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,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,1,1,balanced,0.2166707229614258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,1,1,balanced,0.3154361534118652
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,1,1,balanced,0.503612174987793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,1,1,balanced,0.614532470703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,1,1,balanced,0.6178630447387695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,1,1,balanced,0.6381011199951172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,1,1,balanced,0.6438297271728516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,1,1,balanced,0.6160960006713867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,1,1,balanced,0.6191756820678711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,1,1,balanced,0.6235014343261719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,1,1,balanced,0.6246227264404297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,1,1,balanced,0.6288294219970704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,1,1,balanced,0.6576774597167969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,1,1,balanced,0.8878662109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,1,1,balanced,0.8603584289550781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,1,1,balanced,0.9498156738281249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,1,1,balanced,1.4114028930664062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,1,1,balanced,1.6868928527832032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,1,1,balanced,0.510901107788086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,1,1,balanced,2.5856582641601564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,1,1,balanced,3.368592529296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,1,1,balanced,5.317269897460937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,1,1,balanced,7.549142456054687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,1,1,balanced,12.626966552734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,1,1,balanced,17.556317138671876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,1,1,balanced,27.608151855468748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,1,1,balanced,38.73534423828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,1,1,balanced,0.11595264434814453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,1,1,balanced,0.17772031784057618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,1,1,balanced,0.2728070449829102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,1,1,balanced,0.2767206382751465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,1,1,balanced,0.2926176071166992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,1,1,balanced,0.3086528015136719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,1,1,balanced,0.3021267127990723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,1,1,balanced,0.3120467185974121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,1,1,balanced,0.31654848098754884
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,1,1,balanced,0.29373376846313476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,1,1,balanced,0.3199923133850098
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,1,1,balanced,0.3294246292114258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,1,1,balanced,0.336195182800293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,1,1,balanced,0.4641215896606445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,1,1,balanced,0.49960449218749997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,1,1,balanced,0.2931097602844238
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,1,1,balanced,0.28830528259277344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,1,1,balanced,0.7118931579589843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,1,1,balanced,0.9051564788818359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,1,1,balanced,1.3550559997558593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,1,1,balanced,0.29911615371704103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,1,1,balanced,0.30024448394775394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,1,1,balanced,1.7218669128417967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,1,1,balanced,2.5902438354492188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,1,1,balanced,3.3389715576171874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,1,1,balanced,5.143032836914062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,1,1,balanced,6.710791625976563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,1,1,balanced,0.058556799888610844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,1,1,balanced,0.1590675163269043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,1,1,balanced,0.09025728225708007
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,1,1,balanced,0.16148544311523438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,1,1,balanced,0.15722880363464356
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,1,1,balanced,0.15799936294555664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,1,1,balanced,0.16221952438354492
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,1,1,balanced,0.16400575637817383
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,1,1,balanced,0.16672319412231446
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,1,1,balanced,0.16917055130004882
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,1,1,balanced,0.17164672851562499
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,1,1,balanced,0.17690176010131836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,1,1,balanced,0.16191999435424803
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,1,1,balanced,0.15870976448059082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,1,1,balanced,0.15982848167419433
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,1,1,balanced,0.15921728134155272
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,1,1,balanced,0.15969663619995117
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,1,1,balanced,0.2731475257873535
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,1,1,balanced,0.2872224044799805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,1,1,balanced,0.49877822875976563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,1,1,balanced,0.40263488769531247
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,1,1,balanced,0.7420339202880859
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,1,1,balanced,1.0060620880126954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,1,1,power_law_1.01,0.22336896896362304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,1,1,power_law_1.01,0.3127859115600586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,1,1,balanced,1.4697984313964843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,1,1,power_law_1.01,0.3681062316894531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,1,1,power_law_1.01,0.5087251281738281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,1,1,power_law_1.01,0.6167488098144531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,1,1,power_law_1.01,0.6203756713867188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,1,1,power_law_1.01,0.6233177566528321
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,1,1,power_law_1.01,0.6232966232299805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,1,1,power_law_1.01,0.6323839950561523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,1,1,power_law_1.01,0.6559059143066406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,1,1,power_law_1.01,0.6692530822753906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,1,1,power_law_1.01,0.7149356842041016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,1,1,balanced,1.997151336669922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,1,1,power_law_1.01,0.6229056167602539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,1,1,power_law_1.01,0.7864447784423828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,1,1,power_law_1.01,0.8728524780273437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,1,1,power_law_1.01,1.0206841278076173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,1,1,power_law_1.01,1.2218016052246095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,1,1,power_law_1.01,1.9761695861816406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,1,1,power_law_1.01,1.580067138671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,1,1,power_law_1.01,2.76912841796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,1,1,power_law_1.01,3.5831173706054686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,1,1,power_law_1.01,5.376698608398438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,1,1,power_law_1.01,0.12420479774475097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,1,1,power_law_1.01,0.17378175735473633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,1,1,power_law_1.01,0.19306880950927735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,1,1,power_law_1.01,0.2682732772827149
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,1,1,power_law_1.01,0.2896908760070801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,1,1,power_law_1.01,0.291724796295166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,1,1,power_law_1.01,0.2934304046630859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,1,1,power_law_1.01,7.28867919921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,1,1,power_law_1.01,0.2902169609069824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,1,1,power_law_1.01,0.2977395248413086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,1,1,power_law_1.01,0.2995795249938965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,1,1,power_law_1.01,0.30300672531127926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,1,1,power_law_1.01,0.300254077911377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,1,1,power_law_1.01,0.31117631912231447
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,1,1,power_law_1.01,0.3458911895751953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,1,1,power_law_1.01,0.36111873626708985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,1,1,power_law_1.01,0.37201728820800783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,1,1,power_law_1.01,0.39948928833007813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,1,1,power_law_1.01,0.4813100814819336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,1,1,power_law_1.01,0.585607681274414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,1,1,power_law_1.01,0.7865779113769531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,1,1,power_law_1.01,0.9926080322265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,1,1,power_law_1.01,11.541220703125001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,1,1,power_law_1.01,1.4368748474121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,1,1,power_law_1.01,1.8583168029785155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,1,1,power_law_1.01,2.6264849853515626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,1,1,power_law_1.01,16.04236572265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,1,1,power_law_1.01,0.06118015766143799
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,1,1,power_law_1.01,0.08757568359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,1,1,power_law_1.01,0.11019647598266602
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,1,1,power_law_1.01,0.15498047828674316
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,1,1,power_law_1.01,0.158275203704834
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,1,1,power_law_1.01,0.15871616363525393
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,1,1,power_law_1.01,0.15949695587158202
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,1,1,power_law_1.01,0.1569983959197998
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,1,1,power_law_1.01,0.15776191711425783
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,1,1,power_law_1.01,0.16181888580322265
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,1,1,power_law_1.01,0.16337472915649415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,1,1,power_law_1.01,3.5146060180664067
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,1,1,balanced,2.983489379882813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,1,1,balanced,3.9494943237304687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,1,1,power_law_1.01,0.18721471786499025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,1,1,power_law_1.01,25.54750732421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,1,1,power_law_1.01,5.2852325439453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,1,1,power_law_1.01,0.23537023544311522
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,1,1,power_law_1.01,6.973178100585938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,1,1,power_law_1.01,0.4733465576171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,1,1,power_law_1.01,35.98180908203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,1,1,power_law_1.01,0.1624787139892578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,1,1,power_law_1.01,0.1689567947387695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,1,1,power_law_1.01,0.20714111328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,1,1,power_law_1.01,0.1948179244995117
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,1,1,power_law_1.01,0.30240896224975583
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,1,1,power_law_1.01,0.3555059051513672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,1,1,power_law_1.2,0.22873023986816404
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,1,1,power_law_1.2,0.3056531143188477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,1,1,power_law_1.2,0.3308710479736328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,1,1,power_law_1.2,0.5045971298217774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,1,1,power_law_1.2,0.5928486251831055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,1,1,power_law_1.2,0.6195750427246094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,1,1,power_law_1.2,0.6217580795288086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,1,1,power_law_1.2,0.615035514831543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,1,1,power_law_1.2,0.6270732879638672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,1,1,power_law_1.2,0.6310803222656249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,1,1,power_law_1.2,0.6621465301513672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,1,1,power_law_1.2,0.6877369689941406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,1,1,power_law_1.2,0.7089593505859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,1,1,power_law_1.2,0.7909760284423828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,1,1,power_law_1.2,0.8529452514648437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,1,1,power_law_1.01,0.6007206344604492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,1,1,power_law_1.2,1.0582003021240234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,1,1,power_law_1.2,1.2244895935058593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,1,1,power_law_1.2,1.600984344482422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,1,1,power_law_1.2,1.9911264038085938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,1,1,power_law_1.2,2.783576354980469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,1,1,power_law_1.01,0.915075225830078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,1,1,power_law_1.2,3.554433898925781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,1,1,power_law_1.2,5.331751708984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,1,1,power_law_1.01,1.1157875061035156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,1,1,power_law_1.01,1.714641876220703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,1,1,power_law_1.2,7.358223266601563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,1,1,power_law_1.2,0.1265459156036377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,1,1,power_law_1.2,0.16886720657348633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,1,1,power_law_1.2,0.18084415435791015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,1,1,power_law_1.2,0.2692300796508789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,1,1,power_law_1.2,0.28043264389038086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,1,1,power_law_1.2,0.29125375747680665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,1,1,power_law_1.2,0.29467071533203126
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,1,1,power_law_1.01,2.2204441833496094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,1,1,power_law_1.2,0.28982336044311524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,1,1,power_law_1.2,0.30665599822998046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,1,1,power_law_1.2,0.29878847122192387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,1,1,power_law_1.2,0.3019136047363281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,1,1,power_law_1.2,0.307637767791748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,1,1,power_law_1.2,0.3050009536743164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,1,1,power_law_1.2,0.3456492614746094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,1,1,power_law_1.2,0.3730988693237305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,1,1,power_law_1.2,0.3873881530761719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,1,1,power_law_1.2,0.4008454513549805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,1,1,power_law_1.2,0.47336063385009763
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,1,1,power_law_1.2,0.5806387329101563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,1,1,power_law_1.2,11.757247314453124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,1,1,power_law_1.2,0.7907449340820312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,1,1,power_law_1.01,3.2994644165039064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,1,1,power_law_1.2,0.9873017883300781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,1,1,power_law_1.2,1.4363821411132813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,1,1,power_law_1.2,1.8371910095214843
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,1,1,power_law_1.01,4.334997253417969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,1,1,power_law_1.2,0.06212736129760742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,1,1,power_law_1.2,0.08321279525756836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,1,1,power_law_1.2,0.09566720008850098
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,1,1,power_law_1.2,0.15496000289916992
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,1,1,power_law_1.2,0.15217023849487304
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,1,1,power_law_1.2,0.15866496086120604
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,1,1,power_law_1.2,0.15988927841186523
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,1,1,power_law_1.2,0.15752256393432618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,1,1,power_law_1.2,16.290611572265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,1,1,power_law_1.2,0.1606617546081543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,1,1,power_law_1.2,0.1620319938659668
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,1,1,power_law_1.2,0.1636774444580078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,1,1,power_law_1.2,2.651533508300781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,1,1,power_law_1.2,0.18942848205566407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,1,1,power_law_1.2,0.20055423736572267
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,1,1,power_law_1.2,0.21641216278076172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,1,1,power_law_1.2,3.472105102539062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,1,1,power_law_1.2,5.229480590820312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,1,1,power_law_1.2,25.59571533203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,1,1,power_law_1.2,0.16939008712768555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,1,1,power_law_1.2,0.16543615341186524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,1,1,power_law_1.2,7.0622143554687495
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,1,1,power_law_1.2,0.2379155158996582
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,1,1,power_law_1.2,0.3014476776123047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,1,1,power_law_1.2,36.60083251953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,1,1,power_law_1.2,0.4705311965942383
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,1,1,power_law_1.2,0.35632255554199216
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,1,1,balanced,0.3368473434448242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,1,1,balanced,0.5048614501953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,1,1,balanced,0.8496358489990234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,1,1,balanced,0.8525222778320313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,1,1,power_law_1.2,2.2069549560546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,1,1,balanced,0.9419052886962891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,1,1,balanced,0.9562201690673827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,1,1,balanced,0.9571699523925782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,1,1,balanced,0.9649171447753906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,1,1,balanced,0.9724543762207031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,1,1,balanced,0.9775846099853516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,1,1,balanced,0.9848992156982422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,1,1,balanced,0.9970323181152343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,1,1,balanced,1.0136691284179686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,1,1,balanced,1.0590041351318358
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,1,1,balanced,1.6100250244140626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,1,1,balanced,1.681775360107422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,1,1,balanced,1.7373068237304685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,1,1,balanced,2.5050982666015624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,1,1,balanced,2.995305480957031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,1,1,power_law_1.2,0.6090675354003906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,1,1,balanced,4.524622802734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,1,1,power_law_1.2,0.8813753509521485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,1,1,balanced,5.659319458007813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,1,1,power_law_1.2,1.1371167755126952
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,1,1,balanced,9.210126953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,1,1,power_law_1.2,1.7207128906249998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,1,1,balanced,13.233160400390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,1,1,balanced,0.17555200576782226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,1,1,balanced,0.25978431701660154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,1,1,balanced,0.45029376983642583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,1,1,balanced,0.4517926406860352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,1,1,balanced,0.47727359771728517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,1,1,balanced,0.4779935836791992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,1,1,balanced,0.4818624114990234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,1,1,balanced,0.4899296188354493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,1,1,balanced,0.4881491088867188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,1,1,balanced,0.48860416412353513
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,1,1,balanced,0.4935756683349609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,1,1,balanced,0.4988812637329102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,1,1,balanced,0.500456314086914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,1,1,balanced,0.5103839874267578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,1,1,balanced,0.526163215637207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,1,1,balanced,0.5352659225463867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,1,1,balanced,0.545599365234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,1,1,balanced,0.7651052856445313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,1,1,balanced,0.8358329772949219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,1,1,balanced,1.1229305267333984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,1,1,balanced,1.4710707092285156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,1,1,balanced,2.126445465087891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,1,1,balanced,21.73719482421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,1,1,balanced,2.7569696044921876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,1,1,balanced,4.1070751953124995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,1,1,balanced,5.32185302734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,1,1,balanced,8.171714477539062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,1,1,balanced,30.0182373046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,1,1,balanced,10.696917724609374
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,1,1,power_law_1.2,3.342798767089844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,1,1,balanced,48.29123046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,1,1,power_law_1.2,4.3453784179687505
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,1,1,balanced,0.08108351707458497
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,1,1,balanced,0.24941247940063477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,1,1,balanced,68.0464501953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,1,1,balanced,0.2512928009033203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,1,1,balanced,0.2530041694641113
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,1,1,balanced,0.2558860778808594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,1,1,balanced,0.2563423919677734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,1,1,balanced,0.2618604850769043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,1,1,balanced,0.14286016464233398
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,1,1,balanced,0.26490560531616214
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,1,1,balanced,0.26962303161621093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,1,1,balanced,0.2739628791809082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,1,1,balanced,0.24844991683959963
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,1,1,balanced,0.24722240447998045
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,1,1,balanced,0.24807872772216796
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,1,1,balanced,0.2495961570739746
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,1,1,balanced,0.24754240036010744
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,1,1,balanced,0.2514559936523438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,1,1,balanced,0.48411647796630863
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,1,1,balanced,0.46586559295654295
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,1,1,balanced,0.662616958618164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,1,1,power_law_1.01,0.33835582733154296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,1,1,power_law_1.01,0.48687488555908204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,1,1,power_law_1.01,0.5503398513793945
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,1,1,balanced,0.8879289245605468
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,1,1,balanced,1.3335224914550783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,1,1,power_law_1.01,0.842888946533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,1,1,power_law_1.01,0.9509388732910157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,1,1,power_law_1.01,0.9618476867675781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,1,1,power_law_1.01,0.9599705505371094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,1,1,power_law_1.01,0.9688159942626953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,1,1,power_law_1.01,0.9791622161865234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,1,1,power_law_1.01,1.008951644897461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,1,1,power_law_1.01,1.145187225341797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,1,1,power_law_1.01,1.2053497314453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,1,1,power_law_1.01,1.3468453979492188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,1,1,power_law_1.01,1.4593043518066406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,1,1,power_law_1.01,1.615320281982422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,1,1,power_law_1.01,1.74478271484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,1,1,power_law_1.01,2.091205749511719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,1,1,power_law_1.01,2.7665875244140627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,1,1,power_law_1.01,3.3903314208984376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,1,1,power_law_1.01,4.8604321289062495
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,1,1,balanced,1.7679653930664063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,1,1,power_law_1.01,6.333517456054688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,1,1,power_law_1.01,9.189592895507811
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,1,1,balanced,2.5695513916015624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,1,1,power_law_1.01,12.413107910156251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,1,1,power_law_1.01,0.16878528594970704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,1,1,power_law_1.01,0.24597248077392578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,1,1,power_law_1.01,0.2910560035705566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,1,1,power_law_1.01,0.4464448165893555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,1,1,power_law_1.01,0.47472320556640624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,1,1,power_law_1.01,0.48268993377685543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,1,1,power_law_1.01,0.4826681518554688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,1,1,power_law_1.01,0.4811065673828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,1,1,power_law_1.01,0.48071166992187503
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,1,1,power_law_1.01,0.4897151947021484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,1,1,power_law_1.01,0.49402111053466796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,1,1,power_law_1.01,0.49339649200439456
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,1,1,power_law_1.01,0.5052204895019531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,1,1,power_law_1.01,0.5314310455322266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,1,1,power_law_1.01,0.5317516708374024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,1,1,power_law_1.01,0.5649203109741211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,1,1,power_law_1.01,19.977830810546873
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,1,1,balanced,3.31234619140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,1,1,power_law_1.01,0.6325740814208984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,1,1,power_law_1.01,0.7944153594970703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,1,1,power_law_1.01,0.9631353759765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,1,1,balanced,4.989701232910156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,1,1,power_law_1.01,1.2749779510498047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,1,1,power_law_1.01,1.645562286376953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,1,1,balanced,6.7278369140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,1,1,power_law_1.01,2.2625106811523437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,1,1,power_law_1.01,2.8800115966796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,1,1,power_law_1.01,0.09013567924499512
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,1,1,power_law_1.01,0.1353388786315918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,1,1,power_law_1.01,0.16002687454223632
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,1,1,power_law_1.01,0.24444864273071287
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,1,1,power_law_1.01,0.2488038444519043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,1,1,power_law_1.01,0.24897727966308594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,1,1,power_law_1.01,27.837946777343753
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,1,1,power_law_1.01,0.25062591552734376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,1,1,power_law_1.01,0.24687807083129884
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,1,1,power_law_1.01,0.24809984207153318
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,1,1,power_law_1.01,4.182837829589844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,1,1,power_law_1.01,0.25420799255371096
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,1,1,power_law_1.01,0.25548799514770504
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,1,1,power_law_1.01,0.30581504821777344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,1,1,power_law_1.01,5.559483642578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,1,1,power_law_1.01,8.274191894531251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,1,1,power_law_1.01,10.962708740234374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,1,1,power_law_1.01,43.95783203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,1,1,power_law_1.01,0.25397247314453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,1,1,power_law_1.01,0.31679296493530273
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,1,1,power_law_1.01,0.268159351348877
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,1,1,power_law_1.01,0.3394112014770508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,1,1,power_law_1.01,62.610546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,1,1,power_law_1.01,0.39005950927734373
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,1,1,power_law_1.01,0.5149043273925782
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,1,1,power_law_1.01,0.6222342300415039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,1,1,power_law_1.01,0.8370220947265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,1,1,power_law_1.2,0.3441158294677734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,1,1,power_law_1.2,0.4631315231323242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,1,1,power_law_1.01,1.0495378875732422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,1,1,power_law_1.2,0.5361356735229492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,1,1,power_law_1.2,0.8394322967529296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,1,1,power_law_1.2,0.9170873260498047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,1,1,power_law_1.2,0.955578842163086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,1,1,power_law_1.01,1.4985836791992189
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,1,1,power_law_1.2,0.9606470489501954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,1,1,power_law_1.2,0.964748764038086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,1,1,power_law_1.2,1.0219750213623047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,1,1,power_law_1.2,1.0193043518066407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,1,1,power_law_1.2,1.2029497528076172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,1,1,power_law_1.2,1.2453196716308592
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,1,1,power_law_1.2,1.3024493408203124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,1,1,power_law_1.2,1.4456576538085937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,1,1,power_law_1.01,1.967587890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,1,1,power_law_1.2,1.5771763610839844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,1,1,power_law_1.2,1.7200498962402342
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,1,1,power_law_1.2,2.0456723022460936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,1,1,power_law_1.2,2.73673095703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,1,1,power_law_1.2,3.3717471313476564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,1,1,power_law_1.2,4.823176879882813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,1,1,power_law_1.01,2.74352783203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,1,1,power_law_1.2,6.278477783203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,1,1,power_law_1.01,3.5875732421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,1,1,power_law_1.2,9.344408569335936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,1,1,power_law_1.2,0.17201856613159178
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,1,1,power_law_1.2,0.239050235748291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,1,1,power_law_1.2,0.2655052757263184
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,1,1,power_law_1.2,0.4453023910522461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,1,1,power_law_1.2,0.46593345642089845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,1,1,power_law_1.2,0.4837516784667969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,1,1,power_law_1.2,0.48439361572265627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,1,1,power_law_1.2,0.4821510314941406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,1,1,power_law_1.2,0.488647689819336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,1,1,power_law_1.2,0.48883518218994143
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,1,1,power_law_1.2,0.494060173034668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,1,1,power_law_1.2,12.625550537109374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,1,1,power_law_1.2,0.49668800354003906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,1,1,power_law_1.2,0.4984415817260742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,1,1,power_law_1.2,0.5264575958251954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,1,1,power_law_1.2,0.5399692916870118
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,1,1,power_law_1.2,0.5806272125244141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,1,1,power_law_1.2,0.6509728240966797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,1,1,power_law_1.2,0.7835244750976562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,1,1,power_law_1.2,0.9564685058593749
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,1,1,power_law_1.01,5.327532958984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,1,1,power_law_1.2,1.2743462371826173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,1,1,power_law_1.2,1.6243820190429688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,1,1,power_law_1.2,2.263368377685547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,1,1,power_law_1.2,20.017362060546876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,1,1,power_law_1.2,2.89649658203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,1,1,power_law_1.01,7.053709716796876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,1,1,power_law_1.2,0.091146240234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,1,1,power_law_1.2,0.1298419189453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,1,1,power_law_1.2,0.14578495979309083
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,1,1,power_law_1.2,0.2454662322998047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,1,1,power_law_1.2,0.24165376663208007
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,1,1,power_law_1.2,0.2504870414733887
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,1,1,power_law_1.2,0.2516262435913086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,1,1,power_law_1.2,0.2482579231262207
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,1,1,power_law_1.2,0.25274431228637695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,1,1,power_law_1.2,0.2557043266296387
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,1,1,power_law_1.2,0.25663808822631834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,1,1,power_law_1.2,4.218179931640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,1,1,power_law_1.2,0.277259521484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,1,1,power_law_1.2,0.3042745590209961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,1,1,power_law_1.2,28.14002685546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,1,1,power_law_1.2,5.611912231445312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,1,1,power_law_1.2,8.281248779296876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,1,1,power_law_1.2,45.4371728515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,1,1,power_law_1.2,10.94812744140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,1,1,power_law_1.2,0.25900928497314457
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,1,1,power_law_1.2,0.3252096176147461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,1,1,power_law_1.2,0.3498796844482422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,1,1,power_law_1.2,1.5091290283203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,1,1,power_law_1.2,0.4061337661743164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,1,1,power_law_1.2,63.976923828124995
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,1,1,power_law_1.2,0.5177503967285156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,1,1,power_law_1.2,0.6144691085815429
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,1,1,power_law_1.2,0.8529241943359376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,1,1,power_law_1.2,1.0660768127441407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,1,balanced,0.15471039772033693
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,1,balanced,0.30179008483886716
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,1,balanced,0.5349184036254883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,1,balanced,0.9576774597167969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,1,1,power_law_1.2,1.9397760009765626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,1,balanced,1.8769535827636719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,1,balanced,3.7472320556640626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,1,balanced,3.747958984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,1,balanced,3.7395526123046876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,1,balanced,3.86741943359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,1,balanced,3.7723751831054684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,1,balanced,3.8451513671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,1,1,power_law_1.2,2.829395751953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,1,balanced,3.895339660644531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,1,balanced,3.786639404296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,1,balanced,3.988218994140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,1,balanced,4.0599667358398435
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,1,1,power_law_1.2,3.655242309570313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,1,balanced,4.096431274414063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,1,balanced,4.119120788574219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,1,balanced,4.186981811523437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,1,balanced,4.2999270629882815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,1,balanced,4.528385925292969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,1,balanced,4.808871765136719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,1,balanced,7.406639404296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,1,balanced,7.843748779296876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,1,balanced,11.2230224609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,1,balanced,0.09402943611145019
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,1,balanced,14.963756103515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,1,balanced,0.2763180732727051
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,1,balanced,0.15777215957641602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,1,balanced,0.4902617645263672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,1,balanced,1.875684509277344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,1,balanced,0.9450188446044923
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,1,balanced,24.07965087890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,1,1,power_law_1.2,5.4858984374999995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,1,balanced,1.8848844909667968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,1,balanced,1.888135681152344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,1,balanced,1.88500732421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,1,balanced,1.941009216308594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,1,balanced,1.93644287109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,1,balanced,1.8925625610351564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,1,balanced,1.9085894775390624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,1,balanced,1.8967385864257813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,1,balanced,35.20619384765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,1,1,power_law_1.2,7.012777709960938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,1,balanced,2.1001434326171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,1,balanced,2.130260467529297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,1,balanced,2.9349588012695316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,1,balanced,2.1445298767089844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,1,balanced,2.2038348388671873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,1,balanced,2.2492652893066407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,1,balanced,2.608649597167969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,1,balanced,2.326153564453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,1,balanced,4.209439392089844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,1,balanced,2.4624351501464843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,1,balanced,4.557811889648438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,1,balanced,6.390443115234374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,1,balanced,7.961442260742187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,1,1,balanced,0.08352640151977539
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,1,1,balanced,0.11246015548706054
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,1,1,balanced,0.17032127380371093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,1,1,balanced,0.2998284721374512
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,1,1,balanced,1.1386930847167969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,1,1,balanced,1.1347238159179687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,1,1,balanced,1.138563232421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,1,1,balanced,0.5693459320068359
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,1,1,balanced,1.138461456298828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,1,1,balanced,1.1265331268310548
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,1,1,balanced,1.142053756713867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,1,1,balanced,1.1500768280029297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,1,1,balanced,1.155849609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,1,1,balanced,1.1642189025878906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,1,1,balanced,1.1797068786621092
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,1,1,balanced,1.1272537231445312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,1,1,balanced,1.1953855895996095
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,1,1,balanced,1.2230508422851563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,1,1,balanced,1.273166046142578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,1,1,balanced,1.3013235473632814
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,1,1,balanced,1.4426797485351561
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,1,1,balanced,1.4947999572753905
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,1,1,balanced,1.143804168701172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,1,power_law_1.01,0.5337855911254883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,1,power_law_1.01,0.957193603515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,1,power_law_1.01,1.4601600646972657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,1,power_law_1.01,1.9075366210937499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,1,power_law_1.01,0.153187837600708
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,1,power_law_1.01,2.7561056518554685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,1,power_law_1.01,0.31065088272094726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,1,power_law_1.01,2.9478353881835937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,1,power_law_1.01,2.9135321044921874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,1,power_law_1.01,3.0135162353515623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,1,power_law_1.01,3.478934326171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,1,power_law_1.01,3.5895513916015624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,1,power_law_1.01,3.172854309082031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,1,power_law_1.01,3.917386169433594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,1,power_law_1.01,3.602569580078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,1,power_law_1.01,3.6457748413085938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,1,power_law_1.01,3.8813977050781245
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,1,power_law_1.01,4.251034240722656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,1,power_law_1.01,4.655491943359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,1,power_law_1.01,5.365815429687499
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,1,1,balanced,2.37219970703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,1,1,balanced,2.4949708557128907
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,1,1,balanced,3.700435791015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,1,power_law_1.01,6.20005615234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,1,1,balanced,4.793498840332031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,1,power_law_1.01,7.784539794921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,1,power_law_1.01,0.2758515167236328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,1,power_law_1.01,0.49147392272949214
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,1,power_law_1.01,9.405234375000001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,1,power_law_1.01,0.7434893035888672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,1,power_law_1.01,0.9925567626953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,1,power_law_1.01,1.3673548889160156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,1,power_law_1.01,0.15653632164001466
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,1,power_law_1.01,1.457342071533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,1,power_law_1.01,12.603273925781249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,1,power_law_1.01,1.4781356811523438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,1,power_law_1.01,1.5770918273925782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,1,power_law_1.01,1.6247302246093749
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,1,power_law_1.01,1.6435212707519533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,1,power_law_1.01,1.6716748046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,1,power_law_1.01,15.905111083984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,1,power_law_1.01,1.871337585449219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,1,power_law_1.01,0.09472064018249512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,1,power_law_1.01,1.8802764892578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,1,power_law_1.01,1.9522642517089843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,1,power_law_1.01,2.0501177978515623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,1,power_law_1.01,1.4923143005371093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,1,power_law_1.01,2.1650489807128905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,1,power_law_1.01,2.356669464111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,1,power_law_1.01,2.648478698730469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,1,power_law_1.01,2.97161865234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,1,power_law_1.01,22.11152587890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,1,power_law_1.01,3.412076110839844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,1,1,power_law_1.01,0.09191679954528809
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,1,1,power_law_1.01,0.12611007690429688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,1,power_law_1.01,4.263786926269531
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,1,1,power_law_1.01,0.1810220718383789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,1,1,power_law_1.01,0.3028563117980957
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,1,1,power_law_1.01,0.4469574356079102
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,1,1,power_law_1.01,0.570376968383789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,1,1,power_law_1.01,0.8119551849365234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,1,power_law_1.01,5.121564331054687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,1,1,power_law_1.01,0.8591372680664062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,1,1,power_law_1.01,0.8793804931640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,1,1,power_law_1.01,0.8878758239746094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,1,power_law_1.01,7.049346313476562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,1,1,power_law_1.01,0.9459705352783203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,1,1,power_law_1.01,0.97587646484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,1,1,power_law_1.01,1.005631332397461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,1,1,power_law_1.01,0.9805868530273438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,1,power_law_1.01,8.82100830078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,1,power_law_1.01,28.580263671875002
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,1,1,power_law_1.01,1.0421075439453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,1,1,power_law_1.01,1.0756633758544922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,1,1,power_law_1.01,1.1177894592285156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,1,1,power_law_1.01,1.3181517028808594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,1,1,power_law_1.01,1.193984603881836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,1,1,power_law_1.01,1.5000352478027343
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,1,1,power_law_1.01,1.879613494873047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,1,1,power_law_1.01,2.0547865295410155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,1,power_law_1.2,0.15618240356445312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,1,power_law_1.2,0.30897663116455076
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,1,1,power_law_1.01,2.479722900390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,1,power_law_1.2,1.3281452941894532
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,1,1,power_law_1.01,3.3175149536132813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,1,1,power_law_1.01,4.124783325195312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,1,power_law_1.2,2.697049560546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,1,power_law_1.2,2.8353146362304686
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,1,1,power_law_1.01,5.726314697265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,1,power_law_1.2,0.5297798538208008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,1,power_law_1.2,3.0623181152343752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,1,power_law_1.2,3.2759750366210936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,1,power_law_1.2,3.4121893310546874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,1,power_law_1.2,2.8771813964843753
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,1,power_law_1.2,0.8833792114257812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,1,power_law_1.2,4.116986389160156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,1,power_law_1.2,3.684901733398438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,1,power_law_1.2,1.7509266662597658
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,1,power_law_1.2,3.7928955078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,1,power_law_1.2,3.6014529418945314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,1,power_law_1.2,4.001685791015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,1,power_law_1.2,4.433914794921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,1,power_law_1.2,4.889102172851563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,1,power_law_1.2,5.750504150390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,1,power_law_1.2,8.023991699218751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,1,power_law_1.2,9.664393310546874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,1,power_law_1.2,6.335240478515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,1,power_law_1.2,0.2754463958740234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,1,power_law_1.2,0.460134391784668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,1,power_law_1.2,0.6773388671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,1,power_law_1.2,12.841412353515626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,1,1,power_law_1.01,7.4960498046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,1,power_law_1.2,0.9326258850097655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,1,power_law_1.2,1.3450854492187498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,1,power_law_1.2,1.4466432189941405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,1,power_law_1.2,1.4119091796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,1,power_law_1.2,15.896612548828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,1,power_law_1.2,0.09603391647338867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,1,power_law_1.2,1.5038648986816407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,1,power_law_1.2,1.5568722534179689
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,1,power_law_1.2,1.5950579833984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,1,power_law_1.2,1.649956512451172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,1,power_law_1.2,1.690985565185547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,1,power_law_1.2,1.9150936889648438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,1,power_law_1.2,1.9450022888183593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,1,power_law_1.2,1.9950079345703124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,1,power_law_1.2,2.1468333435058593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,1,power_law_1.2,0.15798784255981446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,1,power_law_1.2,2.2386483764648437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,1,power_law_1.2,2.5050151062011716
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,1,power_law_1.2,2.6686355590820314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,1,power_law_1.2,3.0996978759765623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,1,power_law_1.2,22.57955810546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,1,1,power_law_1.2,0.09189184188842774
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,1,1,power_law_1.2,0.125282564163208
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,1,power_law_1.2,3.545761413574219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,1,1,power_law_1.2,0.1783308792114258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,1,power_law_1.2,4.408948364257812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,1,1,power_law_1.2,0.2832678413391113
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,1,1,power_law_1.2,0.4084352111816406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,1,power_law_1.2,5.373815307617187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,1,1,power_law_1.2,0.5369939041137696
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,1,1,power_law_1.2,0.7983494567871093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,1,1,power_law_1.2,0.8405043029785156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,1,1,power_law_1.2,0.8671469116210938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,1,power_law_1.2,7.279332275390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,1,1,power_law_1.2,0.8997612762451173
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,1,1,power_law_1.2,0.9310880279541015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,1,1,power_law_1.2,0.9664249420166016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,1,1,power_law_1.2,0.9914150238037109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,1,power_law_1.2,29.19541259765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,1,1,power_law_1.2,1.0373043060302733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,1,power_law_1.2,8.950413208007813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,1,1,power_law_1.2,1.0831231689453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,1,1,power_law_1.2,1.1070655822753905
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,1,1,power_law_1.2,1.4168089294433595
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,1,1,power_law_1.2,1.1709798431396483
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,1,1,power_law_1.2,1.3246202087402343
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,1,1,power_law_1.2,1.7289260864257812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,1,1,power_law_1.2,1.9274150085449218
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,1,1,power_law_1.2,2.4940077209472653
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,1,1,power_law_1.2,3.09576904296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,1,1,power_law_1.2,4.31889404296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,1,balanced,0.27845823287963867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,1,balanced,0.13200639724731444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,1,balanced,0.466363525390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,1,1,power_law_1.2,5.658638916015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,1,balanced,3.2019921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,1,balanced,3.2493185424804687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,1,balanced,1.5934573364257811
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,1,balanced,3.2124499511718754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,1,balanced,3.2116433715820314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,1,balanced,3.3346533203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,1,balanced,0.8177267456054688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,1,balanced,3.3773867797851564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,1,balanced,3.3736569213867185
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,1,balanced,3.357969970703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,1,balanced,3.387469482421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,1,balanced,3.5049069213867186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,1,balanced,3.5156204223632814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,1,balanced,3.5745959472656255
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,1,balanced,3.623661499023437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,1,balanced,3.708661804199219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,1,balanced,3.902739868164063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,1,balanced,4.175812988281249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,1,balanced,6.305714721679688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,1,balanced,6.753640136718749
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,1,balanced,9.802548217773438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,1,balanced,0.24174079895019532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,1,balanced,0.42198017120361325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,1,balanced,0.8103692626953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,1,balanced,12.783966064453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,1,balanced,1.6174105834960937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,1,balanced,1.6319987487792968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,1,balanced,0.08778176307678223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,1,balanced,1.6340185546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,1,balanced,1.6387103271484373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,1,balanced,0.13734527587890627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,1,balanced,1.6145651245117185
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,1,balanced,1.6356973266601564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,1,balanced,1.651561584472656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,1,balanced,19.997344970703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,1,balanced,1.6711776733398438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,1,balanced,1.8382566833496092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,1,balanced,1.8547628784179686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,1,balanced,1.8994400024414062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,1,balanced,1.9344761657714844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,1,balanced,1.9409735107421873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,1,balanced,2.042580413818359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,1,balanced,2.122216949462891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,1,balanced,1.658239288330078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,1,balanced,2.3477183532714845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,1,balanced,27.32332763671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,1,balanced,2.5332415771484373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,1,balanced,3.708836975097656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,1,balanced,3.9331137084960934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,1,balanced,5.646086425781251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,1,balanced,7.030709838867187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,1,1,balanced,0.07915455818176269
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,1,1,balanced,0.10355392456054688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,1,1,balanced,0.151592960357666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,1,1,balanced,0.9604972839355469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,1,1,balanced,0.25915391921997066
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,1,1,balanced,0.9575052642822266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,1,1,balanced,0.960549087524414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,1,1,balanced,0.9659597015380859
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,1,1,balanced,0.9637222290039062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,1,1,balanced,0.4890822219848633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,1,1,balanced,0.9640262603759766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,1,1,balanced,0.9731289672851563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,1,1,balanced,0.9800953674316407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,1,1,balanced,0.9775052642822265
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,1,1,balanced,0.9885369873046874
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,1,1,balanced,0.9600198364257813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,1,1,balanced,0.9966207885742188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,1,1,balanced,1.0263750457763672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,1,1,balanced,1.0400953674316407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,1,1,balanced,1.0689933013916015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,1,1,balanced,1.119375991821289
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,1,1,balanced,1.2212576293945312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,1,1,balanced,1.3033792114257812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,1,power_law_1.01,0.26768640518188475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,1,power_law_1.01,0.12904640197753908
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,1,power_law_1.01,0.46115390777587895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,1,power_law_1.01,0.8261055755615235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,1,power_law_1.01,1.2445855712890626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,1,power_law_1.01,1.6150419616699216
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,1,power_law_1.01,2.3246424865722655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,1,power_law_1.01,2.518901824951172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,1,power_law_1.01,2.569926452636719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,1,power_law_1.01,2.596337890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,1,power_law_1.01,2.752732849121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,1,power_law_1.01,2.976856994628906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,1,power_law_1.01,3.063916931152344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,1,1,balanced,2.096910095214844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,1,power_law_1.01,3.3763659667968753
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,1,1,balanced,2.254281005859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,1,power_law_1.01,3.0701889038085937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,1,1,balanced,3.210043640136719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,1,power_law_1.01,3.1033114624023437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,1,power_law_1.01,3.3018008422851564
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,1,1,balanced,4.207666625976563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,1,power_law_1.01,3.595749816894531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,1,power_law_1.01,4.701868896484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,1,power_law_1.01,6.729934082031249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,1,power_law_1.01,0.08918911933898925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,1,power_law_1.01,3.921770324707031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,1,power_law_1.01,8.058909912109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,1,power_law_1.01,0.13797311782836913
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,1,power_law_1.01,0.4272806549072266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,1,power_law_1.01,0.6443321228027343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,1,power_law_1.01,5.346630249023438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,1,power_law_1.01,0.8543014526367188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,1,power_law_1.01,10.854466552734376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,1,power_law_1.01,1.1845164489746094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,1,power_law_1.01,1.246432647705078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,1,power_law_1.01,1.2715731048583985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,1,power_law_1.01,1.288656005859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,1,power_law_1.01,1.3547077941894532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,1,power_law_1.01,1.3856723022460937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,1,power_law_1.01,0.23969728469848633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,1,power_law_1.01,13.652396240234376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,1,power_law_1.01,1.4074604797363282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,1,power_law_1.01,1.4303359985351562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,1,power_law_1.01,1.6229376220703124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,1,power_law_1.01,1.6250674438476562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,1,power_law_1.01,1.6761016845703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,1,power_law_1.01,1.76998779296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,1,power_law_1.01,1.8707955932617188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,1,power_law_1.01,2.045224304199219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,1,power_law_1.01,2.2558796691894534
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,1,power_law_1.01,2.607646789550781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,1,power_law_1.01,3.0064556884765627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,1,power_law_1.01,19.062359619140626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,1,power_law_1.01,3.8068966674804683
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,1,power_law_1.01,4.566054382324219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,1,1,power_law_1.01,0.16167552947998048
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,1,1,power_law_1.01,0.2638790321350098
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,1,1,power_law_1.01,0.385557746887207
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,1,1,power_law_1.01,0.4911967849731445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,1,power_law_1.01,6.110464477539063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,1,1,power_law_1.01,0.6947328186035155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,1,power_law_1.01,7.695522460937499
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,1,1,power_law_1.01,0.7320390319824218
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,1,1,power_law_1.01,0.7573849487304687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,1,power_law_1.01,24.349416503906248
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,1,1,power_law_1.01,0.7462016296386719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,1,1,power_law_1.01,0.7994265747070313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,1,1,power_law_1.01,0.8268473815917968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,1,1,power_law_1.01,0.835607681274414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,1,1,power_law_1.01,0.8662681579589844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,1,1,power_law_1.01,0.8958048248291016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,1,1,power_law_1.01,0.9031948852539063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,1,1,power_law_1.01,0.9580000305175782
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,1,1,power_law_1.01,1.13970947265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,1,1,power_law_1.01,0.08563455581665039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,1,1,power_law_1.01,0.11554495811462402
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,1,1,power_law_1.01,1.0276908874511719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,1,1,power_law_1.01,1.2974751281738282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,1,power_law_1.2,0.13080960273742676
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,1,1,power_law_1.01,1.458406982421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,1,1,power_law_1.01,1.8581330871582034
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,1,power_law_1.2,0.2677964782714844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,1,1,power_law_1.01,2.159055938720703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,1,power_law_1.2,0.4643891143798828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,1,power_law_1.2,0.7558060455322265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,1,power_law_1.2,1.1304326629638672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,1,1,power_law_1.01,2.89421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,1,power_law_1.2,1.519812469482422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,1,power_law_1.2,2.260810241699219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,1,power_law_1.2,2.390184326171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,1,power_law_1.2,2.4910624694824217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,1,power_law_1.2,2.5806124877929686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,1,power_law_1.2,2.792236938476562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,1,power_law_1.2,2.9867129516601563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,1,power_law_1.2,3.1787078857421873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,1,power_law_1.2,3.501068115234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,1,power_law_1.2,3.1479559326171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,1,power_law_1.2,3.2797747802734376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,1,power_law_1.2,3.4199719238281245
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,1,power_law_1.2,3.807709350585937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,1,power_law_1.2,4.122917785644531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,1,power_law_1.2,4.922228393554688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,1,power_law_1.2,5.494361572265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,1,1,power_law_1.01,3.6314041137695314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,1,power_law_1.2,6.890721435546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,1,1,power_law_1.01,5.000983581542969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,1,power_law_1.2,8.327605590820312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,1,power_law_1.2,0.24419776916503907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,1,power_law_1.2,0.3982067108154297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,1,power_law_1.2,0.5825260925292969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,1,power_law_1.2,10.874039306640626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,1,power_law_1.2,0.8028550720214843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,1,power_law_1.2,1.1627526092529297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,1,power_law_1.2,0.14118080139160155
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,1,1,power_law_1.01,6.568982543945313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,1,power_law_1.2,1.2222470092773439
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,1,power_law_1.2,1.2634854125976562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,1,power_law_1.2,0.08693632125854492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,1,power_law_1.2,13.7933544921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,1,power_law_1.2,1.2900679016113281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,1,power_law_1.2,1.3300262451171876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,1,power_law_1.2,1.373832244873047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,1,power_law_1.2,1.4495802307128907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,1,power_law_1.2,1.4130694580078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,1,power_law_1.2,1.6481146240234374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,1,power_law_1.2,1.6681651306152343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,1,power_law_1.2,1.724610595703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,1,power_law_1.2,1.8506501770019532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,1,power_law_1.2,1.9240646362304688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,1,power_law_1.2,2.178052520751953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,1,power_law_1.2,2.348957366943359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,1,power_law_1.2,2.7016915893554687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,1,power_law_1.2,19.097686767578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,1,power_law_1.2,3.1419961547851565
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,1,1,power_law_1.2,0.08589823722839356
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,1,1,power_law_1.2,0.11525888442993164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,1,1,power_law_1.2,0.1593388843536377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,1,power_law_1.2,3.8609100341796876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,1,1,power_law_1.2,0.24657920837402342
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,1,1,power_law_1.2,0.3521171188354492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,1,power_law_1.2,4.682935791015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,1,1,power_law_1.2,0.46216449737548826
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,1,1,power_law_1.2,0.6830623626708985
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,1,1,power_law_1.2,0.7196249389648438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,1,1,power_law_1.2,0.7357714843750001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,1,power_law_1.2,6.191864624023437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,1,1,power_law_1.2,0.76906494140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,1,1,power_law_1.2,0.7956377410888672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,1,1,power_law_1.2,0.8207507324218749
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,1,1,power_law_1.2,0.8500985717773437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,1,1,power_law_1.2,0.8826348876953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,1,1,power_law_1.2,0.9198111724853515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,1,power_law_1.2,7.829706420898438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,1,power_law_1.2,24.77604736328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,1,1,power_law_1.2,0.9510975646972657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,1,1,power_law_1.2,1.0153907012939452
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,1,1,power_law_1.2,1.1339142608642578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,1,1,power_law_1.2,1.2309222412109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,1,1,power_law_1.2,1.4757901000976563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,1,1,power_law_1.2,2.09716796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,1,balanced,0.054494080543518064
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,1,1,power_law_1.2,2.1706040954589847
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,1,balanced,0.07905792236328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,1,balanced,0.13263360023498535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,1,balanced,0.22671936035156248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,1,balanced,0.23046079635620118
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,1,balanced,0.2276326370239258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,1,balanced,0.23186431884765626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,1,balanced,0.23250047683715822
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,1,1,power_law_1.2,2.7130804443359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,1,balanced,0.23312383651733398
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,1,1,power_law_1.2,3.8023431396484377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,1,balanced,0.24903680801391603
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,1,balanced,0.24614400863647462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,1,balanced,0.25497983932495116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,1,balanced,0.25419584274291995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,1,balanced,0.24992383956909178
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,1,balanced,0.26237695693969726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,1,balanced,0.2754560089111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,1,balanced,0.2994534492492676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,1,balanced,0.40993793487548824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,1,balanced,0.041047677993774415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,1,balanced,0.4564147186279297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,1,balanced,0.2340236854553223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,1,balanced,0.6148083114624023
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,1,1,power_law_1.2,5.048609313964844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,1,balanced,0.8092505645751953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,1,balanced,0.044038400650024415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,1,balanced,0.06648640155792236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,1,balanced,1.1965113830566405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,1,balanced,0.03377856016159057
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,1,balanced,0.09756928443908691
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,1,balanced,0.15737855911254883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,1,balanced,0.15879232406616212
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,1,balanced,1.6043218994140624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,1,balanced,0.1583852767944336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,1,balanced,0.17115903854370118
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,1,balanced,0.1596876811981201
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,1,balanced,0.16094720840454102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,1,balanced,0.17726079940795897
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,1,balanced,0.1810086441040039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,1,balanced,0.18175296783447265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,1,balanced,2.3432147216796873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,1,balanced,0.1820140838623047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,1,balanced,0.18489919662475587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,1,balanced,0.19379392623901368
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,1,balanced,0.19881088256835938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,1,balanced,0.20731712341308595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,1,balanced,0.22903743743896485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,1,balanced,0.24274623870849607
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,1,balanced,3.1276953124999998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,1,balanced,0.3587596893310547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,1,balanced,0.4000236892700196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,1,balanced,0.5618092727661133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,1,balanced,0.7527206420898438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,1,balanced,1.0469254302978517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,1,balanced,1.381964874267578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,1,balanced,0.15593983650207519
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,1,1,balanced,0.03651135921478271
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,1,1,balanced,0.04062655925750732
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,1,1,balanced,0.047374720573425295
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,1,1,balanced,0.09301119804382324
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,1,1,balanced,0.09284223556518555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,1,1,balanced,0.09392191886901854
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,1,1,balanced,0.09374208450317383
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,1,1,balanced,0.09243776321411132
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,1,1,balanced,0.09389823913574219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,1,1,balanced,0.09492032051086426
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,1,1,balanced,0.09635647773742675
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,1,1,balanced,0.09776896476745606
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,1,1,balanced,0.09821375846862793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,1,1,balanced,0.09974847793579102
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,1,1,balanced,0.10322815895080566
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,1,1,balanced,0.10770175933837892
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,1,1,balanced,0.11335295677185057
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,1,1,balanced,0.12408448219299317
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,1,1,balanced,0.13335616111755372
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,1,1,balanced,0.06112639904022217
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,1,1,balanced,0.09213952064514161
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,1,1,balanced,0.21345279693603514
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,1,1,balanced,0.19188863754272462
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,1,1,balanced,0.2991993522644043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,1,power_law_1.01,0.05424448013305664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,1,1,balanced,0.3838943862915039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,1,power_law_1.01,0.07884416103363037
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,1,power_law_1.01,0.11800191879272462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,1,power_law_1.01,0.13439552307128905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,1,power_law_1.01,0.18969280242919923
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,1,1,balanced,0.5800032043457032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,1,power_law_1.01,0.20236799240112308
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,1,power_law_1.01,0.20773632049560548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,1,power_law_1.01,0.2170342445373535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,1,power_law_1.01,0.22669311523437502
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,1,power_law_1.01,0.23733503341674805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,1,power_law_1.01,0.24170175552368164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,1,power_law_1.01,0.2430835151672363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,1,power_law_1.01,0.25470848083496095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,1,power_law_1.01,0.2701132774353027
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,1,power_law_1.01,0.27526336669921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,1,power_law_1.01,0.28709375381469726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,1,power_law_1.01,0.3205875015258789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,1,power_law_1.01,0.3538463973999023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,1,power_law_1.01,0.4319500732421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,1,power_law_1.01,0.5188166427612304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,1,power_law_1.01,0.7034636688232422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,1,power_law_1.01,0.8735110473632812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,1,power_law_1.01,1.264170913696289
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,1,1,balanced,0.7591737365722656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,1,power_law_1.01,1.6756396484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,1,power_law_1.01,0.044452481269836426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,1,power_law_1.01,0.06625088214874267
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,1,power_law_1.01,0.08185855865478515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,1,power_law_1.01,0.0990835189819336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,1,power_law_1.01,0.1364236831665039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,1,power_law_1.01,2.4876614379882813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,1,power_law_1.01,0.14186944007873534
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,1,power_law_1.01,0.1467411231994629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,1,power_law_1.01,0.15815551757812502
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,1,power_law_1.01,0.15620160102844238
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,1,power_law_1.01,0.1741414451599121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,1,power_law_1.01,0.175601921081543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,1,power_law_1.01,3.2439404296874996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,1,power_law_1.01,0.17909376144409178
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,1,power_law_1.01,0.1866579246520996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,1,power_law_1.01,0.150316801071167
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,1,power_law_1.01,0.1986355209350586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,1,power_law_1.01,0.1991187286376953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,1,power_law_1.01,0.20660800933837892
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,1,power_law_1.01,0.22411264419555663
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,1,power_law_1.01,0.2621183967590332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,1,power_law_1.01,0.3030316734313965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,1,power_law_1.01,0.3777433776855469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,1,power_law_1.01,0.03407232046127319
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,1,power_law_1.01,0.03657536029815674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,1,power_law_1.01,0.4547488021850586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,1,power_law_1.01,0.6164089584350586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,1,power_law_1.01,0.7801273345947266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,1,1,power_law_1.01,0.03653503894805908
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,1,1,power_law_1.01,0.04270143985748291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,1,power_law_1.01,1.1024703979492188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,1,1,power_law_1.01,0.05581696033477783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,1,power_law_1.01,1.4271398925781251
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,1,1,power_law_1.01,0.061850881576538085
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,1,1,power_law_1.01,0.08077823638916015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,1,1,power_law_1.01,0.08506048202514649
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,1,1,power_law_1.01,0.04889279842376709
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,1,1,power_law_1.01,0.087260160446167
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,1,1,power_law_1.01,0.0893836784362793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,1,1,power_law_1.01,0.08981760025024414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,1,1,power_law_1.01,0.09263615608215332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,1,1,power_law_1.01,0.09483072280883789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,1,1,power_law_1.01,0.09707072257995605
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,1,1,power_law_1.01,0.10200639724731446
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,1,1,power_law_1.01,0.10703424453735351
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,1,1,power_law_1.01,0.11009984016418457
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,1,1,power_law_1.01,0.11910207748413086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,1,1,power_law_1.01,0.1557414436340332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,1,1,power_law_1.01,0.13688511848449708
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,1,1,power_law_1.01,0.1991628837585449
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,1,1,power_law_1.01,0.24466623306274413
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,1,1,power_law_1.01,0.3360742568969727
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,1,1,power_law_1.01,0.6323219299316406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,1,1,power_law_1.01,0.8359571075439453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,1,power_law_1.2,0.07623744010925293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,1,1,power_law_1.01,1.2528966522216796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,1,power_law_1.2,0.10868991851806639
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,1,1,power_law_1.01,1.6761798095703124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,1,power_law_1.2,0.19466880798339842
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,1,power_law_1.2,0.12966208457946776
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,1,power_law_1.2,0.21529920578002928
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,1,power_law_1.2,0.20498239517211916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,1,power_law_1.2,0.22165184020996093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,1,power_law_1.2,0.03852479934692383
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,1,1,power_law_1.01,0.43195327758789065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,1,power_law_1.2,0.2227065658569336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,1,power_law_1.2,0.23948608398437501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,1,power_law_1.2,0.24398143768310546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,1,power_law_1.2,0.24487039566040042
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,1,power_law_1.2,0.2628895950317383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,1,power_law_1.2,0.2754636764526367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,1,power_law_1.2,0.27862079620361324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,1,power_law_1.2,0.2892019271850586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,1,power_law_1.2,0.32681793212890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,1,power_law_1.2,0.36066112518310545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,1,power_law_1.2,0.442748794555664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,1,power_law_1.2,0.5227308654785157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,1,power_law_1.2,0.7030643463134766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,1,power_law_1.2,0.8833542633056641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,1,power_law_1.2,0.034879360198974606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,1,power_law_1.2,0.042858881950378416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,1,power_law_1.2,0.06484096050262451
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,1,power_law_1.2,0.07644415855407714
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,1,power_law_1.2,0.09628543853759766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,1,power_law_1.2,0.13660287857055664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,1,power_law_1.2,0.1447654438018799
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,1,power_law_1.2,0.14967552185058594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,1,power_law_1.2,0.15275520324707031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,1,power_law_1.2,1.2308908843994142
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,1,power_law_1.2,0.15697343826293947
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,1,power_law_1.2,0.0541593599319458
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,1,power_law_1.2,0.1522982406616211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,1,power_law_1.2,0.173504638671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,1,power_law_1.2,0.17463359832763672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,1,power_law_1.2,0.18275007247924804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,1,power_law_1.2,0.18734655380249024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,1,power_law_1.2,0.1902579116821289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,1,power_law_1.2,0.1984115219116211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,1,power_law_1.2,0.21149951934814454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,1,power_law_1.2,0.2285862350463867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,1,power_law_1.2,0.268603515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,1,power_law_1.2,0.3811948776245117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,1,power_law_1.2,0.463884162902832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,1,power_law_1.2,1.6643469238281248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,1,power_law_1.2,0.621640968322754
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,1,1,power_law_1.2,0.036512000560760496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,1,power_law_1.2,0.3016460800170898
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,1,1,power_law_1.2,0.04300352096557617
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,1,1,power_law_1.2,0.05315072059631347
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,1,1,power_law_1.2,0.060163841247558594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,1,1,power_law_1.2,0.08154175758361817
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,1,1,power_law_1.2,0.08550848007202148
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,1,1,power_law_1.2,0.09035391807556152
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,1,power_law_1.2,0.7858124542236328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,1,1,power_law_1.2,0.09123904228210448
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,1,1,power_law_1.2,0.09378239631652832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,1,1,power_law_1.2,0.09770432472229004
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,1,1,power_law_1.2,0.10021632194519044
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,1,1,power_law_1.2,0.10648639678955078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,1,1,power_law_1.2,0.11245887756347657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,1,1,power_law_1.2,0.11668160438537598
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,1,power_law_1.2,2.421948089599609
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,1,1,power_law_1.2,0.04675007820129394
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,1,1,power_law_1.2,0.12968192100524903
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,1,power_law_1.2,1.1090579223632813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,1,1,power_law_1.2,0.156495361328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,1,1,power_law_1.2,0.0868825626373291
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,1,1,power_law_1.2,0.18515775680541993
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,1,1,power_law_1.2,0.24903295516967772
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,1,power_law_1.2,3.1429125976562498
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,1,1,power_law_1.2,0.315467529296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,1,1,power_law_1.2,0.4552876663208008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,1,balanced,0.14199040412902833
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,1,balanced,0.24979711532592774
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,1,1,power_law_1.2,0.597960319519043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,1,balanced,0.44700992584228516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,1,balanced,0.80648193359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,1,balanced,0.8184230041503906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,1,balanced,0.8296173095703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,1,balanced,0.8113510131835937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,1,balanced,0.08226048469543457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,1,balanced,0.8239193725585938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,1,balanced,0.8177523040771485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,1,balanced,0.8250745391845703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,1,balanced,0.8980544281005859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,1,balanced,0.9177900695800781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,1,1,power_law_1.2,0.8776108551025391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,1,balanced,0.9160717010498047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,1,balanced,0.9346015930175782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,1,balanced,0.9400434875488282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,1,balanced,0.9999231719970704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,1,power_law_1.2,1.4286994934082031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,1,balanced,1.0682841491699218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,1,balanced,1.5909228515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,1,balanced,1.7616184997558595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,1,balanced,0.9339788818359376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,1,balanced,2.454554901123047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,1,balanced,3.0725439453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,1,balanced,0.060947837829589846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,1,balanced,0.0910649585723877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,1,balanced,0.14161151885986328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,1,balanced,0.2388870429992676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,1,balanced,4.496433410644531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,1,balanced,0.4269977569580078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,1,balanced,0.4356800079345703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,1,balanced,0.43784446716308595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,1,balanced,0.4402668762207031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,1,balanced,0.43947265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,1,balanced,5.841602172851562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,1,balanced,0.4419820785522461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,1,balanced,0.4415251159667969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,1,balanced,0.5057401657104492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,1,balanced,0.4959372711181641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,1,balanced,0.49008960723876954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,1,balanced,0.49992511749267576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,1,balanced,0.5097766494750976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,1,balanced,0.5170451354980469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,1,balanced,0.5396812820434571
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,1,balanced,0.5648889541625977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,1,balanced,0.5976870346069336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,1,1,power_law_1.2,1.150719985961914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,1,balanced,0.6541350555419922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,1,balanced,0.9804108428955078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,1,balanced,1.0857234954833985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,1,balanced,1.5198873901367187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,1,balanced,8.921024169921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,1,balanced,1.970782012939453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,1,balanced,2.892294921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,1,balanced,3.723697204589844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,1,balanced,12.587269287109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,1,1,balanced,0.04792384147644043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,1,1,balanced,0.062427520751953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,1,1,balanced,0.2673702430725098
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,1,1,balanced,0.08964223861694336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,1,1,balanced,0.26697984695434573
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,1,1,power_law_1.2,1.709736328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,1,1,balanced,0.26896831512451175
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,1,1,balanced,0.269477767944336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,1,1,balanced,0.2702924728393555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,1,1,balanced,0.14639424324035644
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,1,1,balanced,0.2736454391479492
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,1,1,balanced,0.27638208389282226
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,1,1,balanced,0.27937343597412106
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,1,1,balanced,0.28264959335327144
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,1,1,balanced,0.2887833595275879
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,1,1,balanced,0.2995903968811035
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,1,1,balanced,0.3106131172180176
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,1,1,balanced,0.26653888702392575
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,1,1,balanced,0.33370880126953123
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,1,1,balanced,0.3558726501464844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,1,1,balanced,0.2671699142456055
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,1,1,balanced,0.2713068771362305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,1,1,balanced,0.5338911819458008
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,1,1,balanced,0.5897004699707031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,1,1,balanced,0.8702706909179687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,1,power_law_1.01,0.37287296295166017
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,1,1,balanced,1.1181581115722656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,1,1,power_law_1.2,2.276458282470703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,1,power_law_1.01,0.4352915191650391
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,1,1,balanced,1.6572972106933594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,1,power_law_1.01,0.6792537689208984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,1,power_law_1.01,0.7124422454833985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,1,power_law_1.01,0.7335564422607421
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,1,1,balanced,2.2198112487792967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,1,power_law_1.01,0.7776236724853516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,1,power_law_1.01,0.784967041015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,1,power_law_1.01,0.8555999755859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,1,power_law_1.01,0.08192255973815918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,1,power_law_1.01,0.8703609466552734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,1,power_law_1.01,0.8820082855224609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,1,power_law_1.01,0.9055538940429688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,1,power_law_1.01,0.9470124816894531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,1,power_law_1.01,0.13747967720031737
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,1,power_law_1.01,1.0477619171142578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,1,power_law_1.01,1.2059321594238281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,1,power_law_1.01,1.3454354858398436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,1,power_law_1.01,1.6803596496582032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,1,power_law_1.01,2.065930938720703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,1,power_law_1.01,2.6970840454101563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,1,power_law_1.01,3.449684448242188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,1,power_law_1.01,0.9590169525146484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,1,power_law_1.01,0.24086591720581058
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,1,power_law_1.01,0.09179903984069823
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,1,power_law_1.01,0.25219072341918947
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,1,power_law_1.01,0.36565696716308593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,1,power_law_1.01,4.805695495605469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,1,power_law_1.01,0.3855084609985352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,1,power_law_1.01,0.3987635040283203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,1,power_law_1.01,0.4080441665649414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,1,power_law_1.01,0.411773452758789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,1,power_law_1.01,6.1105194091796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,1,power_law_1.01,0.4239916610717773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,1,power_law_1.01,0.4688639831542969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,1,power_law_1.01,0.472276496887207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,1,power_law_1.01,0.4806387329101563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,1,power_law_1.01,0.496759033203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,1,power_law_1.01,0.139933443069458
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,1,power_law_1.01,0.19536895751953126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,1,power_law_1.01,0.5026623916625976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,1,power_law_1.01,0.530744972229004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,1,power_law_1.01,0.05792640209197998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,1,power_law_1.01,0.5502048110961913
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,1,power_law_1.01,0.5997196960449218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,1,power_law_1.01,0.6862226867675781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,1,power_law_1.01,0.8032319641113281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,1,power_law_1.01,8.850462036132813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,1,power_law_1.01,1.0153190612792968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,1,power_law_1.01,1.180794219970703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,1,power_law_1.01,1.7002041625976563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,1,power_law_1.01,2.120336608886719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,1,1,power_law_1.01,0.12149760246276856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,1,power_law_1.01,2.9906072998046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,1,1,power_law_1.01,0.0658239984512329
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,1,power_law_1.01,11.51634765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,1,1,power_law_1.01,0.09071935653686523
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,1,1,power_law_1.01,0.14857279777526855
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,1,1,power_law_1.01,0.22537599563598634
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,1,1,power_law_1.01,0.2452716827392578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,1,1,power_law_1.01,0.050079998970031736
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,1,1,power_law_1.01,0.2396678352355957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,1,power_law_1.01,3.834490966796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,1,1,power_law_1.01,0.25378047943115234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,1,1,power_law_1.01,0.27015104293823244
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,1,1,power_law_1.01,0.2745747184753418
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,1,1,power_law_1.01,0.2650643157958984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,1,1,power_law_1.01,0.28661376953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,1,1,power_law_1.01,0.29944128036499024
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,1,1,power_law_1.01,0.3290496063232422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,1,1,power_law_1.01,0.25283327102661135
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,1,1,power_law_1.01,0.306550407409668
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,1,1,power_law_1.01,0.37233470916748046
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,1,1,power_law_1.01,0.4188991928100586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,1,1,power_law_1.01,0.5261555099487305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,1,power_law_1.2,0.08309951782226563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,1,1,power_law_1.01,0.6426502227783203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,1,power_law_1.2,0.23388223648071288
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,1,1,power_law_1.01,0.8703910064697264
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,1,1,power_law_1.01,1.1126016235351561
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,1,power_law_1.2,0.3264633560180664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,1,power_law_1.2,0.4063865661621094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,1,power_law_1.2,0.6782150268554688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,1,power_law_1.2,0.7223270416259766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,1,power_law_1.2,0.7501197052001953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,1,power_law_1.2,0.13749567985534666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,1,power_law_1.2,0.7990290832519531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,1,power_law_1.2,0.8040076446533202
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,1,power_law_1.2,0.8791584014892578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,1,power_law_1.2,0.8877017974853516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,1,power_law_1.2,0.8926630401611328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,1,power_law_1.2,0.9266893005371093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,1,power_law_1.2,0.967034912109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,1,power_law_1.2,0.9772806549072266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,1,power_law_1.2,1.045704345703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,1,power_law_1.2,1.2384390258789062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,1,power_law_1.2,1.4003269958496094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,1,power_law_1.2,1.746812744140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,1,power_law_1.2,2.074040985107422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,1,power_law_1.2,2.704532470703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,1,power_law_1.2,3.3211431884765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,1,power_law_1.2,4.759717102050781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,1,1,power_law_1.01,1.593797149658203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,1,power_law_1.2,0.09281408309936523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,1,power_law_1.2,0.13566207885742188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,1,1,power_law_1.01,2.087541809082031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,1,power_law_1.2,0.1783238410949707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,1,power_law_1.2,0.2331955146789551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,1,power_law_1.2,0.36664447784423826
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,1,power_law_1.2,0.39035648345947266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,1,power_law_1.2,0.40598529815673834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,1,power_law_1.2,0.41357505798339844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,1,1,power_law_1.01,4.176206665039063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,1,power_law_1.2,0.41446720123291014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,1,power_law_1.2,0.425164794921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,1,power_law_1.2,0.05741824150085449
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,1,power_law_1.2,0.47152446746826177
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,1,1,power_law_1.01,3.1079974365234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,1,power_law_1.2,0.47888961791992185
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,1,power_law_1.2,0.4962777709960937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,1,power_law_1.2,0.5009331130981446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,1,power_law_1.2,6.209862060546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,1,power_law_1.2,0.5105817413330078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,1,power_law_1.2,0.528766098022461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,1,power_law_1.2,0.555792007446289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,1,power_law_1.2,0.6139321517944336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,1,power_law_1.2,0.7146144104003906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,1,power_law_1.2,0.8204895782470704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,1,power_law_1.2,1.0392530822753907
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,1,1,power_law_1.2,0.050378241539001466
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,1,power_law_1.2,1.265905303955078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,1,1,power_law_1.2,0.06511295795440673
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,1,1,power_law_1.2,0.08831487655639648
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,1,1,power_law_1.2,0.11289152145385742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,1,1,power_law_1.2,0.13949567794799805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,1,1,power_law_1.2,0.22732223510742186
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,1,1,power_law_1.2,0.24239871978759764
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,1,1,power_law_1.2,0.24975679397583006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,1,power_law_1.2,1.6569471740722654
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,1,1,power_law_1.2,0.258341121673584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,1,1,power_law_1.2,0.25851711273193356
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,1,1,power_law_1.2,0.2677702331542969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,1,1,power_law_1.2,0.2776896095275879
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,1,1,power_law_1.2,0.28210111618041994
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,1,power_law_1.2,2.125909729003906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,1,power_law_1.2,8.8942138671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,1,1,power_law_1.2,0.3128550338745117
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,1,1,power_law_1.2,0.2981439971923828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,1,1,power_law_1.2,0.32223232269287105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,1,power_law_1.2,2.9965625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,1,1,power_law_1.2,0.35306495666503906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,1,1,power_law_1.2,0.4172019195556641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,1,power_law_1.2,11.618270263671874
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,1,1,power_law_1.2,0.4846944046020508
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,1,1,power_law_1.2,0.6383168029785156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,1,1,power_law_1.2,0.7934188842773438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,1,1,balanced,0.15471808433532713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,1,1,balanced,0.3305132675170898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,1,1,balanced,0.5339347076416016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,1,1,power_law_1.2,1.1127200317382813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,1,1,balanced,1.0174368286132813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,1,power_law_1.2,3.8832281494140624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,1,1,power_law_1.2,1.4446905517578126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,1,1,balanced,2.603027954101562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,1,1,balanced,2.6038079833984376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,1,1,balanced,2.0009767150878908
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,1,1,balanced,2.5362144470214845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,1,1,balanced,2.63814208984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,1,1,balanced,2.6278707885742185
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,1,1,balanced,2.649400329589844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,1,1,balanced,2.690687255859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,1,1,balanced,2.804856872558594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,1,1,balanced,2.8116229248046873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,1,1,balanced,2.8220794677734373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,1,1,balanced,2.8330541992187497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,1,1,balanced,2.879356689453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,1,1,balanced,2.99315185546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,1,1,balanced,3.1221978759765627
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,1,1,power_law_1.2,2.108441619873047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,1,1,balanced,4.748849182128906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,1,1,balanced,4.991125793457032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,1,1,balanced,7.222673950195312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,1,1,balanced,8.934308471679689
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,1,1,balanced,0.11402367591857909
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,1,1,balanced,0.17236928939819335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,1,1,balanced,0.2914182472229004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,1,1,balanced,0.5191891098022461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,1,1,power_law_1.2,2.786705322265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,1,1,balanced,12.27232666015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,1,1,balanced,1.0005165100097657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,1,1,balanced,1.260240020751953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,1,1,balanced,1.2748799896240235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,1,1,balanced,1.2792908477783205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,1,1,balanced,1.2569670104980468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,1,1,balanced,1.2828933715820312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,1,1,balanced,1.2610797119140624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,1,1,balanced,1.2747187042236328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,1,1,balanced,18.120546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,1,1,balanced,1.4382284545898438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,1,1,balanced,1.4474803161621093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,1,1,balanced,1.4618853759765624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,1,1,balanced,1.4876864624023436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,1,1,balanced,1.4914291381835938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,1,1,balanced,1.535361328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,1,1,balanced,1.5829522705078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,1,1,balanced,1.689056701660156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,1,1,balanced,1.7814694213867188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,1,1,balanced,2.690804443359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,1,1,balanced,2.824398193359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,1,1,balanced,3.999882202148437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,1,1,power_law_1.2,4.182415466308593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,1,1,balanced,27.29921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,1,1,balanced,4.975646057128906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,1,1,balanced,6.53717529296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,1,1,balanced,8.60309814453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,1,1,balanced,37.1352978515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,1,1,power_law_1.2,5.5426049804687505
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,1,1,balanced,0.07466688156127929
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,1,1,balanced,0.10568960189819336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,1,1,balanced,0.16992128372192383
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,1,1,balanced,0.7398969268798827
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,1,1,balanced,0.741325454711914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,1,1,balanced,0.7494771575927734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,1,1,balanced,0.30728639602661134
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,1,1,balanced,0.7480140686035156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,1,1,balanced,0.7475654602050781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,1,1,balanced,0.7625286102294921
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,1,1,balanced,0.7622764587402344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,1,1,balanced,0.593051528930664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,1,1,balanced,0.7663091278076172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,1,1,balanced,0.7863603210449218
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,1,1,balanced,0.8157241821289063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,1,1,balanced,0.8160409545898437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,1,1,balanced,0.8730220794677734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,1,1,balanced,0.9102143859863281
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,1,1,balanced,0.7363609313964844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,1,1,balanced,0.7354156494140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,1,1,power_law_1.01,0.1597599983215332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,1,1,power_law_1.01,0.3288889694213867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,1,1,power_law_1.01,0.5363199996948242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,1,1,power_law_1.01,0.8665939331054687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,1,1,power_law_1.01,1.1993170928955077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,1,1,power_law_1.01,1.9992498779296874
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,1,1,balanced,0.7362918090820313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,1,1,power_law_1.01,2.1431648254394533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,1,1,power_law_1.01,2.147984619140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,1,1,power_law_1.01,2.299022674560547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,1,1,power_law_1.01,2.3793414306640623
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,1,1,balanced,1.5298739624023436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,1,1,power_law_1.01,2.6208230590820314
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,1,1,balanced,1.6592108154296874
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,1,1,balanced,2.270362243652344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,1,1,power_law_1.01,2.779516296386719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,1,1,balanced,2.928283386230469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,1,1,power_law_1.01,2.573723449707031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,1,1,power_law_1.01,2.7154296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,1,1,balanced,3.8022637939453126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,1,1,power_law_1.01,2.7675628662109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,1,1,power_law_1.01,2.8102911376953124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,1,1,power_law_1.01,3.0127679443359376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,1,1,power_law_1.01,3.4772152709960933
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,1,1,balanced,5.265789184570313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,1,1,power_law_1.01,3.862323913574219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,1,1,power_law_1.01,4.665450134277344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,1,1,power_law_1.01,0.1162060832977295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,1,1,power_law_1.01,0.17227392196655272
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,1,1,power_law_1.01,5.552747802734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,1,1,power_law_1.01,0.29249216079711915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,1,1,power_law_1.01,0.44944385528564457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,1,1,power_law_1.01,0.6119795227050782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,1,1,power_law_1.01,7.288762817382812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,1,1,power_law_1.01,0.9788140869140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,1,1,power_law_1.01,1.0359442901611327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,1,1,power_law_1.01,1.0879814147949218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,1,1,power_law_1.01,1.1071289825439454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,1,1,power_law_1.01,8.954782104492187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,1,1,power_law_1.01,1.1321875000000001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,1,1,power_law_1.01,1.1703110504150391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,1,1,power_law_1.01,1.1930188751220703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,1,1,power_law_1.01,1.3291116333007813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,1,1,power_law_1.01,1.375096893310547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,1,1,power_law_1.01,1.367550048828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,1,1,power_law_1.01,1.3849778747558594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,1,1,power_law_1.01,1.437401580810547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,1,1,power_law_1.01,12.2713134765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,1,1,power_law_1.01,1.5853382873535158
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,1,1,power_law_1.01,1.6379180908203126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,1,1,power_law_1.01,1.8048448181152346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,1,1,power_law_1.01,2.1049708557128906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,1,1,power_law_1.01,2.5802386474609373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,1,1,power_law_1.01,3.0437799072265626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,1,1,power_law_1.01,15.77047119140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,1,1,power_law_1.01,0.08285632133483886
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,1,1,power_law_1.01,0.11429247856140137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,1,1,power_law_1.01,3.9773504638671873
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,1,1,power_law_1.01,0.17283199310302735
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,1,1,power_law_1.01,0.26412864685058596
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,1,1,power_law_1.01,0.3543667221069336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,1,1,power_law_1.01,4.892914428710937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,1,1,power_law_1.01,0.577136001586914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,1,1,power_law_1.01,0.6124979019165039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,1,1,power_law_1.01,0.6319187164306641
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,1,1,power_law_1.01,0.6558195495605469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,1,1,power_law_1.01,0.6715481567382813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,1,1,power_law_1.01,6.737539672851563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,1,1,power_law_1.01,0.6878662109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,1,1,power_law_1.01,0.7093984222412109
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,1,1,power_law_1.01,0.7129542541503906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,1,1,power_law_1.01,0.7598387145996094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,1,1,power_law_1.01,0.7659398651123046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,1,1,power_law_1.01,8.551583251953124
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,1,1,power_law_1.01,0.7699257659912109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,1,1,power_law_1.01,22.5866552734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,1,1,power_law_1.01,1.0230796813964844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,1,1,power_law_1.01,29.966738281250002
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,1,1,power_law_1.01,0.854800033569336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,1,1,power_law_1.01,0.9291680145263672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,1,1,power_law_1.01,1.192987518310547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,1,1,power_law_1.01,1.4349395751953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,1,1,power_law_1.2,0.15963264465332033
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,1,1,power_law_1.01,1.8331027221679688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,1,1,power_law_1.2,0.3305747222900391
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,1,1,power_law_1.01,2.2964306640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,1,1,power_law_1.2,0.5213804626464844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,1,1,power_law_1.2,0.7889260864257812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,1,1,power_law_1.2,1.0955398559570313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,1,1,power_law_1.2,1.9305740356445313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,1,1,power_law_1.01,3.1264135742187498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,1,1,power_law_1.2,2.048703308105469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,1,1,power_law_1.2,2.2079302978515623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,1,1,power_law_1.2,2.3289939880371096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,1,1,power_law_1.2,2.5104435729980468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,1,1,power_law_1.2,2.669921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,1,1,power_law_1.2,2.8802239990234377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,1,1,power_law_1.2,2.691068115234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,1,1,power_law_1.2,2.768944091796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,1,1,power_law_1.2,2.8939117431640624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,1,1,power_law_1.2,2.9645675659179687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,1,1,power_law_1.2,3.150637512207031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,1,1,power_law_1.01,4.046402587890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,1,1,power_law_1.2,3.5350738525390626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,1,1,power_law_1.2,3.9217593383789064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,1,1,power_law_1.2,4.812076110839844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,1,1,power_law_1.2,5.630577392578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,1,1,power_law_1.2,7.3617919921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,1,1,power_law_1.2,9.08315673828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,1,1,power_law_1.2,0.17441856384277343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,1,1,power_law_1.2,0.28328256607055663
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,1,1,power_law_1.2,0.4046656036376953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,1,1,power_law_1.2,0.5654431915283202
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,1,1,power_law_1.2,12.404022216796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,1,1,power_law_1.2,0.11638208389282227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,1,1,power_law_1.2,0.9682201385498047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,1,1,power_law_1.2,1.0377139282226562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,1,1,power_law_1.2,1.0871437072753907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,1,1,power_law_1.2,1.1367340850830079
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,1,1,power_law_1.01,5.836875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,1,1,power_law_1.2,1.15963134765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,1,1,power_law_1.2,1.1933535766601562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,1,1,power_law_1.2,1.2213382720947266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,1,1,power_law_1.2,1.36546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,1,1,power_law_1.2,1.3953459167480469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,1,1,power_law_1.2,16.12964111328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,1,1,power_law_1.2,1.430330810546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,1,1,power_law_1.2,1.477139892578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,1,1,power_law_1.2,1.497681884765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,1,1,power_law_1.01,7.739783935546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,1,1,power_law_1.2,1.6042752075195312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,1,1,power_law_1.2,1.7145248413085938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,1,1,power_law_1.2,1.9004159545898438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,1,1,power_law_1.2,2.1576512145996096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,1,1,power_law_1.2,2.6176690673828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,1,1,power_law_1.2,3.038197021484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,1,1,power_law_1.2,0.08189248085021973
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,1,1,power_law_1.2,0.11069439888000487
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,1,1,power_law_1.2,0.16898303985595703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,1,1,power_law_1.2,3.980381469726562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,1,1,power_law_1.2,0.23911615371704104
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,1,1,power_law_1.2,0.32894081115722656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,1,1,power_law_1.2,0.5636511993408203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,1,1,power_law_1.2,4.915145568847656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,1,1,power_law_1.2,0.6054496002197266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,1,1,power_law_1.2,0.6437190246582032
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,1,1,power_law_1.2,0.6666937255859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,1,1,power_law_1.2,0.6864556884765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,1,1,power_law_1.2,0.7089644622802734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,1,1,power_law_1.2,0.7254771423339844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,1,1,power_law_1.2,0.7471871948242187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,1,1,power_law_1.2,0.7696672058105468
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,1,1,power_law_1.2,0.8258131408691407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,1,1,power_law_1.2,0.8606822204589843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,1,1,power_law_1.2,23.92079833984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,1,1,power_law_1.2,6.762622680664063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,1,1,power_law_1.2,8.757735595703124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,1,1,power_law_1.2,31.178125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,1,1,power_law_1.2,0.8885266876220703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,1,1,power_law_1.2,1.002740478515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,1,1,power_law_1.2,1.1361542510986329
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,1,1,power_law_1.2,1.390045471191406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,1,balanced,0.15987135887145995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,1,balanced,0.30958335876464843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,1,balanced,0.5357548904418945
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,1,1,power_law_1.2,1.7192588806152345
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,1,1,power_law_1.2,2.2515423583984377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,1,balanced,0.9583513641357422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,1,balanced,1.8916268920898438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,1,balanced,3.766687927246094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,1,balanced,5.7806420898437505
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,1,balanced,5.823276977539062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,1,balanced,5.748057250976562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,1,1,power_law_1.2,2.890417785644531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,1,balanced,5.807791137695313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,1,balanced,5.850634155273438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,1,balanced,5.8029443359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,1,balanced,5.9043688964843755
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,1,1,power_law_1.2,4.2322802734375005
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,1,balanced,5.901707763671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,1,balanced,5.95084228515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,1,balanced,5.93780029296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,1,balanced,6.129349975585937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,1,1,power_law_1.2,5.593720092773437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,1,balanced,6.182264404296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,1,balanced,6.2913275146484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,1,balanced,6.4958056640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,1,balanced,6.7057666015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,1,balanced,7.346837768554687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,1,1,power_law_1.2,8.398908081054689
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,1,balanced,0.09746815681457519
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,1,balanced,0.27518272399902344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,1,balanced,0.15990015983581543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,1,balanced,0.493644790649414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,1,balanced,0.9490322875976563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,1,balanced,10.938363037109374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,1,balanced,11.71896484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,1,balanced,1.8886618041992187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,1,balanced,2.8246426391601562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,1,balanced,2.8229144287109373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,1,balanced,2.8387628173828126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,1,balanced,2.8744351196289064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,1,balanced,2.8514739990234377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,1,balanced,2.871850891113281
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,1,1,power_law_1.2,11.234422607421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,1,balanced,2.8908831787109373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,1,balanced,17.375440673828127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,1,balanced,2.9020736694335936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,1,balanced,2.897244873046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,1,balanced,2.9139678955078123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,1,balanced,3.1909881591796876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,1,balanced,3.258482055664062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,1,balanced,3.3227380371093753
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,1,balanced,3.460172119140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,1,balanced,3.468681030273437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,1,balanced,3.7187347412109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,1,balanced,3.931997375488281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,1,balanced,4.398327026367188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,1,balanced,25.02170654296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,1,balanced,6.224437866210938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,1,balanced,6.979654541015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,1,balanced,9.366218872070313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,1,balanced,39.7851708984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,1,1,balanced,0.10690688133239745
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,1,1,balanced,0.10663871765136719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,1,1,balanced,0.173253116607666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,1,1,balanced,0.10845184326171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,1,1,balanced,0.11081151962280274
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,1,1,balanced,0.1867718315124512
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,1,1,balanced,0.21376256942749022
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,1,1,balanced,0.11805760383605957
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,1,1,balanced,0.13280703544616698
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,1,1,balanced,0.14585280418395996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,1,1,balanced,0.15993791580200195
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,1,1,balanced,0.2665644836425781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,1,1,balanced,0.2962860870361328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,1,1,balanced,0.3524070358276367
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,1,1,balanced,0.4285164642333984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,1,1,balanced,0.48554622650146484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,1,1,balanced,0.615035514831543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,1,1,balanced,0.8833036804199219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,1,1,balanced,1.1567027282714843
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,1,1,balanced,1.6907891845703127
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,1,1,balanced,2.240647735595703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,1,power_law_1.01,0.3138956832885742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,1,1,balanced,3.30288330078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,1,power_law_1.01,0.5326380920410156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,1,1,balanced,4.377622375488281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,1,power_law_1.01,0.9587699127197264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,1,power_law_1.01,1.759344024658203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,1,power_law_1.01,0.15608256340026855
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,1,1,balanced,6.515502319335937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,1,power_law_1.01,2.474500427246094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,1,power_law_1.01,2.879647216796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,1,power_law_1.01,3.8072076416015626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,1,power_law_1.01,3.9385406494140627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,1,power_law_1.01,4.322133178710938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,1,power_law_1.01,4.113395690917969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,1,power_law_1.01,4.417803649902344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,1,power_law_1.01,4.6879815673828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,1,power_law_1.01,4.94774658203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,1,power_law_1.01,5.24574462890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,1,power_law_1.01,5.85601806640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,1,power_law_1.01,5.10171142578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,1,power_law_1.01,5.488460693359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,1,power_law_1.01,5.9894677734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,1,power_law_1.01,6.805653686523438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,1,power_law_1.01,7.639485473632813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,1,power_law_1.01,9.303518676757813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,1,power_law_1.01,10.93701904296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,1,power_law_1.01,0.09837632179260254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,1,power_law_1.01,14.061842041015627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,1,power_law_1.01,0.160118408203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,1,power_law_1.01,0.27673408508300784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,1,power_law_1.01,0.49292545318603515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,1,power_law_1.01,17.23572509765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,1,power_law_1.01,0.8777401733398438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,1,power_law_1.01,1.268008346557617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,1,power_law_1.01,1.4933351135253905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,1,power_law_1.01,1.9019206237792967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,1,power_law_1.01,2.0012722778320313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,1,power_law_1.01,2.088636779785156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,1,power_law_1.01,2.1580288696289065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,1,power_law_1.01,2.1921881103515624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,1,power_law_1.01,2.256053771972656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,1,power_law_1.01,23.830820312500002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,1,power_law_1.01,2.340531768798828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,1,power_law_1.01,2.3797669982910157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,1,power_law_1.01,2.425778503417969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,1,power_law_1.01,2.6700250244140626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,1,power_law_1.01,2.7769683837890624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,1,power_law_1.01,2.9216180419921876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,1,power_law_1.01,3.1124475097656252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,1,power_law_1.01,3.3661599731445313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,1,power_law_1.01,3.780939636230469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,1,power_law_1.01,30.2439453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,1,power_law_1.01,4.198921508789063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,1,power_law_1.01,5.1418310546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,1,1,power_law_1.01,0.10697664260864256
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,1,1,power_law_1.01,0.10679360389709473
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,1,power_law_1.01,6.0605517578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,1,1,power_law_1.01,0.1092262363433838
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,1,1,power_law_1.01,0.11124992370605469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,1,1,power_law_1.01,0.13279680252075193
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,1,1,power_law_1.01,0.11824383735656738
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,1,1,power_law_1.01,0.14598464012145995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,1,power_law_1.01,7.948460083007812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,1,1,power_law_1.01,0.16000959396362305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,1,1,power_law_1.01,0.1732499122619629
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,1,1,power_law_1.01,0.18671808242797852
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,1,1,power_law_1.01,0.21361215591430663
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,1,power_law_1.01,9.766615600585938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,1,1,power_law_1.01,0.26671552658081055
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,1,1,power_law_1.01,0.2949305534362793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,1,1,power_law_1.01,0.3540467071533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,1,1,power_law_1.01,0.42840641021728515
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,1,1,power_law_1.01,0.4862080001831055
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,1,1,power_law_1.01,0.6183014297485352
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,1,1,power_law_1.01,0.8853772735595703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,1,1,power_law_1.01,1.1505926513671876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,1,1,power_law_1.01,1.6835494995117188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,1,1,power_law_1.01,2.223553924560547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,1,1,power_law_1.01,3.302167663574219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,1,power_law_1.2,0.15327679634094238
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,1,power_law_1.2,0.3094150352478027
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,1,power_law_1.2,0.5306963348388672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,1,power_law_1.2,1.5799885559082032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,1,power_law_1.2,0.9541158294677734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,1,1,power_law_1.01,4.377156372070313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,1,power_law_1.2,2.2453759765625003
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,1,1,power_law_1.01,6.5389892578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,1,power_law_1.2,2.6564184570312497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,1,power_law_1.2,3.6430419921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,1,power_law_1.2,3.8271243286132814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,1,power_law_1.2,3.981845703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,1,power_law_1.2,4.127080993652344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,1,power_law_1.2,4.440804443359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,1,power_law_1.2,4.659698486328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,1,power_law_1.2,4.961019592285156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,1,power_law_1.2,5.379678344726562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,1,power_law_1.2,5.219297485351563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,1,power_law_1.2,5.716688232421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,1,power_law_1.2,5.821921997070312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,1,power_law_1.2,6.143240966796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,1,power_law_1.2,7.034174194335938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,1,power_law_1.2,7.947361450195312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,1,power_law_1.2,9.60346435546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,1,power_law_1.2,11.369969482421876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,1,power_law_1.2,14.4669677734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,1,power_law_1.2,0.0983347225189209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,1,power_law_1.2,0.15869248390197754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,1,power_law_1.2,0.27702592849731444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,1,power_law_1.2,0.49304382324218754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,1,power_law_1.2,0.8047456359863281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,1,power_law_1.2,17.642064208984372
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,1,power_law_1.2,1.1580774688720703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,1,power_law_1.2,1.3777529907226562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,1,power_law_1.2,1.8686175537109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,1,power_law_1.2,1.9449407958984373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,1,power_law_1.2,2.022520294189453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,1,power_law_1.2,2.0963948059082034
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,1,power_law_1.2,2.174332733154297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,1,power_law_1.2,2.234132537841797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,1,power_law_1.2,2.298830108642578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,1,power_law_1.2,2.3636543273925783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,1,power_law_1.2,2.433829803466797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,1,power_law_1.2,2.6878341674804687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,1,power_law_1.2,24.0846240234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,1,power_law_1.2,2.875166015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,1,power_law_1.2,2.9768804931640624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,1,power_law_1.2,3.2313522338867187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,1,power_law_1.2,3.4907040405273433
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,1,power_law_1.2,4.027031555175781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,1,power_law_1.2,30.51282958984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,1,power_law_1.2,4.482245788574219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,1,power_law_1.2,5.362092895507812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,1,1,power_law_1.2,0.10718848228454589
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,1,1,power_law_1.2,0.10849023818969727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,1,power_law_1.2,6.295755615234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,1,1,power_law_1.2,0.11134464263916015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,1,1,power_law_1.2,0.11840191841125489
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,1,1,power_law_1.2,0.13197504043579103
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,1,1,power_law_1.2,0.14652159690856933
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,1,power_law_1.2,8.166623535156251
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,1,1,power_law_1.2,0.15980863571166992
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,1,1,power_law_1.2,0.1727724838256836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,1,1,power_law_1.2,0.18746688842773437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,1,1,power_law_1.2,0.26764223098754886
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,1,1,power_law_1.2,0.21346176147460935
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,1,1,power_law_1.2,0.2949241638183594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,1,1,power_law_1.2,0.35269760131835937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,1,1,power_law_1.2,0.4290694427490235
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,1,1,power_law_1.2,0.48591041564941406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,1,power_law_1.2,9.852579345703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,1,1,power_law_1.2,0.10585408210754395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,1,1,power_law_1.2,0.6144992065429687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,1,1,power_law_1.2,0.8854879760742188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,1,1,power_law_1.2,1.1548339080810546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,1,1,power_law_1.2,1.6907571411132813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,1,balanced,0.07037888050079347
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,1,balanced,0.10201279640197754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,1,balanced,0.18837312698364256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,1,balanced,0.3410553741455078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,1,balanced,0.6110528182983399
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,1,balanced,1.2014348602294922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,1,balanced,1.2125215911865235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,1,balanced,1.2622541046142577
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,1,1,power_law_1.2,2.2308589172363282
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,1,1,power_law_1.2,3.3149920654296876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,1,balanced,1.2166207885742186
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,1,1,power_law_1.2,4.366041564941407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,1,balanced,1.260080032348633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,1,balanced,1.213870086669922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,1,balanced,1.293409881591797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,1,balanced,1.1895616149902344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,1,balanced,1.3447576904296876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,1,balanced,1.2052825927734374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,1,balanced,1.3563346862792969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,1,balanced,1.3591194152832031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,1,balanced,1.3973574829101563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,1,balanced,1.4311737060546874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,1,balanced,1.5270854187011718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,1,balanced,1.6670008850097655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,1,balanced,2.4355442810058596
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,1,balanced,2.6899423217773437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,1,balanced,3.6942196655273434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,1,balanced,4.832948608398437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,1,balanced,0.08477567672729493
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,1,1,power_law_1.2,6.516975708007813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,1,balanced,0.1173740768432617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,1,balanced,0.33597057342529296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,1,balanced,7.25854248046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,1,balanced,0.6171244812011719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,1,balanced,0.6354956817626953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,1,balanced,0.6263328170776367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,1,balanced,0.6422886657714844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,1,balanced,0.6336851119995117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,1,balanced,0.6429420471191406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,1,balanced,0.6403174591064453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,1,balanced,0.6530182647705078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,1,balanced,9.999913940429687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,1,balanced,0.6621497344970704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,1,balanced,0.7537817382812501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,1,balanced,0.7446387481689454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,1,balanced,0.7495455932617188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,1,balanced,0.7820063781738281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,1,balanced,0.7849779510498047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,1,balanced,0.8042240142822266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,1,balanced,0.8365465545654297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,1,balanced,0.9394201660156251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,1,balanced,0.05081024169921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,1,balanced,0.19285120010375975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,1,balanced,1.027832336425781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,1,balanced,1.5395359802246094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,1,balanced,1.716926727294922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,1,balanced,3.1249472045898434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,1,balanced,2.4432505798339843
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,1,1,balanced,0.06434368133544922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,1,1,balanced,0.07392576217651367
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,1,1,balanced,0.09052800178527833
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,1,1,balanced,0.12805120468139647
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,1,1,balanced,0.3895353698730469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,1,1,balanced,0.3912704086303711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,1,1,balanced,0.39162368774414064
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,1,1,balanced,0.3913395309448242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,1,1,balanced,0.3916249465942383
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,1,1,balanced,0.3929926300048828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,1,1,balanced,0.39359169006347655
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,1,1,balanced,0.3958848190307617
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,1,1,balanced,0.4009542465209961
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,1,1,balanced,0.4083366394042969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,1,1,balanced,0.4015321731567383
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,1,1,balanced,0.4050048065185547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,1,1,balanced,0.41989185333251955
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,1,1,balanced,0.4317030334472657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,1,1,balanced,0.44946239471435545
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,1,1,balanced,0.4718585586547851
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,1,1,balanced,0.51285888671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,1,1,balanced,0.21204288482666014
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,1,1,balanced,0.5583327865600586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,1,power_law_1.01,0.06902463912963867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,1,power_law_1.01,0.10386560440063478
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,1,power_law_1.01,0.1892108726501465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,1,power_law_1.01,0.34321662902832034
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,1,power_law_1.01,0.4901235198974609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,1,power_law_1.01,0.6176160049438477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,1,power_law_1.01,0.8527814483642577
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,1,1,balanced,0.9244127655029297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,1,power_law_1.01,0.9236383819580078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,1,1,balanced,0.9611328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,1,1,balanced,1.3982400512695312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,1,power_law_1.01,0.9600832366943359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,1,power_law_1.01,0.9258790588378906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,1,power_law_1.01,1.0166796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,1,1,balanced,1.7959942626953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,1,power_law_1.01,1.0827436828613282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,1,power_law_1.01,1.1579481506347657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,1,power_law_1.01,1.2366719818115235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,1,power_law_1.01,1.1963244628906249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,1,power_law_1.01,1.204700164794922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,1,power_law_1.01,1.276094741821289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,1,power_law_1.01,1.4908709716796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,1,power_law_1.01,1.745983428955078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,1,power_law_1.01,1.968934326171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,1,power_law_1.01,2.511602630615234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,1,power_law_1.01,1.3644601440429687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,1,power_law_1.01,3.02276611328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,1,power_law_1.01,0.05280447959899902
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,1,power_law_1.01,0.11815232276916503
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,1,power_law_1.01,0.19634624481201174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,1,power_law_1.01,4.137992858886719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,1,power_law_1.01,0.2744268798828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,1,power_law_1.01,0.3672256088256836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,1,power_law_1.01,0.472465934753418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,1,power_law_1.01,0.49127166748046874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,1,power_law_1.01,0.5036588668823242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,1,power_law_1.01,0.5071865463256836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,1,power_law_1.01,0.5320191955566406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,1,power_law_1.01,7.188707885742187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,1,power_law_1.01,0.5528057479858399
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,1,power_law_1.01,0.5467468643188477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,1,power_law_1.01,0.07813695907592774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,1,power_law_1.01,0.6829119873046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,1,power_law_1.01,0.7035148620605469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,1,power_law_1.01,0.7438976287841796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,1,power_law_1.01,9.364357910156249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,1,power_law_1.01,0.8599065399169922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,1,power_law_1.01,5.017705688476562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,1,power_law_1.01,0.914645767211914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,1,power_law_1.01,1.0643852996826173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,1,power_law_1.01,1.2615974426269532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,1,power_law_1.01,1.6761619567871093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,1,power_law_1.01,2.0126739501953126
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,1,1,power_law_1.01,0.0638150405883789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,1,1,power_law_1.01,0.07827263832092285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,1,power_law_1.01,2.740717468261719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,1,power_law_1.01,0.5655116653442382
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,1,1,power_law_1.01,0.131593599319458
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,1,1,power_law_1.01,0.17478912353515624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,1,power_law_1.01,0.6360678482055664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,1,power_law_1.01,3.448359069824219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,1,1,power_law_1.01,0.2136672019958496
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,1,1,power_law_1.01,0.28960575103759767
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,1,power_law_1.01,0.6398259353637695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,1,1,power_law_1.01,0.31382080078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,1,1,power_law_1.01,0.3319436645507813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,1,1,power_law_1.01,0.34428031921386715
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,1,1,power_law_1.01,0.30549055099487304
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,1,1,power_law_1.01,0.34886783599853516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,1,1,power_law_1.01,0.3715571212768555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,1,1,power_law_1.01,0.3790784072875977
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,1,1,power_law_1.01,0.4029542541503906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,1,1,power_law_1.01,0.0973804759979248
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,1,1,power_law_1.01,0.3171334457397461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,1,1,power_law_1.01,0.43828479766845707
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,1,1,power_law_1.01,0.48480384826660156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,1,1,power_law_1.01,0.353900146484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,1,1,power_law_1.01,0.5709971237182617
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,1,1,power_law_1.01,0.655511703491211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,1,power_law_1.2,0.07073984146118165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,1,power_law_1.2,0.10541760444641113
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,1,1,power_law_1.01,0.8404825592041016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,1,power_law_1.2,0.18935232162475585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,1,power_law_1.2,0.3083456039428711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,1,power_law_1.2,0.4401401519775391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,1,power_law_1.2,0.582374382019043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,1,power_law_1.2,0.8506707000732423
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,1,power_law_1.2,0.8918182373046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,1,power_law_1.2,0.9328031921386719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,1,power_law_1.2,0.9638105773925781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,1,power_law_1.2,1.014642562866211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,1,power_law_1.2,1.1002918243408204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,1,power_law_1.2,1.1680850982666016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,1,power_law_1.2,1.290011444091797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,1,power_law_1.2,1.2231168365478515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,1,power_law_1.2,1.2563033294677735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,1,power_law_1.2,1.30434814453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,1,power_law_1.2,1.4617190551757813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,1,power_law_1.2,1.5810861206054687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,1,power_law_1.2,1.864405059814453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,1,power_law_1.2,2.1059896850585935
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,1,power_law_1.2,2.6103265380859377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,1,power_law_1.2,3.1427294921874998
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,1,1,power_law_1.01,1.0330879974365235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,1,power_law_1.2,4.19185791015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,1,power_law_1.2,0.05374591827392579
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,1,power_law_1.2,0.07966784000396729
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,1,power_law_1.2,0.1153983974456787
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,1,power_law_1.2,0.17849088668823243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,1,power_law_1.2,5.252203369140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,1,1,power_law_1.01,1.442487030029297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,1,power_law_1.2,0.2518918418884277
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,1,power_law_1.2,0.35589950561523437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,1,1,power_law_1.01,1.857379150390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,1,power_law_1.2,0.4634944152832031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,1,power_law_1.2,0.4830860900878906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,1,power_law_1.2,0.49635326385498046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,1,power_law_1.2,0.5110246276855468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,1,power_law_1.2,0.5259174346923828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,1,1,power_law_1.01,2.7165728759765626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,1,power_law_1.2,0.5412345504760742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,1,power_law_1.2,0.5594380950927734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,1,power_law_1.2,0.5738451385498047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,1,power_law_1.2,0.6525190734863282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,1,power_law_1.2,0.6645382690429688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,1,power_law_1.2,7.327610473632812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,1,power_law_1.2,0.6762649536132812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,1,power_law_1.2,0.7409343719482422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,1,1,power_law_1.01,3.599896240234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,1,power_law_1.2,0.7807078552246094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,1,power_law_1.2,0.8762528228759766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,1,power_law_1.2,0.9251065826416015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,1,power_law_1.2,1.1416185760498048
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,1,power_law_1.2,1.3291993713378907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,1,power_law_1.2,1.7042655944824219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,1,1,power_law_1.2,0.06401343822479248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,1,power_law_1.2,2.0060377502441407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,1,1,power_law_1.2,0.07758399963378906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,1,1,power_law_1.2,0.0970207977294922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,1,1,power_law_1.2,0.12627391815185546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,1,power_law_1.2,2.775593566894531
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,1,1,power_law_1.2,0.1628223991394043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,1,power_law_1.2,9.365714721679687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,1,1,power_law_1.2,0.20307392120361328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,1,1,power_law_1.2,0.2843526458740234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,1,1,power_law_1.2,0.3014252853393555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,1,1,power_law_1.2,0.3088121604919434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,1,power_law_1.2,3.5095297241210934
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,1,1,power_law_1.2,0.3215315246582031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,1,1,power_law_1.2,0.3298739242553711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,1,1,power_law_1.2,0.3421913528442383
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,1,1,power_law_1.2,0.35592830657958985
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,1,1,power_law_1.2,0.37133758544921874
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,1,1,power_law_1.2,0.3862489700317383
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,1,1,power_law_1.2,0.4004627227783203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,1,1,power_law_1.2,0.49250686645507813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,1,1,power_law_1.2,0.5521414566040039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,1,1,power_law_1.2,0.4304659271240235
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,1,1,power_law_1.2,0.6869593811035156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,1,1,power_law_1.2,0.8245075225830079
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,1,1,power_law_1.2,1.1253670501708986
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,1,1,power_law_1.2,1.4559603881835936
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,1,1,power_law_1.2,2.142079315185547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,1,1,power_law_1.2,2.8996005249023438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,1,1,power_law_1.2,4.475379333496094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,1,balanced,0.5840889739990234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,1,balanced,0.054675841331481935
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,1,balanced,0.25649280548095704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,1,balanced,0.08174079895019531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,1,balanced,0.4355551910400391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,1,balanced,0.1362649631500244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,1,balanced,0.5903257751464844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,1,balanced,0.5940659332275391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,1,balanced,0.581649284362793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,1,balanced,0.6089811325073242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,1,balanced,0.5925555038452148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,1,balanced,0.5924998474121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,1,balanced,0.5989107131958008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,1,balanced,0.598542709350586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,1,balanced,0.6065881729125977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,1,balanced,0.7579494476318359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,1,balanced,0.651944351196289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,1,balanced,0.6814713287353515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,1,balanced,0.8863795471191406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,1,1,power_law_1.2,6.183097534179687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,1,balanced,0.7709311676025391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,1,balanced,0.8092998504638672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,1,balanced,1.1063878631591797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,1,balanced,1.1669197082519531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,1,balanced,2.0599647521972657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,1,balanced,1.967589111328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,1,balanced,2.8964178466796873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,1,balanced,3.134405212402344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,1,1,balanced,0.04215680122375488
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,1,1,balanced,0.22035072326660154
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,1,1,balanced,0.22100799560546874
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,1,1,balanced,0.05194431781768799
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,1,1,balanced,0.22305856704711916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,1,1,balanced,0.22399551391601563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,1,1,balanced,0.2243622398376465
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,1,1,balanced,0.22767744064331055
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,1,1,balanced,0.22925376892089844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,1,1,balanced,0.23146944046020507
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,1,1,balanced,0.23555839538574216
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,1,1,balanced,0.24316032409667968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,1,1,balanced,0.06875648021697997
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,1,1,balanced,0.2510867118835449
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,1,1,balanced,0.2663507270812988
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,1,1,balanced,0.10302847862243651
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,1,1,balanced,0.28333248138427736
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,1,1,balanced,0.16681280136108398
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,1,1,balanced,0.2184992027282715
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,1,1,balanced,0.22141632080078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,1,1,balanced,0.21935295104980468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,1,power_law_1.01,0.05680704116821289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,1,power_law_1.01,0.08293439865112304
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,1,1,balanced,0.43431232452392576
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,1,1,balanced,0.6872390747070313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,1,1,balanced,0.774673309326172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,1,power_law_1.01,0.2930937576293945
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,1,1,balanced,0.4671059036254883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,1,power_law_1.01,0.4516396713256836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,1,power_law_1.01,0.526690559387207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,1,power_law_1.01,0.22149568557739258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,1,1,balanced,1.1879251098632813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,1,1,balanced,1.574097900390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,1,power_law_1.01,0.4695148849487305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,1,power_law_1.01,0.6342272186279297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,1,power_law_1.01,0.13812288284301757
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,1,power_law_1.01,0.4998432159423828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,1,power_law_1.01,0.5196627044677735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,1,power_law_1.01,0.5474835205078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,1,power_law_1.01,0.7041548919677735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,1,power_law_1.01,0.7226636505126953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,1,power_law_1.01,0.5683065414428711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,1,power_law_1.01,0.6008108901977539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,1,power_law_1.01,0.6553945922851563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,1,power_law_1.01,1.301832275390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,1,power_law_1.01,1.5069906616210937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,1,power_law_1.01,1.853166046142578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,1,power_law_1.01,0.8608934020996093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,1,power_law_1.01,2.204136962890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,1,power_law_1.01,0.7779923248291015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,1,1,power_law_1.01,0.043227519989013675
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,1,1,power_law_1.01,0.05691775798797607
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,1,1,power_law_1.01,0.07222911834716797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,1,1,power_law_1.01,0.0967347240447998
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,1,1,power_law_1.01,0.11953215599060059
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,1,power_law_1.01,2.908370056152344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,1,1,power_law_1.01,0.17187135696411132
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,1,power_law_1.01,0.8728832244873047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,1,1,power_law_1.01,0.18055871963500975
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,1,1,power_law_1.01,0.18967552185058595
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,1,1,power_law_1.01,0.1970636749267578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,1,1,power_law_1.01,0.20059839248657227
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,1,1,power_law_1.01,0.20347135543823242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,1,1,power_law_1.01,0.20961599349975585
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,1,1,power_law_1.01,0.21355520248413087
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,1,1,power_law_1.01,0.22172096252441403
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,1,power_law_1.01,0.9536466979980469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,1,1,power_law_1.01,0.22747392654418946
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,1,1,power_law_1.01,0.23322687149047852
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,1,1,power_law_1.01,0.25100799560546877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,1,power_law_1.01,3.6487136840820313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,1,1,power_law_1.01,0.28222911834716796
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,1,1,power_law_1.01,0.31386367797851566
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,1,1,power_law_1.01,0.38299583435058593
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,1,1,power_law_1.01,0.4582265472412109
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,1,1,power_law_1.01,0.60285888671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,1,1,power_law_1.01,0.7686022186279298
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,1,1,power_law_1.01,1.0989933013916016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,1,power_law_1.2,0.2016896057128906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,1,1,power_law_1.01,1.430167694091797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,1,power_law_1.2,0.2722835159301758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,1,power_law_1.2,0.5086105728149415
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,1,1,power_law_1.01,2.1033894348144533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,1,power_law_1.2,0.523130226135254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,1,power_law_1.2,0.5320844650268555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,1,power_law_1.2,0.5598451232910155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,1,power_law_1.2,0.08301759719848632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,1,power_law_1.2,0.5972774505615235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,1,power_law_1.2,0.6591769409179687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,1,power_law_1.2,0.054504318237304686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,1,power_law_1.2,0.47732288360595704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,1,power_law_1.2,0.44890174865722654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,1,power_law_1.2,0.700733413696289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,1,power_law_1.2,0.7539353942871093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,1,power_law_1.2,0.9118515014648437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,1,power_law_1.2,0.8925312042236329
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,1,power_law_1.2,0.9853977966308595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,1,power_law_1.2,1.3901068115234376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,1,power_law_1.2,0.6244992065429688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,1,power_law_1.2,0.13426367759704588
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,1,power_law_1.2,0.7415245056152344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,1,power_law_1.2,0.7910399627685547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,1,1,power_law_1.2,0.04299456119537354
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,1,power_law_1.2,2.2641632080078127
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,1,1,power_law_1.2,0.055751042366027834
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,1,1,power_law_1.2,0.07078656196594238
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,1,1,power_law_1.2,0.09008255958557129
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,1,1,power_law_1.2,0.11269824028015137
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,1,1,power_law_1.2,0.17153215408325195
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,1,1,power_law_1.2,0.18101951599121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,1,1,power_law_1.2,0.19274303436279297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,1,1,power_law_1.2,0.19796415328979494
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,1,1,power_law_1.2,0.20112768173217774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,1,power_law_1.2,2.905439453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,1,1,power_law_1.2,0.20791999816894532
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,1,1,power_law_1.2,0.21551103591918946
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,1,1,power_law_1.2,0.22143808364868164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,1,1,power_law_1.2,0.232031364440918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,1,power_law_1.2,1.5468537902832031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,1,power_law_1.2,1.889475860595703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,1,power_law_1.2,3.614229125976563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,1,1,power_law_1.01,2.799727478027344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,1,1,power_law_1.2,0.25070655822753907
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,1,1,power_law_1.2,0.24056896209716797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,1,1,power_law_1.2,0.2702867126464844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,1,1,power_law_1.2,0.3124716758728027
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,1,1,power_law_1.2,0.3581126403808594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,1,1,power_law_1.2,0.4629292678833008
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,1,1,power_law_1.2,0.5712019348144531
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,1,1,power_law_1.2,0.8099807739257813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,1,balanced,0.8641907501220704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,1,balanced,0.2771091270446777
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,1,balanced,0.4643558502197266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,1,balanced,1.6801446533203126
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,1,1,power_law_1.2,1.060022430419922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,1,1,power_law_1.2,1.5668070983886717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,1,balanced,3.2634988403320313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,1,1,power_law_1.2,2.098663024902344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,1,balanced,4.826768493652343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,1,balanced,4.837474670410156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,1,1,power_law_1.2,3.2120346069335937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,1,balanced,4.88218994140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,1,balanced,4.8902740478515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,1,balanced,4.900634155273438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,1,balanced,4.911233215332031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,1,1,power_law_1.2,4.32905029296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,1,balanced,4.964094848632813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,1,balanced,4.961850891113281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,1,balanced,5.015132141113281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,1,balanced,5.04310791015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,1,balanced,5.728991088867188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,1,balanced,5.41127197265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,1,balanced,5.851261596679687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,1,balanced,7.860721435546876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,1,balanced,6.37119384765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,1,balanced,6.586000366210937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,1,balanced,10.037592773437499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,1,balanced,0.283175048828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,1,balanced,9.36052734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,1,balanced,0.17651519775390626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,1,balanced,0.5082470321655274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,1,balanced,0.9536927795410156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,1,balanced,15.673291015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,1,balanced,2.76541748046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,1,balanced,16.55815185546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,1,balanced,2.743957824707031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,1,balanced,1.871189727783203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,1,balanced,2.817718505859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,1,balanced,2.7612564086914064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,1,balanced,2.8454022216796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,1,balanced,2.7987072753906252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,1,balanced,2.9186105346679687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,1,balanced,2.943580017089844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,1,balanced,2.9568197631835935
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,1,balanced,3.0271954345703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,1,balanced,23.4174658203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,1,balanced,2.823392028808594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,1,balanced,25.34194580078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,1,balanced,2.866629638671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,1,balanced,3.3904684448242186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,1,balanced,3.5192205810546873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,1,balanced,3.7474667358398435
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,1,balanced,3.9673440551757815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,1,balanced,5.9682812499999995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,1,balanced,6.387862548828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,1,balanced,8.957855834960938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,1,balanced,9.620548706054688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,1,balanced,14.582916259765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,1,balanced,17.80101806640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,1,1,balanced,0.16297088623046876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,1,1,balanced,0.16302080154418946
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,1,1,balanced,0.23329536437988282
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,1,1,balanced,0.1644691276550293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,1,1,balanced,0.2472217559814453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,1,1,balanced,0.16748479843139646
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,1,1,balanced,0.27554624557495117
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,1,1,balanced,0.1742835235595703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,1,1,balanced,0.19024768829345703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,1,1,balanced,0.20344703674316406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,1,1,balanced,0.21783168792724608
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,1,1,balanced,0.3348646545410156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,1,1,balanced,0.36554302215576173
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,1,1,balanced,0.42120193481445317
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,1,1,balanced,0.5179097747802734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,1,1,balanced,0.5760332870483398
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,1,1,balanced,0.7346720123291016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,1,1,balanced,1.0488800048828124
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,1,1,balanced,1.365983428955078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,1,1,balanced,1.9957720947265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,1,1,balanced,2.640220947265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,1,1,balanced,3.8984826660156253
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,1,1,balanced,5.169503784179687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,1,power_law_1.01,0.8722271728515626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,1,power_law_1.01,0.2747052764892578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,1,power_law_1.01,1.5267251586914061
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,1,power_law_1.01,2.148951110839844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,1,1,balanced,7.71794921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,1,power_law_1.01,0.4634643173217773
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,1,1,balanced,10.2700048828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,1,power_law_1.01,4.104537048339844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,1,power_law_1.01,3.799578857421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,1,power_law_1.01,3.9805599975585935
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,1,power_law_1.01,4.173102111816407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,1,power_law_1.01,4.570631103515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,1,power_law_1.01,4.313417663574219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,1,power_law_1.01,3.5554324340820314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,1,power_law_1.01,4.467189636230469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,1,power_law_1.01,4.908006286621093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,1,power_law_1.01,5.2851513671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,1,power_law_1.01,6.114758911132812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,1,power_law_1.01,5.70177001953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,1,power_law_1.01,7.2902026367187505
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,1,power_law_1.01,6.2753228759765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,1,power_law_1.01,7.0867156982421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,1,power_law_1.01,7.8517565917968755
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,1,power_law_1.01,9.447565917968749
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,1,power_law_1.01,14.282309570312501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,1,power_law_1.01,11.76677001953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,1,power_law_1.01,16.938482666015624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,1,power_law_1.01,0.5037273788452149
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,1,power_law_1.01,0.17525951385498045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,1,power_law_1.01,0.2841036796569824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,1,power_law_1.01,0.8551570892333984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,1,power_law_1.01,2.3162918090820312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,1,power_law_1.01,22.16031982421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,1,power_law_1.01,1.2694771575927735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,1,power_law_1.01,2.3830195617675782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,1,power_law_1.01,2.0786483764648436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,1,power_law_1.01,2.4493484497070312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,1,power_law_1.01,2.212718048095703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,1,power_law_1.01,2.482183685302734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,1,power_law_1.01,2.575578308105469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,1,power_law_1.01,2.6011444091796876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,1,power_law_1.01,2.6673406982421874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,1,power_law_1.01,27.21902587890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,1,power_law_1.01,2.7257876586914063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,1,power_law_1.01,2.8147738647460936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,1,power_law_1.01,2.9810635375976564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,1,power_law_1.01,3.3743911743164063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,1,power_law_1.01,3.5450042724609374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,1,power_law_1.01,4.020600891113281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,1,power_law_1.01,4.483972473144531
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,1,1,power_law_1.01,0.162161922454834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,1,power_law_1.01,6.447395629882813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,1,power_law_1.01,8.380420532226562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,1,1,power_law_1.01,0.16475200653076172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,1,1,power_law_1.01,0.1623366355895996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,1,1,power_law_1.01,0.16733695983886718
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,1,1,power_law_1.01,0.17521087646484373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,1,power_law_1.01,10.32861328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,1,1,power_law_1.01,0.20339263916015624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,1,1,power_law_1.01,0.2183308792114258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,1,1,power_law_1.01,0.23354047775268555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,1,1,power_law_1.01,0.27496511459350587
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,1,1,power_law_1.01,0.24759103775024416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,1,power_law_1.01,14.29264892578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,1,power_law_1.01,5.430974731445312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,1,1,power_law_1.01,0.4208563232421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,1,power_law_1.01,18.253701171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,1,1,power_law_1.01,0.18894784927368163
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,1,1,power_law_1.01,0.33543296813964846
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,1,1,power_law_1.01,0.36427070617675783
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,1,1,power_law_1.01,0.5180902481079102
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,1,1,power_law_1.01,0.5755744171142578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,1,1,power_law_1.01,1.0464230346679688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,1,1,power_law_1.01,0.7328243255615234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,1,1,power_law_1.01,1.368270721435547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,1,1,power_law_1.01,1.9955775451660156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,1,1,power_law_1.01,2.635623779296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,1,1,power_law_1.01,3.9030572509765626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,1,1,power_law_1.01,5.181975097656251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,1,power_law_1.2,0.46257278442382815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,1,power_law_1.2,0.8725926208496094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,1,1,power_law_1.01,7.689957275390626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,1,power_law_1.2,1.3939468383789062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,1,power_law_1.2,3.5017599487304687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,1,power_law_1.2,0.2705331230163574
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,1,1,power_law_1.01,10.232266845703126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,1,power_law_1.2,1.912200927734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,1,power_law_1.2,3.934603576660156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,1,power_law_1.2,4.1060461425781245
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,1,power_law_1.2,3.7987640380859373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,1,power_law_1.2,4.62822265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,1,power_law_1.2,4.877536010742188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,1,power_law_1.2,5.120991821289063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,1,power_law_1.2,4.240574035644531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,1,power_law_1.2,4.4098181152343745
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,1,power_law_1.2,5.899445800781249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,1,power_law_1.2,5.368755493164063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,1,power_law_1.2,6.272664184570313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,1,power_law_1.2,7.546151733398437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,1,power_law_1.2,6.475348510742188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,1,power_law_1.2,7.274769897460938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,1,power_law_1.2,8.160087890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,1,power_law_1.2,12.23656494140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,1,power_law_1.2,0.17565183639526366
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,1,power_law_1.2,9.598565063476562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,1,power_law_1.2,0.28264768600463863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,1,power_law_1.2,0.5035417556762696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,1,power_law_1.2,14.709493408203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,1,power_law_1.2,2.0305690002441406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,1,power_law_1.2,0.7880499267578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,1,power_law_1.2,1.1265382385253906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,1,power_law_1.2,2.207153930664062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,1,power_law_1.2,2.278803253173828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,1,power_law_1.2,2.3790879821777344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,1,power_law_1.2,2.485130920410156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,1,power_law_1.2,2.5483251953124997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,1,power_law_1.2,2.5997018432617187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,1,power_law_1.2,17.100865478515622
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,1,power_law_1.2,2.6524197387695314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,1,power_law_1.2,2.7974298095703127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,1,power_law_1.2,2.862049255371094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,1,power_law_1.2,2.917922058105469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,1,power_law_1.2,3.10745849609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,1,power_law_1.2,3.4833752441406247
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,1,power_law_1.2,3.7071563720703127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,1,power_law_1.2,4.120334777832031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,1,power_law_1.2,4.638799438476562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,1,power_law_1.2,5.6005401611328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,1,1,power_law_1.2,0.1628883171081543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,1,power_law_1.2,6.516999511718749
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,1,power_law_1.2,22.33419677734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,1,1,power_law_1.2,0.16342784881591796
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,1,1,power_law_1.2,0.1645721626281738
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,1,1,power_law_1.2,0.1904249572753906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,1,1,power_law_1.2,0.17517440795898437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,1,1,power_law_1.2,0.20448896408081058
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,1,1,power_law_1.2,0.21881280899047853
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,1,power_law_1.2,8.55833984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,1,power_law_1.2,27.6705615234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,1,1,power_law_1.2,0.2337785530090332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,1,1,power_law_1.2,0.24752319335937498
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,1,1,power_law_1.2,0.3350048065185547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,1,1,power_law_1.2,0.3639583969116211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,1,power_law_1.2,10.4882666015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,1,1,power_law_1.2,0.5174803161621094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,1,power_law_1.2,14.197515869140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,1,1,power_law_1.2,0.16722496032714845
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,1,1,power_law_1.2,0.2747871971130371
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,1,1,power_law_1.2,0.4213727951049805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,1,power_law_1.2,18.185457763671874
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,1,1,power_law_1.2,0.5759737777709961
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,1,1,power_law_1.2,0.7335417938232422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,1,1,power_law_1.2,1.0471814727783202
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,1,1,power_law_1.2,1.3654425048828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,1,1,power_law_1.2,5.1714196777343755
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,1,2,balanced,0.162105598449707
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,1,1,power_law_1.2,7.690360107421876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,1,2,balanced,0.21827199935913083
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,1,1,power_law_1.2,1.9990226745605468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,1,2,balanced,0.41304065704345705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,1,2,balanced,0.41473152160644533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,1,2,balanced,0.4199436950683594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,1,2,balanced,0.40957183837890626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,1,2,balanced,0.42001918792724613
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,1,2,balanced,0.4240153503417969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,1,2,balanced,0.42622848510742184
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,1,2,balanced,0.4270195388793946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,1,2,balanced,0.3123238372802734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,1,2,balanced,0.43332481384277344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,1,2,balanced,0.43753601074218745
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,1,2,balanced,0.5669247817993164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,1,2,balanced,0.5620415878295899
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,1,2,balanced,0.5692927932739258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,1,2,balanced,0.7671155548095703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,1,2,balanced,0.9551718139648437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,1,2,balanced,1.381966094970703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,1,2,balanced,0.44871551513671876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,1,2,balanced,1.785103302001953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,1,1,power_law_1.2,2.6365005493164064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,1,2,balanced,2.8010302734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,1,2,balanced,3.946369323730469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,1,1,power_law_1.2,3.909328002929688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,1,2,balanced,6.311417846679687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,1,2,balanced,0.11838080406188964
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,1,2,balanced,8.700772094726563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,1,2,balanced,0.1847871971130371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,1,2,balanced,0.1826278305053711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,1,2,balanced,0.09007488250732422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,1,2,balanced,0.18689023971557617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,1,2,balanced,0.18797824859619142
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,1,2,balanced,0.18782079696655274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,1,2,balanced,0.18937215805053711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,1,2,balanced,0.19196672439575196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,1,2,balanced,0.1939481544494629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,1,2,balanced,0.19524480819702147
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,1,2,balanced,0.17200767517089843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,1,2,balanced,0.19762687683105468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,1,2,balanced,0.20628223419189454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,1,2,balanced,0.21063295364379883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,1,2,balanced,0.21548543930053712
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,1,2,balanced,0.21458303451538088
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,1,2,balanced,0.27405824661254885
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,1,2,balanced,0.2955878448486328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,1,2,balanced,0.4167244720458984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,1,2,balanced,0.49780094146728515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,1,2,balanced,13.823787841796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,1,2,balanced,0.7608000183105469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,1,2,balanced,0.978519058227539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,1,2,balanced,1.4903155517578124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,1,2,balanced,2.9337841796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,1,2,balanced,3.895459899902344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,1,1,power_law_1.2,10.238733520507813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,1,2,balanced,19.824644775390624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,1,2,balanced,1.9065638732910155
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,1,2,balanced,0.04226687908172607
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,1,2,balanced,0.09069439888000488
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,1,2,balanced,0.09077887535095215
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,1,2,balanced,0.09110527992248535
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,1,2,balanced,0.059773440361022945
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,1,2,balanced,0.0907750415802002
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,1,2,balanced,0.09135999679565429
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,1,2,balanced,0.0914419174194336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,1,2,balanced,0.09451135635375976
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,1,2,balanced,0.09375103950500488
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,1,2,balanced,0.09524736404418946
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,1,2,balanced,0.09634816169738769
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,1,2,balanced,0.09894911766052246
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,1,2,balanced,0.08923904418945312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,1,2,balanced,0.12669952392578127
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,1,2,balanced,0.09008640289306641
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,1,2,balanced,0.09004544258117676
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,1,2,balanced,0.14919424057006836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,1,2,balanced,0.1413094425201416
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,1,2,balanced,0.20645120620727536
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,1,2,power_law_1.01,0.20244991302490237
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,1,2,power_law_1.01,0.228590087890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,1,2,balanced,0.3683340835571289
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,1,2,balanced,0.2576255989074707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,1,2,power_law_1.01,0.2682649612426758
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,1,2,balanced,0.4702016067504883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,1,2,power_law_1.01,0.41024768829345704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,1,2,power_law_1.01,0.42140926361083986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,1,2,power_law_1.01,0.4553036880493164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,1,2,power_law_1.01,0.465082893371582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,1,2,power_law_1.01,0.41342208862304686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,1,2,power_law_1.01,0.4696076965332031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,1,2,power_law_1.01,0.48206718444824215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,1,2,power_law_1.01,0.515464973449707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,1,2,power_law_1.01,0.5692287826538086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,1,2,power_law_1.01,0.6218112182617188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,1,2,power_law_1.01,0.6658278656005859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,1,2,power_law_1.01,0.41733249664306643
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,1,2,power_law_1.01,0.80447998046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,1,2,power_law_1.01,0.409172477722168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,1,2,power_law_1.01,1.1510514831542968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,1,2,power_law_1.01,1.314661102294922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,1,2,power_law_1.01,1.9534757995605467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,1,2,power_law_1.01,2.755810546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,1,2,power_law_1.01,3.507516174316406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,1,2,power_law_1.01,5.370999145507812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,1,2,power_law_1.01,7.884910278320312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,1,2,power_law_1.01,0.11529472351074219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,1,2,power_law_1.01,0.12584832191467285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,1,2,power_law_1.01,0.13698944091796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,1,2,power_law_1.01,0.18423551559448242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,1,2,power_law_1.01,0.18423423767089844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,1,2,power_law_1.01,0.187139835357666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,1,2,power_law_1.01,0.18881664276123047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,1,2,power_law_1.01,0.18901887893676758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,1,2,power_law_1.01,0.1893401527404785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,1,2,power_law_1.01,11.781197509765624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,1,2,balanced,1.5082456970214844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,1,2,power_law_1.01,0.19145984649658204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,1,2,power_law_1.01,0.20416383743286132
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,1,2,power_law_1.01,0.21247743606567382
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,1,2,power_law_1.01,0.22886783599853516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,1,2,power_law_1.01,0.2346918487548828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,1,2,balanced,0.6971942138671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,1,2,balanced,0.9123468780517578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,1,2,power_law_1.01,0.2563699150085449
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,1,2,power_law_1.01,0.5038886260986328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,1,2,balanced,1.97299072265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,1,2,power_law_1.01,0.6458137512207032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,1,2,power_law_1.01,19.98124267578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,1,2,power_law_1.01,0.8178701019287109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,1,2,power_law_1.01,0.3178598403930664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,1,2,power_law_1.01,0.19828224182128906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,1,2,power_law_1.01,0.36979328155517577
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,1,2,power_law_1.01,1.275662078857422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,1,2,power_law_1.01,0.06307456016540527
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,1,2,power_law_1.01,0.07410560131072999
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,1,2,power_law_1.01,0.09039487838745117
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,1,2,power_law_1.01,0.0902899169921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,1,2,power_law_1.01,0.08693375587463378
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,1,2,power_law_1.01,0.0906265640258789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,1,2,power_law_1.01,0.09112319946289063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,1,2,power_law_1.01,0.09171199798583984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,1,2,power_law_1.01,0.09210368156433105
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,1,2,power_law_1.01,0.09291775703430176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,1,2,power_law_1.01,2.519866943359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,1,2,power_law_1.01,0.11647232055664063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,1,2,power_law_1.01,0.1959846305847168
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,1,2,power_law_1.01,0.11680383682250976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,1,2,power_law_1.01,1.7974386596679686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,1,2,power_law_1.01,3.81362060546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,1,2,power_law_1.01,25.944931640624997
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,1,2,power_law_1.01,0.21953792572021485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,1,2,power_law_1.01,4.825710144042969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,1,2,power_law_1.01,0.0485811185836792
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,1,2,power_law_1.01,0.10546560287475586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,1,2,power_law_1.01,0.12870400428771972
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,1,2,power_law_1.01,0.126876163482666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,1,2,power_law_1.01,0.1468505573272705
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,1,2,power_law_1.01,0.2005081558227539
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,1,2,power_law_1.01,0.31356927871704104
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,1,2,power_law_1.2,0.1848294448852539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,1,2,power_law_1.2,0.23029504776000978
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,1,2,power_law_1.2,0.2692403221130371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,1,2,power_law_1.2,0.409431037902832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,1,2,power_law_1.2,0.41098751068115236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,1,2,power_law_1.2,0.4134848022460938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,1,2,power_law_1.2,0.41999870300292963
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,1,2,power_law_1.2,0.4252211380004883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,1,2,power_law_1.2,0.45667072296142575
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,1,2,power_law_1.2,0.46488193511962894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,1,2,power_law_1.2,0.4702105712890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,1,2,power_law_1.2,0.48680576324462893
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,1,2,power_law_1.2,0.500613136291504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,1,2,power_law_1.2,0.5693350219726563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,1,2,power_law_1.2,0.618716163635254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,1,2,power_law_1.2,0.6735731506347656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,1,2,power_law_1.2,0.8341887664794921
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,1,2,power_law_1.01,0.3683494567871094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,1,2,power_law_1.2,1.1535782623291015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,1,2,power_law_1.01,0.5230220794677735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,1,2,power_law_1.2,2.1831257629394534
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,1,2,power_law_1.2,1.185560302734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,1,2,power_law_1.2,2.4176013183593748
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,1,2,power_law_1.01,0.6824934387207031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,1,2,power_law_1.01,1.9714828491210938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,1,2,power_law_1.2,3.6216513061523434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,1,2,power_law_1.2,5.474872436523437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,1,2,power_law_1.2,0.1187827205657959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,1,2,power_law_1.2,0.11696895599365234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,1,2,power_law_1.01,1.0861132812499998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,1,2,power_law_1.2,0.19415807723999023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,1,2,power_law_1.2,0.19025152206420898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,1,2,power_law_1.2,0.18869375228881835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,1,2,power_law_1.2,0.18851072311401368
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,1,2,power_law_1.2,0.19064191818237303
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,1,2,power_law_1.01,1.8273330688476563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,1,2,power_law_1.2,0.19088384628295899
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,1,2,power_law_1.2,0.20999040603637695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,1,2,power_law_1.2,0.19555583953857422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,1,2,power_law_1.2,0.2063884735107422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,1,2,power_law_1.2,0.2034150314331055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,1,2,power_law_1.2,0.21271167755126955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,1,2,power_law_1.2,7.835595703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,1,2,power_law_1.2,0.22810752868652343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,1,2,power_law_1.2,0.23507968902587892
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,1,2,power_law_1.2,0.35096576690673825
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,1,2,power_law_1.2,0.2536460876464844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,1,2,power_law_1.2,0.3950681686401367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,1,2,power_law_1.2,0.4917683029174805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,1,2,power_law_1.2,0.13167360305786133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,1,2,power_law_1.2,0.6524966430664063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,1,2,power_law_1.2,11.200982666015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,1,2,power_law_1.2,1.0528268432617187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,1,2,power_law_1.2,1.2586662292480468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,1,2,power_law_1.2,1.8266586303710937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,1,2,power_law_1.2,0.0497651195526123
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,1,2,power_law_1.2,0.05874944210052491
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,1,2,power_law_1.2,0.070731520652771
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,1,2,power_law_1.2,0.09208831787109376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,1,2,power_law_1.2,0.08988800048828124
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,1,2,power_law_1.2,0.09107583999633789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,1,2,power_law_1.2,0.09075712203979493
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,1,2,power_law_1.2,0.09195775985717773
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,1,2,power_law_1.2,0.0919654369354248
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,1,2,power_law_1.2,0.09303936004638672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,1,2,power_law_1.2,0.0936793613433838
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,1,2,power_law_1.2,0.10254591941833495
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,1,2,power_law_1.2,0.11702655792236327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,1,2,power_law_1.2,2.6915609741210935
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,1,2,power_law_1.01,2.771101379394531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,1,2,power_law_1.2,20.026810302734376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,1,2,power_law_1.2,3.902113342285156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,1,2,power_law_1.2,26.20279052734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,1,2,power_law_1.2,0.3295654296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,1,2,power_law_1.2,0.11361663818359376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,1,2,power_law_1.2,0.12672512054443358
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,1,2,power_law_1.2,4.770055541992187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,1,2,power_law_1.2,0.12979328155517578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,1,2,power_law_1.2,0.9785958099365235
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,1,2,power_law_1.2,0.1453388786315918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,1,2,power_law_1.2,0.20316415786743164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,2,1,balanced,0.12031359672546386
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,2,1,balanced,0.17945856094360352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,2,1,balanced,0.27671295166015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,2,1,balanced,0.2800089645385742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,2,1,balanced,0.32089279174804686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,2,1,balanced,0.32247104644775393
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,2,1,balanced,0.32432510375976564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,2,1,balanced,0.3258163070678711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,2,1,balanced,0.3270585632324219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,2,1,balanced,0.32804161071777344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,2,1,balanced,0.32981952667236325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,2,1,balanced,0.33253822326660154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,2,1,balanced,0.3352012634277344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,2,1,balanced,0.3462188720703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,2,1,balanced,0.45528640747070315
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,2,1,balanced,0.48486785888671874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,2,1,balanced,0.4953644943237305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,1,2,power_law_1.2,0.2294963264465332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,2,1,balanced,0.7347519683837891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,2,1,balanced,0.8741107177734374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,2,1,balanced,1.3327462768554688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,2,1,balanced,1.6227462768554688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,2,1,balanced,2.556291198730469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,2,1,balanced,3.318215637207031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,2,1,balanced,4.942400512695313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,1,2,power_law_1.2,2.7271807861328123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,2,1,balanced,6.967365112304688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,2,1,balanced,0.0772223997116089
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,2,1,balanced,0.10382783889770508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,2,1,balanced,0.1584620761871338
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,2,1,balanced,0.15845824241638184
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,2,1,balanced,0.17015552520751953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,2,1,balanced,0.17087615966796874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,2,1,balanced,0.17296895980834961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,2,1,balanced,0.17355775833129883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,2,1,balanced,0.17507392883300782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,2,1,balanced,0.17475519180297852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,2,1,balanced,0.17785728454589844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,2,1,balanced,0.17760255813598633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,2,1,balanced,0.1806342315673828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,2,1,balanced,0.18361471176147462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,2,1,balanced,0.19059072494506837
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,2,1,balanced,0.1936960029602051
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,2,1,balanced,0.19365760803222656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,2,1,balanced,0.2601363182067871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,2,1,balanced,0.2741779136657715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,2,1,balanced,0.3729958343505859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,2,1,balanced,0.4725747299194336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,2,1,balanced,0.7052973175048829
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,2,1,balanced,11.92056396484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,2,1,balanced,0.9187897491455079
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,1,2,power_law_1.2,0.40668544769287107
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,2,1,balanced,1.3987750244140624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,1,2,power_law_1.2,0.5836492919921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,2,1,balanced,1.8121542358398437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,2,1,balanced,2.6200506591796873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,2,1,balanced,3.5194784545898434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,2,1,balanced,17.16797607421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,1,2,power_law_1.2,0.6267750549316407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,1,2,power_law_1.2,1.5471731567382814
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,1,2,power_law_1.2,2.143057861328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,2,1,balanced,0.0361075210571289
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,2,1,balanced,0.09221759796142578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,2,1,balanced,0.09250304222106934
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,2,1,balanced,0.05766592025756836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,2,1,balanced,0.09227392196655274
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,2,1,balanced,0.09251584053039551
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,2,1,balanced,0.09535103797912597
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,2,1,balanced,0.09477888107299805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,2,1,balanced,0.09643391609191895
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,2,1,balanced,0.09867327690124512
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,2,1,balanced,0.10119423866271973
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,2,1,balanced,0.09162752151489258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,2,1,balanced,0.09105536460876465
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,2,1,balanced,0.09181695938110351
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,2,1,balanced,0.09275456428527831
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,2,1,balanced,0.09254143714904786
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,2,1,balanced,0.09229503631591797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,2,1,balanced,0.1369497585296631
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,2,1,balanced,0.14408639907836912
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,2,1,balanced,0.19487360000610351
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,2,1,balanced,0.24638656616210936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,2,1,power_law_1.01,0.17771520614624023
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,2,1,balanced,0.3534585571289063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,2,1,power_law_1.01,0.11851008415222168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,2,1,power_law_1.01,0.21375040054321287
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,2,1,power_law_1.01,0.27855615615844725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,2,1,power_law_1.01,0.3211705780029297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,2,1,power_law_1.01,0.3244364929199219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,2,1,power_law_1.01,0.3273894500732422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,2,1,power_law_1.01,0.32295745849609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,2,1,balanced,0.45939071655273434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,2,1,power_law_1.01,0.32974334716796877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,2,1,power_law_1.01,0.3426163101196289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,2,1,power_law_1.01,0.3747654342651367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,2,1,power_law_1.01,0.380761604309082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,2,1,power_law_1.01,0.3903744125366211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,2,1,power_law_1.01,0.4223513412475586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,2,1,power_law_1.01,0.4580006408691406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,2,1,power_law_1.01,0.5337702560424804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,2,1,power_law_1.01,0.6156844711303711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,2,1,power_law_1.01,0.7760691070556641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,2,1,power_law_1.01,0.9786399841308594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,2,1,power_law_1.01,1.3951770019531249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,2,1,power_law_1.01,1.7816217041015627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,2,1,power_law_1.01,2.5729159545898437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,2,1,power_law_1.01,3.3376293945312496
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,2,1,balanced,0.6960313415527344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,2,1,power_law_1.01,0.0778278398513794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,2,1,power_law_1.01,0.10124416351318359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,2,1,power_law_1.01,0.11099264144897461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,2,1,power_law_1.01,0.15601344108581544
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,2,1,power_law_1.01,0.1699180793762207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,2,1,power_law_1.01,0.17149375915527343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,2,1,power_law_1.01,0.17291391372680665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,2,1,power_law_1.01,0.1702079963684082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,2,1,power_law_1.01,0.1705094337463379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,2,1,power_law_1.01,0.17554752349853514
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,2,1,power_law_1.01,0.1777529525756836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,2,1,power_law_1.01,0.17529279708862305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,2,1,power_law_1.01,0.18114240646362306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,2,1,power_law_1.01,5.086056213378906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,2,1,power_law_1.01,0.19572160720825196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,2,1,power_law_1.01,0.20816640853881835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,2,1,power_law_1.01,0.21527936935424802
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,2,1,power_law_1.01,0.22367872238159178
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,2,1,power_law_1.01,0.26291776657104493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,2,1,power_law_1.01,0.3176851272583008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,2,1,power_law_1.01,0.4112345504760742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,2,1,balanced,0.9553440093994141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,2,1,power_law_1.01,0.509760627746582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,2,1,power_law_1.01,6.804226074218749
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,2,1,power_law_1.01,0.7093292999267578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,2,1,balanced,1.5011167907714844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,2,1,balanced,1.9489056396484377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,2,1,power_law_1.01,0.927675552368164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,2,1,power_law_1.01,0.04371840000152588
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,2,1,power_law_1.01,0.0582041597366333
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,2,1,power_law_1.01,0.07126207828521729
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,2,1,power_law_1.01,0.09052607536315918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,2,1,power_law_1.01,0.09168512344360351
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,2,1,power_law_1.01,0.09169919967651367
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,2,1,power_law_1.01,0.09148032188415527
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,2,1,power_law_1.01,0.09037311553955078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,2,1,power_law_1.01,0.09086784362792968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,2,1,power_law_1.01,0.09282112121582031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,2,1,power_law_1.01,0.09362560272216797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,2,1,power_law_1.01,0.09549311637878419
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,2,1,power_law_1.01,0.10277055740356444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,2,1,power_law_1.01,1.410592041015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,2,1,power_law_1.01,1.8443872070312501
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,2,1,power_law_1.01,0.1251807975769043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,2,1,power_law_1.01,10.849840087890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,2,1,power_law_1.01,2.687334289550781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,2,1,power_law_1.01,0.23696895599365236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,2,1,power_law_1.01,3.537342224121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,2,1,power_law_1.01,15.45453369140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,2,1,power_law_1.01,0.4149343872070313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,2,1,power_law_1.01,0.09312128067016602
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,2,1,power_law_1.01,0.10551487922668457
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,2,1,power_law_1.01,0.11125503540039064
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,2,1,power_law_1.01,0.15244288444519044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,2,1,power_law_1.2,0.11957375526428222
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,2,1,power_law_1.2,0.16689855575561524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,2,1,power_law_1.2,0.18869504928588868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,2,1,power_law_1.2,0.27432191848754883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,2,1,power_law_1.2,0.3197747230529785
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,2,1,power_law_1.01,0.1815558433532715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,2,1,power_law_1.2,0.32863487243652345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,2,1,power_law_1.2,0.3262604904174805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,2,1,power_law_1.2,0.3286528015136719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,2,1,power_law_1.2,0.32989887237548826
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,2,1,power_law_1.2,0.3465203094482422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,2,1,power_law_1.2,0.3774905776977539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,2,1,power_law_1.2,0.3804230499267578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,2,1,power_law_1.2,0.3873017501831055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,2,1,power_law_1.2,0.41917312622070313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,2,1,power_law_1.2,0.45200576782226565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,2,1,power_law_1.2,0.5499654388427734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,2,1,power_law_1.2,0.6157043075561524
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,2,1,power_law_1.01,0.29300607681274415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,2,1,power_law_1.2,0.7878828430175782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,2,1,power_law_1.2,0.9794624328613282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,2,1,power_law_1.2,1.3474598693847657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,2,1,power_law_1.2,1.7644703674316407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,2,1,power_law_1.01,0.5368352127075195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,2,1,power_law_1.2,2.5637905883789065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,2,1,power_law_1.2,3.311802978515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,2,1,power_law_1.01,0.8077894592285156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,2,1,power_law_1.2,0.07977407932281494
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,2,1,power_law_1.2,0.09921216011047364
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,2,1,power_law_1.2,0.10444479942321779
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,2,1,power_law_1.2,0.15439040184020997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,2,1,power_law_1.2,0.16070207595825195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,2,1,power_law_1.2,0.17079296112060546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,2,1,power_law_1.2,0.17186304092407229
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,2,1,power_law_1.2,0.16921279907226563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,2,1,power_law_1.2,0.17392000198364257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,2,1,power_law_1.2,0.1742393684387207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,2,1,power_law_1.2,0.17673343658447266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,2,1,power_law_1.2,0.17783424377441406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,2,1,power_law_1.2,0.17771711349487304
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,2,1,power_law_1.01,1.0408038330078124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,2,1,power_law_1.2,0.1950899124145508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,2,1,power_law_1.2,0.21432256698608398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,2,1,power_law_1.2,0.22098751068115235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,2,1,power_law_1.2,0.22169408798217774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,2,1,power_law_1.2,0.2619897651672363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,2,1,power_law_1.2,0.31660224914550783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,2,1,power_law_1.2,5.005201416015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,2,1,power_law_1.2,0.41072830200195315
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,2,1,power_law_1.2,0.5079097747802734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,2,1,power_law_1.01,1.6076287841796877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,2,1,power_law_1.2,0.6944953918457031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,2,1,power_law_1.2,0.9384326171874999
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,2,1,power_law_1.2,6.780712280273437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,2,1,power_law_1.2,0.040108160972595217
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,2,1,power_law_1.01,2.1813075256347654
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,2,1,power_law_1.2,0.05279744148254395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,2,1,power_law_1.2,0.06081984043121338
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,2,1,power_law_1.2,0.09012543678283691
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,2,1,power_law_1.2,0.08951168060302735
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,2,1,power_law_1.2,0.09189824104309083
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,2,1,power_law_1.2,0.09245439529418945
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,2,1,power_law_1.2,0.0907027244567871
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,2,1,power_law_1.2,0.09259327888488769
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,2,1,power_law_1.2,0.09312447547912597
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,2,1,power_law_1.2,0.09430272102355956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,2,1,power_law_1.2,1.3763885498046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,2,1,power_law_1.2,0.10684224128723145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,2,1,power_law_1.2,1.819983367919922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,2,1,power_law_1.2,0.11657471656799316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,2,1,power_law_1.2,11.010477294921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,2,1,power_law_1.2,0.1550387191772461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,2,1,power_law_1.2,0.1833203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,2,1,power_law_1.2,2.7164422607421876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,2,1,power_law_1.2,0.09520000457763672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,2,1,power_law_1.2,0.09514944076538086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,2,1,power_law_1.2,0.10859840393066407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,2,1,power_law_1.2,15.35468017578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,2,1,power_law_1.2,0.12399423599243165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,2,1,power_law_1.2,3.5480673217773435
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,2,1,power_law_1.2,0.24156864166259764
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,1,2,balanced,0.23393407821655274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,1,2,balanced,0.32827903747558596
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,1,2,balanced,0.4986521530151367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,1,2,balanced,0.628746223449707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,1,2,balanced,0.6289088058471679
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,1,2,balanced,0.6292339324951172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,1,2,balanced,0.6468428802490235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,1,2,balanced,0.6491276550292968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,1,2,balanced,0.6425151824951172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,1,2,balanced,0.6555404663085938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,1,2,balanced,0.6298150253295899
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,1,2,balanced,0.6237273788452148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,1,2,balanced,0.6286681747436523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,1,2,balanced,0.6571968078613282
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,2,1,power_law_1.2,0.29899648666381834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,1,2,balanced,0.9406835174560546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,1,2,balanced,0.9224742126464843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,1,2,balanced,0.9035008239746094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,1,2,balanced,1.3520204162597655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,1,2,balanced,1.5867404174804687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,1,2,balanced,2.3274943542480466
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,2,1,power_law_1.2,0.42241470336914066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,1,2,balanced,3.0174719238281247
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,2,1,power_law_1.2,0.5411379241943359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,1,2,balanced,5.0464807128906255
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,1,2,balanced,6.968276367187499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,1,2,balanced,10.870137939453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,1,2,balanced,0.1119820785522461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,1,2,balanced,0.1602367973327637
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,2,1,power_law_1.2,1.7311001586914063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,2,1,power_law_1.2,0.8330419158935547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,1,2,balanced,0.2889612770080566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,1,2,balanced,0.27456384658813476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,1,2,balanced,0.2665382385253906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,1,2,balanced,0.26863616943359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,1,2,balanced,0.26933376312255863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,1,2,balanced,0.2718272018432617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,1,2,balanced,0.27196287155151366
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,1,2,balanced,0.27715967178344725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,1,2,balanced,0.2799910354614258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,1,2,balanced,0.2822080039978027
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,1,2,balanced,0.29428096771240236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,1,2,balanced,0.2986163139343262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,1,2,balanced,0.30562175750732423
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,1,2,balanced,0.3079257583618164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,1,2,balanced,0.43080703735351566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,1,2,balanced,0.46015361785888675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,1,2,balanced,0.5920652770996094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,1,2,balanced,15.021373291015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,1,2,balanced,0.7916710662841797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,1,2,balanced,1.1806566619873047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,1,2,balanced,1.5689651489257812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,1,2,balanced,0.2510361671447754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,1,2,balanced,2.2879231262207034
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,1,2,balanced,3.047016906738281
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,2,1,power_law_1.2,1.075720977783203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,1,2,balanced,4.5450918579101565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,1,2,balanced,5.830247802734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,1,2,balanced,23.45931884765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,1,2,balanced,33.664584960937496
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,2,1,power_law_1.2,2.2136448669433593
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,1,2,balanced,0.08121600151062011
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,1,2,balanced,0.05584383964538574
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,1,2,balanced,0.1421337604522705
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,1,2,balanced,0.1423859214782715
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,1,2,balanced,0.14507519721984863
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,1,2,balanced,0.14012672424316405
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,1,2,balanced,0.144651517868042
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,1,2,balanced,0.14605183601379396
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,1,2,balanced,0.1476863956451416
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,1,2,balanced,0.14766207695007325
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,1,2,balanced,0.15173248291015623
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,1,2,balanced,0.14212991714477538
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,1,2,balanced,0.14040575981140138
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,1,2,balanced,0.14117631912231446
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,1,2,balanced,0.14163711547851562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,1,2,balanced,0.1422003173828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,1,2,balanced,0.14235008239746094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,1,2,balanced,0.34040958404541016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,1,2,balanced,0.24691455841064452
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,1,2,balanced,0.2381439971923828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,1,2,power_law_1.01,0.2545907211303711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,1,2,power_law_1.01,0.3190323257446289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,1,2,power_law_1.01,0.3557376098632813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,1,2,balanced,0.4545369720458984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,1,2,power_law_1.01,0.6082035064697265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,1,2,power_law_1.01,0.6461811065673828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,1,2,power_law_1.01,0.6416063690185547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,1,2,power_law_1.01,0.6518246459960937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,1,2,power_law_1.01,0.6563852691650391
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,1,2,balanced,0.6266150283813476
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,1,2,balanced,0.8792716979980468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,1,2,power_law_1.01,0.6597990417480469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,1,2,power_law_1.01,0.6604390716552734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,1,2,power_law_1.01,0.6837324523925782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,1,2,power_law_1.01,0.7040460968017579
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,1,2,power_law_1.01,0.7561305236816407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,1,2,power_law_1.01,0.876842269897461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,1,2,power_law_1.01,1.024664306640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,1,2,power_law_1.01,1.1193612670898436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,1,2,power_law_1.01,1.3136717224121095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,1,2,power_law_1.01,1.824204864501953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,1,2,power_law_1.01,1.9103269958496092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,1,2,power_law_1.01,3.232265014648438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,1,2,power_law_1.01,4.317498779296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,1,2,power_law_1.01,6.058689575195312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,1,2,balanced,1.1878195190429688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,1,2,power_law_1.01,7.771652832031251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,1,2,power_law_1.01,0.11152640342712403
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,1,2,power_law_1.01,0.14286463737487792
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,1,2,power_law_1.01,0.16065408706665038
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,1,2,power_law_1.01,15.916812744140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,1,2,power_law_1.01,0.26810752868652343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,1,2,power_law_1.01,0.26531328201293947
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,1,2,power_law_1.01,0.25745023727416994
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,1,2,power_law_1.01,0.26878463745117187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,1,2,power_law_1.01,0.27043071746826175
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,1,2,power_law_1.01,0.2725350379943848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,1,2,power_law_1.01,0.2749951934814453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,1,2,power_law_1.01,0.2841049575805664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,1,2,power_law_1.01,0.2898944091796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,1,2,power_law_1.01,0.3041753578186035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,1,2,power_law_1.01,0.35547393798828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,1,2,power_law_1.01,0.36448768615722654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,1,2,power_law_1.01,0.4159014511108398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,1,2,power_law_1.01,0.47882751464843754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,1,2,power_law_1.01,0.26034175872802734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,1,2,power_law_1.01,18.212403564453126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,1,2,power_law_1.01,0.5913689422607422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,1,2,power_law_1.01,0.8738598632812501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,1,2,power_law_1.01,1.0715686035156249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,1,2,power_law_1.01,1.5553919982910156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,1,2,power_law_1.01,2.18123779296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,1,2,balanced,1.7198144531250001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,1,2,power_law_1.01,3.026578063964844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,1,2,power_law_1.01,0.06891392230987549
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,1,2,power_law_1.01,0.08464896202087402
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,1,2,power_law_1.01,0.09253760337829589
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,1,2,power_law_1.01,0.13406847953796386
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,1,2,power_law_1.01,0.14048383712768556
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,1,2,power_law_1.01,0.1413094425201416
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,1,2,power_law_1.01,0.1409817600250244
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,1,2,power_law_1.01,0.14161279678344726
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,1,2,power_law_1.01,0.14187647819519042
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,1,2,power_law_1.01,0.14279295921325683
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,1,2,power_law_1.01,0.14350720405578615
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,1,2,power_law_1.01,4.409964904785157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,1,2,power_law_1.01,28.64899658203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,1,2,power_law_1.01,6.040835571289063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,1,2,power_law_1.01,0.16667264938354492
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,1,2,balanced,2.504779510498047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,1,2,balanced,3.2961856079101564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,1,2,power_law_1.01,9.813316040039062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,1,2,power_law_1.01,43.9918798828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,1,2,power_law_1.01,0.5144051361083984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,1,2,power_law_1.01,0.14789631843566894
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,1,2,power_law_1.01,0.15730560302734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,1,2,power_law_1.01,0.21682815551757811
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,1,2,power_law_1.01,0.2036800003051758
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,1,2,power_law_1.01,0.2544934463500977
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,1,2,power_law_1.01,0.29130111694335936
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,1,2,power_law_1.01,0.38094463348388674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,1,2,power_law_1.2,0.24167039871215817
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,1,2,power_law_1.2,0.3035494422912598
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,1,2,power_law_1.2,0.3311283111572266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,1,2,power_law_1.2,0.61198974609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,1,2,power_law_1.2,0.632552947998047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,1,2,power_law_1.2,0.6399103927612304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,1,2,power_law_1.2,0.6517913818359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,1,2,power_law_1.2,0.6566130828857422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,1,2,power_law_1.2,0.6583551788330079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,1,2,power_law_1.2,0.6593574523925781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,1,2,power_law_1.2,0.6649868774414063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,1,2,power_law_1.2,0.6910387420654297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,1,2,power_law_1.2,0.748828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,1,2,power_law_1.2,0.889977569580078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,1,2,power_law_1.2,1.0265177917480468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,1,2,power_law_1.2,1.1478527832031251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,1,2,power_law_1.2,1.3192729187011718
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,1,2,power_law_1.01,0.6580710601806641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,1,2,power_law_1.2,1.9560383605957032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,1,2,power_law_1.2,2.2983538818359373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,1,2,power_law_1.2,3.295397033691406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,1,2,power_law_1.2,4.025472106933593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,1,2,power_law_1.2,6.8635888671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,1,2,power_law_1.2,8.5012939453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,1,2,power_law_1.01,1.3654144287109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,1,2,power_law_1.01,1.0067903900146484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,1,2,power_law_1.01,1.9585075378417969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,1,2,power_law_1.2,0.13675647735595703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,1,2,power_law_1.2,0.15554304122924806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,1,2,power_law_1.2,0.1708403205871582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,1,2,power_law_1.2,0.262741756439209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,1,2,power_law_1.2,0.2640537643432617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,1,2,power_law_1.2,0.26565887451171877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,1,2,power_law_1.2,0.2680601692199707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,1,2,power_law_1.2,0.26867584228515623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,1,2,power_law_1.2,0.2707072067260742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,1,2,power_law_1.2,0.27155839920043945
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,1,2,power_law_1.01,2.594607238769531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,1,2,power_law_1.2,0.2753113555908203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,1,2,power_law_1.2,0.3047065544128418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,1,2,power_law_1.2,0.2917158317565918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,1,2,power_law_1.2,0.3086016082763672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,1,2,power_law_1.2,13.8785205078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,1,2,power_law_1.2,0.3330918502807617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,1,2,power_law_1.2,0.3768819046020508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,1,2,power_law_1.2,0.40311294555664057
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,1,2,power_law_1.2,0.4777638244628906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,1,2,power_law_1.2,0.8696678161621094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,1,2,power_law_1.2,0.5786547088623047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,1,2,power_law_1.2,1.0836441802978516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,1,2,power_law_1.2,1.6795327758789065
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,1,2,power_law_1.01,3.6682559204101564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,1,2,power_law_1.2,2.140112609863281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,1,2,power_law_1.2,18.285169677734377
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,1,2,power_law_1.2,0.05637887954711914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,1,2,power_law_1.2,0.08139776229858399
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,1,2,power_law_1.2,0.08632063865661621
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,1,2,power_law_1.2,0.13436415672302246
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,1,2,power_law_1.2,0.14120832443237305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,1,2,power_law_1.2,0.14187904357910158
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,1,2,power_law_1.2,0.14203776359558104
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,1,2,power_law_1.2,0.14229632377624513
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,1,2,power_law_1.2,0.1424563217163086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,1,2,power_law_1.2,2.992591247558594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,1,2,power_law_1.2,0.1434278392791748
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,1,2,power_law_1.2,0.15537792205810547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,1,2,power_law_1.2,0.16063488006591797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,1,2,power_law_1.2,4.309434814453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,1,2,power_law_1.2,0.20375680923461914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,1,2,power_law_1.2,0.21326976776123047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,1,2,power_law_1.01,5.329957275390624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,1,2,power_law_1.2,0.31112831115722656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,1,2,power_law_1.2,6.47759765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,1,2,power_law_1.2,29.848281250000003
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,1,2,power_law_1.2,8.697067260742188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,1,2,power_law_1.2,0.15185919761657715
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,1,2,power_law_1.2,0.1643827247619629
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,1,2,power_law_1.2,0.24163583755493162
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,1,2,power_law_1.2,48.003701171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,1,2,power_law_1.2,0.3663398361206055
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,1,2,power_law_1.2,0.5571059036254883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,2,1,balanced,0.16934463500976563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,2,1,balanced,0.28284351348876957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,2,1,balanced,0.4728838348388672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,2,1,balanced,0.4728115081787109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,2,1,balanced,0.5421932983398438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,2,1,balanced,0.5442412948608398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,2,1,balanced,0.5503308868408203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,2,1,balanced,0.5506022262573242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,2,1,balanced,0.5392486572265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,2,1,balanced,0.5430534362792969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,2,1,balanced,0.5189273452758789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,2,1,balanced,0.5200844955444336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,2,1,balanced,0.5250476837158203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,2,1,balanced,0.54052734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,2,1,balanced,0.7914867401123047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,2,1,balanced,0.8219821166992187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,2,1,balanced,0.8754201507568359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,2,1,balanced,1.2116928100585938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,2,1,balanced,1.493887939453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,1,2,power_law_1.2,0.6220902252197266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,2,1,balanced,2.241815643310547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,2,1,balanced,2.7216986083984374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,2,1,balanced,4.2600775146484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,2,1,balanced,5.525631713867187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,1,2,power_law_1.2,1.007105255126953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,2,1,balanced,8.407434692382813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,1,2,power_law_1.2,1.4491148376464844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,2,1,balanced,12.445313720703124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,2,1,balanced,0.14818943977355956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,2,1,balanced,0.2511180877685547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,1,2,power_law_1.2,3.942837829589844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,2,1,balanced,0.09128512382507324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,2,1,balanced,0.2755475234985352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,2,1,balanced,0.2661471939086914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,2,1,balanced,0.269745922088623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,2,1,balanced,0.27002944946289065
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,1,2,power_law_1.2,1.7543309020996094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,2,1,balanced,0.2695801544189453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,2,1,balanced,0.2748633575439453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,2,1,balanced,0.2513996887207031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,2,1,balanced,0.30832256317138673
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,2,1,balanced,0.27739391326904295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,2,1,balanced,0.28219392776489255
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,2,1,balanced,0.284703369140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,2,1,balanced,0.29434816360473637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,2,1,balanced,0.296231689453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,2,1,balanced,0.30436607360839846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,2,1,balanced,0.3921331024169922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,2,1,balanced,0.41614017486572263
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,2,1,balanced,0.5751142501831055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,2,1,balanced,0.7356057739257812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,2,1,balanced,1.092339859008789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,2,1,balanced,1.4487263488769533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,2,1,balanced,2.135479736328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,2,1,balanced,2.7946527099609377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,2,1,balanced,4.228072204589844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,2,1,balanced,20.77318603515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,2,1,balanced,5.478446655273437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,1,2,power_law_1.2,2.500767974853516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,2,1,balanced,30.431857910156246
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,1,2,power_law_1.2,3.3170611572265627
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,2,1,balanced,0.04731008052825928
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,2,1,balanced,0.13448896408081054
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,2,1,balanced,0.13425855636596679
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,2,1,balanced,0.13471551895141604
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,2,1,balanced,0.0811366367340088
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,2,1,balanced,0.1364345645904541
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,2,1,balanced,0.13697983741760253
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,2,1,balanced,0.1396281623840332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,2,1,balanced,0.14125247955322268
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,2,1,balanced,0.142225284576416
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,2,1,balanced,0.1460812759399414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,2,1,balanced,0.1332038402557373
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,2,1,balanced,0.13217023849487303
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,2,1,balanced,0.13354944229125976
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,2,1,balanced,0.13413887977600097
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,2,1,balanced,0.1337715244293213
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,2,1,balanced,0.13437696456909182
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,2,1,balanced,0.2290719985961914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,2,1,balanced,0.22129535675048828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,2,1,balanced,0.3130310440063476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,2,1,power_law_1.01,0.1601811218261719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,2,1,power_law_1.01,0.267770881652832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,2,1,power_law_1.01,0.3284377670288086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,2,1,balanced,0.4065254211425781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,2,1,balanced,0.5932889556884765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,2,1,power_law_1.01,0.5142303848266602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,2,1,power_law_1.01,0.544600944519043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,2,1,power_law_1.01,0.5588339233398438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,2,1,power_law_1.01,0.4608787155151367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,2,1,power_law_1.01,0.5503449630737305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,2,1,power_law_1.01,0.5625548934936523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,2,1,power_law_1.01,0.5681452941894531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,2,1,power_law_1.01,0.6148281478881836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,2,1,power_law_1.01,0.6388288116455078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,2,1,power_law_1.01,0.6820416259765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,2,1,power_law_1.01,0.741954574584961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,2,1,power_law_1.01,0.7821529388427735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,2,1,power_law_1.01,0.8584710693359374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,2,1,power_law_1.01,1.0384595489501953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,2,1,power_law_1.01,1.3610272216796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,2,1,balanced,0.7621868896484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,2,1,power_law_1.01,1.7075091552734374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,2,1,power_law_1.01,2.3726483154296876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,2,1,balanced,1.1993280029296876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,2,1,power_law_1.01,3.0074847412109373
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,2,1,balanced,1.6229785156249998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,2,1,power_law_1.01,4.355347290039062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,2,1,power_law_1.01,0.09959551811218262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,2,1,power_law_1.01,0.14099007606506347
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,2,1,power_law_1.01,0.16832895278930665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,2,1,power_law_1.01,0.24858175277709962
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,2,1,power_law_1.01,0.2627667236328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,2,1,power_law_1.01,0.26503999710083004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,2,1,power_law_1.01,0.26641920089721677
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,2,1,power_law_1.01,0.26636032104492186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,2,1,power_law_1.01,5.8808935546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,2,1,power_law_1.01,0.26663295745849613
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,2,1,power_law_1.01,0.27970624923706056
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,2,1,power_law_1.01,0.2745439910888672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,2,1,power_law_1.01,0.27796287536621095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,2,1,power_law_1.01,0.2785612869262695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,2,1,power_law_1.01,0.2915142440795898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,2,1,power_law_1.01,0.2962003135681152
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,2,1,power_law_1.01,0.3103264045715332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,2,1,power_law_1.01,0.3375801467895508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,2,1,power_law_1.01,0.41969345092773436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,2,1,power_law_1.01,0.47912063598632815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,2,1,power_law_1.01,0.6173907089233398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,2,1,power_law_1.01,0.7850342559814453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,2,1,power_law_1.01,8.728233032226562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,2,1,power_law_1.01,1.1608793640136719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,2,1,power_law_1.01,1.5378425598144532
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,2,1,balanced,2.5011302185058595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,2,1,power_law_1.01,11.574952392578124
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,2,1,power_law_1.01,0.04995840072631836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,2,1,power_law_1.01,0.07812672138214112
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,2,1,power_law_1.01,0.09073280334472657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,2,1,power_law_1.01,0.13117312431335448
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,2,1,power_law_1.01,0.13286208152770995
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,2,1,balanced,3.256287231445312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,2,1,power_law_1.01,0.1333043193817139
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,2,1,power_law_1.01,0.13716480255126953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,2,1,power_law_1.01,0.13224831581115723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,2,1,power_law_1.01,2.1948275756835938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,2,1,power_law_1.01,0.13343551635742187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,2,1,power_law_1.01,0.13523520469665526
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,2,1,power_law_1.01,0.14375871658325196
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,2,1,power_law_1.01,0.13634048461914064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,2,1,power_law_1.01,2.8396267700195312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,2,1,power_law_1.01,18.936715087890626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,2,1,power_law_1.01,4.203190307617187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,2,1,power_law_1.01,5.6149658203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,2,1,power_law_1.01,27.256328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,2,1,power_law_1.01,0.1406067180633545
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,2,1,power_law_1.01,0.15891776084899903
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,2,1,power_law_1.01,0.1551187229156494
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,2,1,power_law_1.01,0.1676684761047363
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,2,1,power_law_1.01,0.18975679397583006
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,2,1,power_law_1.01,0.23892799377441407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,2,1,power_law_1.01,0.2860044860839844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,2,1,power_law_1.01,0.3763916778564453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,2,1,power_law_1.2,0.1729862403869629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,2,1,power_law_1.2,0.2535475158691406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,2,1,power_law_1.2,0.2984761619567871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,2,1,power_law_1.2,0.46150657653808597
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,2,1,power_law_1.2,0.5004486465454102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,2,1,power_law_1.2,0.5472582244873048
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,2,1,power_law_1.01,0.480426254272461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,2,1,power_law_1.2,0.563749771118164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,2,1,power_law_1.2,0.5512742233276368
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,2,1,power_law_1.2,0.5711481475830078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,2,1,power_law_1.2,0.5692972946166992
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,2,1,power_law_1.01,0.6783968353271485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,2,1,power_law_1.2,0.6332396697998047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,2,1,power_law_1.2,0.653662109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,2,1,power_law_1.2,0.6772710418701171
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,2,1,power_law_1.2,0.7398912048339843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,2,1,power_law_1.2,0.7863539123535157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,2,1,power_law_1.2,0.8582546997070313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,2,1,power_law_1.2,1.0163494110107423
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,2,1,power_law_1.2,1.3785491943359376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,2,1,power_law_1.01,0.934039077758789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,2,1,power_law_1.2,1.66453369140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,2,1,power_law_1.2,2.3633804321289062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,2,1,power_law_1.01,1.3075538635253907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,2,1,power_law_1.2,3.0289599609374998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,2,1,power_law_1.2,4.422800598144532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,2,1,power_law_1.2,5.816576538085938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,2,1,power_law_1.01,1.8352575683593748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,2,1,power_law_1.2,0.09475520133972168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,2,1,power_law_1.2,0.13977663993835449
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,2,1,power_law_1.2,0.15043904304504394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,2,1,power_law_1.2,0.24899456024169925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,2,1,power_law_1.2,0.26101055145263674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,2,1,power_law_1.2,0.2660870361328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,2,1,power_law_1.2,0.2700428771972656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,2,1,power_law_1.2,0.26805696487426756
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,2,1,power_law_1.2,0.2728473663330078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,2,1,power_law_1.2,0.2703494453430176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,2,1,power_law_1.2,0.2746918487548828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,2,1,power_law_1.2,0.27719295501708985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,2,1,power_law_1.2,0.2754816055297852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,2,1,power_law_1.2,0.2904966354370117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,2,1,power_law_1.2,0.3019001579284668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,2,1,power_law_1.2,0.31673728942871093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,2,1,power_law_1.2,0.3460095977783203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,2,1,power_law_1.2,0.40243583679199213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,2,1,power_law_1.2,0.48450431823730467
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,2,1,power_law_1.01,2.5996084594726563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,2,1,power_law_1.2,0.6121401596069336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,2,1,power_law_1.2,8.7099365234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,2,1,power_law_1.2,0.8069932556152344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,2,1,power_law_1.2,1.1060486602783204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,2,1,power_law_1.2,1.5022451782226562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,2,1,power_law_1.2,11.841563720703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,2,1,power_law_1.2,0.051973118782043456
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,2,1,power_law_1.2,0.07395840167999268
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,2,1,power_law_1.2,0.08338751792907714
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,2,1,power_law_1.2,0.13104767799377443
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,2,1,power_law_1.2,0.1301088047027588
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,2,1,power_law_1.2,0.1330291175842285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,2,1,power_law_1.2,0.13334719657897948
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,2,1,power_law_1.2,0.13253184318542482
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,2,1,power_law_1.2,0.13441216468811035
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,2,1,power_law_1.2,0.13511551856994627
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,2,1,power_law_1.01,3.515672912597656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,2,1,power_law_1.2,0.13626879692077637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,2,1,power_law_1.2,2.1988319396972655
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,2,1,power_law_1.2,0.13925439834594727
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,2,1,power_law_1.2,0.1406112003326416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,2,1,power_law_1.2,2.846846008300781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,2,1,power_law_1.2,19.608469238281252
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,2,1,power_law_1.2,0.2441004753112793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,2,1,power_law_1.2,4.160869750976563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,2,1,power_law_1.2,0.1551097583770752
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,2,1,power_law_1.2,0.17436800003051758
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,2,1,power_law_1.2,0.16205184936523437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,2,1,power_law_1.2,0.4765990447998047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,2,1,power_law_1.2,0.1964249610900879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,2,1,power_law_1.2,5.590550537109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,2,1,power_law_1.2,27.725810546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,2,1,power_law_1.2,0.2867263984680176
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,2,1,power_law_1.2,0.3722681427001953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,2,balanced,0.09769472122192382
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,2,balanced,0.29797504425048826
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,2,balanced,0.521267204284668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,2,balanced,0.14914943695068358
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,2,balanced,0.9464473724365234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,2,balanced,1.8956787109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,2,balanced,1.957891845703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,2,balanced,1.8625202941894532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,2,balanced,1.956275177001953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,2,1,power_law_1.2,0.6926143646240235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,2,balanced,1.8801370239257813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,2,balanced,1.9483366394042967
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,2,1,power_law_1.2,0.8817996978759766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,2,balanced,2.0712179565429687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,2,balanced,2.089852142333984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,2,balanced,2.1044044494628906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,2,balanced,2.1304063415527343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,2,balanced,2.150762176513672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,2,balanced,2.0682150268554684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,2,balanced,2.197139129638672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,2,balanced,2.2491442871093748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,2,balanced,2.385733184814453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,2,balanced,2.535577545166016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,2,balanced,3.891422729492187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,2,balanced,4.177816467285156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,2,1,power_law_1.2,1.3741279602050782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,2,balanced,6.134967041015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,2,balanced,7.723793945312499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,2,balanced,0.1512819194793701
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,2,balanced,0.2736921691894531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,2,balanced,0.5030835342407227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,2,balanced,0.9367616271972656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,2,balanced,0.061223678588867195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,2,balanced,0.9356288146972656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,2,balanced,0.942388458251953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,2,balanced,0.08540800094604492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,2,balanced,0.9453132629394532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,2,balanced,0.9472946929931642
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,2,balanced,0.9418265533447265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,2,balanced,1.0163929748535157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,2,balanced,12.964654541015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,2,balanced,1.0235404968261719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,2,balanced,1.0327129364013672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,2,balanced,1.0384178924560548
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,2,1,power_law_1.2,1.823168029785156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,2,balanced,1.0643341064453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,2,balanced,1.0333618927001953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,2,balanced,1.102233581542969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,2,balanced,1.1435340881347655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,2,balanced,1.1822271728515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,2,balanced,1.3135411071777345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,2,balanced,1.4528294372558594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,2,balanced,1.6190821838378906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,2,balanced,2.333263397216797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,2,balanced,2.6071896362304687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,2,balanced,18.778402099609373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,2,balanced,3.768641357421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,2,balanced,4.660145874023438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,2,1,power_law_1.2,2.680987548828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,1,2,balanced,0.04973696231842041
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,1,2,balanced,0.565483512878418
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,1,2,balanced,0.06651648044586181
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,1,2,balanced,0.15963904380798338
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,1,2,balanced,0.09773695945739747
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,1,2,balanced,0.5672844696044922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,1,2,balanced,0.5676582336425782
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,1,2,balanced,0.5713587188720703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,2,1,power_law_1.2,3.5994931030273434
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,1,2,balanced,0.5697894287109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,1,2,balanced,0.5718054580688476
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,1,2,balanced,0.5757043075561523
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,1,2,balanced,0.5680012893676758
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,1,2,balanced,0.29251071929931644
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,1,2,balanced,0.5910374450683593
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,1,2,balanced,0.5995059204101563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,1,2,balanced,0.6056524658203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,1,2,balanced,0.626855697631836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,1,2,balanced,0.6508889770507812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,1,2,balanced,0.6907673645019531
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,1,2,balanced,0.7333030700683594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,1,2,balanced,0.5625574493408203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,2,power_law_1.01,0.2950527954101562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,2,power_law_1.01,0.40646526336669925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,2,power_law_1.01,0.5780710220336914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,2,power_law_1.01,0.7841331481933593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,2,power_law_1.01,0.14598912239074707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,2,power_law_1.01,1.0446310424804688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,2,power_law_1.01,1.5040460205078126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,2,power_law_1.01,1.4366157531738282
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,1,2,balanced,0.5631513595581055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,2,power_law_1.01,1.5203916931152344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,2,power_law_1.01,1.4576039123535156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,2,power_law_1.01,1.6148658752441407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,2,power_law_1.01,1.7103308105468749
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,2,power_law_1.01,1.7283392333984373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,2,power_law_1.01,1.7966578674316405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,2,power_law_1.01,1.8545689392089844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,2,power_law_1.01,1.9847514343261718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,2,power_law_1.01,2.0322470092773437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,2,power_law_1.01,2.8704461669921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,2,power_law_1.01,2.150846710205078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,2,power_law_1.01,2.3759231567382812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,2,power_law_1.01,3.416234130859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,2,power_law_1.01,4.318307800292969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,2,power_law_1.01,5.256298217773438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,1,2,balanced,0.5806310272216797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,2,power_law_1.01,7.288703002929688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,1,2,balanced,1.1392115020751954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,2,power_law_1.01,0.08799488067626952
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,1,2,balanced,1.2451251220703123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,2,power_law_1.01,0.16133247375488283
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,2,power_law_1.01,0.19942399978637695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,1,2,balanced,1.8633689880371094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,2,power_law_1.01,0.43925121307373044
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,1,2,balanced,2.4861631774902344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,2,power_law_1.01,0.5410291290283202
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,2,power_law_1.01,0.7323776245117187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,2,power_law_1.01,0.7479180908203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,2,power_law_1.01,0.7764185333251954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,2,power_law_1.01,8.989285278320313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,2,power_law_1.01,0.866270751953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,2,power_law_1.01,0.8945101165771485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,2,power_law_1.01,0.8823820495605469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,2,power_law_1.01,0.9123442840576171
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,2,power_law_1.01,0.9716659545898437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,2,power_law_1.01,1.0068211364746094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,2,power_law_1.01,1.1216307067871094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,2,power_law_1.01,1.2143500518798827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,2,power_law_1.01,12.36326416015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,2,power_law_1.01,1.3832550048828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,2,power_law_1.01,0.766033935546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,2,power_law_1.01,0.7803033447265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,2,power_law_1.01,1.6312985229492187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,2,power_law_1.01,1.9069299316406252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,2,power_law_1.01,0.3212108612060547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,1,2,power_law_1.01,0.07481215953826904
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,1,2,power_law_1.01,0.10695936203002929
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,1,2,power_law_1.01,0.14068991661071778
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,1,2,power_law_1.01,0.1933568000793457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,2,power_law_1.01,2.512665557861328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,1,2,power_law_1.01,0.2689638328552246
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,1,2,power_law_1.01,0.3283123016357422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,2,power_law_1.01,2.981475830078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,1,2,power_law_1.01,0.43116928100585933
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,1,2,power_law_1.01,0.4418815994262696
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,1,2,power_law_1.01,0.449224967956543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,1,2,power_law_1.01,0.47794303894042967
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,1,2,power_law_1.01,0.4893273544311524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,2,power_law_1.01,0.8917005157470703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,2,power_law_1.01,4.229736938476562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,1,2,power_law_1.01,0.4929830551147461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,1,2,power_law_1.01,0.5146265411376953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,1,2,power_law_1.01,0.534837760925293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,1,2,power_law_1.01,0.5341158294677735
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,1,2,power_law_1.01,0.564218864440918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,1,2,power_law_1.01,0.5233062362670898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,2,power_law_1.01,5.39994873046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,2,power_law_1.01,16.082451171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,1,2,power_law_1.01,0.6266508865356445
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,1,2,power_law_1.01,0.7185702514648438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,1,2,power_law_1.01,0.6500518035888672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,1,2,power_law_1.01,0.8568102264404297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,1,2,power_law_1.01,1.038856964111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,2,power_law_1.2,0.14875519752502442
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,1,2,power_law_1.01,1.3203916931152344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,2,power_law_1.2,0.2965017509460449
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,1,2,power_law_1.01,1.7559347534179686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,2,power_law_1.2,0.911352310180664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,2,power_law_1.2,0.37488384246826173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,2,power_law_1.2,0.5732390213012695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,2,power_law_1.2,1.3648384094238282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,2,power_law_1.2,1.442571563720703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,2,power_law_1.2,1.4755494689941406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,1,2,power_law_1.01,2.427386932373047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,2,power_law_1.2,1.5694886779785155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,2,power_law_1.2,1.763811798095703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,2,power_law_1.2,1.8963494873046876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,2,power_law_1.2,1.8954393005371095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,2,power_law_1.2,2.0270681762695313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,2,power_law_1.2,2.0554112243652343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,2,power_law_1.2,1.5421055603027345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,2,power_law_1.2,0.7224883270263672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,2,power_law_1.2,2.3902989196777344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,2,power_law_1.2,1.678128662109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,2,power_law_1.2,2.5981118774414065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,2,power_law_1.2,3.0522265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,2,power_law_1.2,3.5374526977539062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,2,power_law_1.2,4.483395690917969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,2,power_law_1.2,5.5565161132812495
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,2,power_law_1.2,0.14893183708190919
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,2,power_law_1.2,0.188220157623291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,2,power_law_1.2,0.2947327995300293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,2,power_law_1.2,0.08471551895141602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,2,power_law_1.2,7.87011474609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,2,power_law_1.2,0.4038835144042968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,2,power_law_1.2,0.73557373046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,2,power_law_1.2,0.7448499298095703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,2,power_law_1.2,0.5051340866088867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,2,power_law_1.2,9.43398681640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,2,power_law_1.2,0.7685555267333984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,2,power_law_1.2,0.8056768035888672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,2,power_law_1.2,0.8818726348876954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,2,power_law_1.2,0.9160192108154297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,2,power_law_1.2,0.9363225555419922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,2,power_law_1.2,12.841484374999999
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,2,power_law_1.2,0.983477783203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,2,power_law_1.2,1.048616943359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,2,power_law_1.2,0.7586598205566406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,2,power_law_1.2,1.161856002807617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,2,power_law_1.2,1.3225715637207032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,2,power_law_1.2,17.890607910156252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,2,power_law_1.2,0.826776351928711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,2,power_law_1.2,1.4220941162109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,2,power_law_1.2,1.7412109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,2,power_law_1.2,2.0963891601562503
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,1,2,power_law_1.2,0.08315520286560059
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,2,power_law_1.2,0.928985595703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,1,2,power_law_1.2,0.10664192199707032
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,1,2,power_law_1.01,3.033630676269531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,2,power_law_1.2,2.735362548828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,1,2,power_law_1.2,0.12430080413818359
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,1,2,power_law_1.2,0.1849126434326172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,2,power_law_1.2,3.146531982421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,1,2,power_law_1.01,3.853360595703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,1,2,power_law_1.2,0.22704639434814453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,1,2,power_law_1.2,0.2889894485473633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,1,2,power_law_1.2,0.4574310302734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,1,2,power_law_1.2,0.44199680328369145
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,1,2,power_law_1.2,0.4621900939941407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,1,2,power_law_1.2,0.4943654251098633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,1,2,power_law_1.2,0.48772224426269534
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,1,2,power_law_1.2,0.4729305648803711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,1,2,power_law_1.2,0.5117427062988281
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,1,2,power_law_1.2,0.5353830337524415
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,1,2,power_law_1.2,0.5599539184570312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,2,power_law_1.2,4.4372583007812505
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,1,2,power_law_1.2,0.5600204849243164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,1,2,power_law_1.2,0.6089011383056641
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,1,2,power_law_1.2,0.6692723083496094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,1,2,power_law_1.2,1.0006143951416016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,2,power_law_1.2,5.630178833007813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,1,2,power_law_1.2,0.8335014343261719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,1,2,power_law_1.2,1.3010252380371095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,1,balanced,0.1065171241760254
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,1,2,power_law_1.2,1.7162661743164063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,1,balanced,0.14882111549377441
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,1,2,power_law_1.2,0.7349900817871093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,1,2,power_law_1.2,2.350435791015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,1,balanced,0.5269734573364258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,1,2,power_law_1.2,2.7714968872070314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,1,balanced,0.9651289367675782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,1,balanced,1.9162495422363282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,1,balanced,0.30309568405151366
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,1,balanced,1.9128262329101564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,1,balanced,1.9321209716796877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,1,balanced,1.899824676513672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,1,2,power_law_1.2,4.9982629394531255
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,1,balanced,1.9268121337890627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,1,balanced,1.94619140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,1,balanced,2.0016333007812497
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,1,2,power_law_1.2,6.2248974609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,1,balanced,2.010981140136719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,1,balanced,2.021143646240234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,1,balanced,2.1205375671386717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,1,balanced,2.1183526611328127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,1,balanced,2.1920979309082034
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,1,balanced,2.2299430847167967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,1,balanced,2.269935302734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,1,balanced,2.5741806030273438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,1,balanced,2.380768585205078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,1,balanced,0.08052607536315917
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,1,balanced,0.10341376304626464
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,1,balanced,0.16198591232299803
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,1,balanced,0.28442304611206054
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,1,balanced,3.8760006713867186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,1,balanced,4.01740234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,1,balanced,0.5141491317749024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,1,balanced,1.002520294189453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,1,balanced,1.0191891479492188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,1,balanced,1.0082054138183594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,1,balanced,1.0100204467773437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,1,balanced,1.0112614440917969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,1,balanced,1.0081394958496095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,1,balanced,5.819855346679687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,1,balanced,1.021489944458008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,1,balanced,1.0353343963623047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,1,balanced,1.049863052368164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,1,balanced,1.157847671508789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,1,balanced,1.1700646209716798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,1,balanced,7.309335327148437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,1,balanced,1.1881510162353517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,1,balanced,1.2002944183349609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,1,balanced,1.2453241729736328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,1,balanced,1.3208274841308594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,1,balanced,1.3389733886718749
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,1,balanced,1.4973574829101561
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,1,balanced,1.6081510925292968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,1,balanced,2.4717254638671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,1,balanced,2.703031005859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,1,balanced,10.872315673828124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,1,balanced,3.806372985839844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,1,balanced,4.786197204589844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,1,balanced,16.087214355468753
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,2,1,balanced,0.06832704067230225
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,2,1,balanced,0.08334272384643555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,2,1,balanced,0.11351488113403321
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,2,1,balanced,0.6155251312255859
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,2,1,balanced,0.6138777542114258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,2,1,balanced,0.17575807571411134
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,2,1,balanced,0.6159283065795899
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,2,1,balanced,0.319234561920166
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,2,1,balanced,0.6189561462402343
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,2,1,balanced,0.6218924713134766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,2,1,balanced,0.6188275146484374
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,2,1,balanced,0.6267776107788086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,2,1,balanced,0.6100934219360352
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,2,1,balanced,0.6305382537841797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,2,1,balanced,0.6414758300781249
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,2,1,balanced,0.6110400009155273
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,2,1,balanced,0.6567501068115235
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,2,1,balanced,0.6747885131835938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,2,1,balanced,0.7012818908691406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,2,1,balanced,0.7435481262207031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,2,1,balanced,0.8143321228027343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,1,power_law_1.01,0.109170560836792
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,2,1,balanced,0.8784499359130858
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,1,power_law_1.01,0.15211520195007325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,1,power_law_1.01,0.3063430404663086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,1,power_law_1.01,0.5391007995605468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,1,power_law_1.01,0.759820785522461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,2,1,balanced,0.6107046508789062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,1,power_law_1.01,0.9770706939697267
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,1,power_law_1.01,1.3760748291015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,1,power_law_1.01,1.4546290588378905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,1,power_law_1.01,1.5002137756347655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,1,power_law_1.01,1.5174771118164063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,1,power_law_1.01,1.6064857482910155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,1,power_law_1.01,1.741045684814453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,1,power_law_1.01,1.814481964111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,1,power_law_1.01,2.0015040588378907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,1,power_law_1.01,1.8885267639160155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,1,power_law_1.01,1.9101210021972654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,1,power_law_1.01,2.0277484130859373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,1,power_law_1.01,2.2145330810546873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,1,power_law_1.01,2.4227577209472657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,1,power_law_1.01,2.8100070190429687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,1,power_law_1.01,3.243747253417969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,1,power_law_1.01,3.9722476196289067
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,1,power_law_1.01,4.762398681640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,1,power_law_1.01,6.439335327148437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,1,power_law_1.01,0.0699129581451416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,1,power_law_1.01,0.09763263702392579
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,1,power_law_1.01,0.159039363861084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,1,power_law_1.01,8.103995971679687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,1,power_law_1.01,0.2823001670837402
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,1,power_law_1.01,0.42273792266845706
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,1,power_law_1.01,0.5496089553833008
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,2,1,balanced,1.3662118530273437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,1,power_law_1.01,0.7408985900878906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,2,1,balanced,1.50338623046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,1,power_law_1.01,0.7758956909179687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,1,power_law_1.01,0.7927865600585937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,2,1,balanced,2.217552032470703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,1,power_law_1.01,0.8026624298095703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,1,power_law_1.01,0.8474732971191405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,1,power_law_1.01,0.873064956665039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,2,1,balanced,2.860999755859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,1,power_law_1.01,0.8807347106933594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,1,power_law_1.01,0.8991545867919921
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,1,power_law_1.01,1.0209152221679687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,1,power_law_1.01,11.350626220703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,1,power_law_1.01,1.0224832153320311
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,1,power_law_1.01,1.0659468841552733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,1,power_law_1.01,1.103861083984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,1,power_law_1.01,1.176833267211914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,1,power_law_1.01,1.3078572082519533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,1,power_law_1.01,1.4402432250976562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,1,power_law_1.01,1.7211174011230468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,1,power_law_1.01,1.9607379150390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,2,1,power_law_1.01,0.07420671939849853
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,1,power_law_1.01,2.50725830078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,2,1,power_law_1.01,0.09238656044006348
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,2,1,power_law_1.01,0.12276800155639649
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,1,power_law_1.01,3.02465087890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,1,power_law_1.01,14.471418457031252
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,2,1,power_law_1.01,0.17991680145263672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,2,1,power_law_1.01,0.2550777626037598
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,2,1,power_law_1.01,0.3198598480224609
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,2,1,power_law_1.01,0.44746238708496094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,2,1,power_law_1.01,0.4717318344116211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,1,power_law_1.01,4.150184936523438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,2,1,power_law_1.01,0.48425472259521485
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,2,1,power_law_1.01,0.4904934310913086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,2,1,power_law_1.01,0.5152518463134765
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,2,1,power_law_1.01,0.5331974411010743
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,1,power_law_1.01,5.220394287109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,2,1,power_law_1.01,0.5411308670043946
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,2,1,power_law_1.01,0.5533695983886718
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,2,1,power_law_1.01,0.5785823822021484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,2,1,power_law_1.01,0.588458251953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,2,1,power_law_1.01,0.6286643218994141
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,2,1,power_law_1.01,0.6861984252929687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,2,1,power_law_1.01,0.7504287719726562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,2,1,power_law_1.01,0.8800838470458985
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,2,1,power_law_1.01,0.9915878295898437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,2,1,power_law_1.01,1.2398265838623046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,1,power_law_1.2,0.10983231544494629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,1,power_law_1.2,0.1522265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,2,1,power_law_1.01,1.523047637939453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,1,power_law_1.2,0.29658559799194334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,1,power_law_1.2,0.479106559753418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,1,power_law_1.2,0.6939974212646485
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,2,1,power_law_1.01,2.0848326110839843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,1,power_law_1.2,0.915313262939453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,2,1,power_law_1.01,2.6834130859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,1,power_law_1.2,1.3417861938476563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,1,power_law_1.2,1.43679931640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,1,power_law_1.2,1.4696365356445313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,1,power_law_1.2,1.5080281066894532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,1,power_law_1.2,1.629808654785156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,1,power_law_1.2,1.818553009033203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,1,power_law_1.2,1.7104953002929686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,1,power_law_1.2,2.053910369873047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,1,power_law_1.2,1.916465301513672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,1,power_law_1.2,1.9629945373535158
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,1,power_law_1.2,2.0459143066406247
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,1,power_law_1.2,2.288465881347656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,1,power_law_1.2,2.530397491455078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,1,power_law_1.2,2.9599124145507814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,1,power_law_1.2,3.2974630737304684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,1,power_law_1.2,4.096827392578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,1,power_law_1.2,0.07394879817962646
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,1,power_law_1.2,4.913281860351562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,1,power_law_1.2,0.09815296173095703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,1,power_law_1.2,0.16292224884033205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,1,power_law_1.2,0.2696703910827637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,1,power_law_1.2,0.38398719787597657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,1,power_law_1.2,0.5297126388549804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,1,power_law_1.2,6.472810668945312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,1,power_law_1.2,0.7264358520507812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,1,power_law_1.2,0.767820816040039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,1,power_law_1.2,0.7866169738769531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,1,power_law_1.2,0.812113265991211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,1,power_law_1.2,8.2370166015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,1,power_law_1.2,0.8452748870849609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,1,power_law_1.2,0.8640652465820313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,1,power_law_1.2,0.8943135833740234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,1,power_law_1.2,0.9124172973632814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,1,power_law_1.2,1.0325990295410157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,1,power_law_1.2,1.0673862457275392
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,1,power_law_1.2,1.065185317993164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,1,power_law_1.2,1.1615853118896484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,1,power_law_1.2,1.2331763458251952
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,1,power_law_1.2,11.406595458984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,1,power_law_1.2,1.3675180053710938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,1,power_law_1.2,1.5023890686035155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,1,power_law_1.2,1.7774758911132813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,1,power_law_1.2,2.0545330810546876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,1,power_law_1.2,2.603851013183594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,2,1,power_law_1.2,0.07343232154846191
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,1,power_law_1.2,3.1527783203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,1,power_law_1.2,14.72935791015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,2,1,power_law_1.2,0.09243904113769531
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,2,1,power_law_1.2,0.11974207878112793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,1,power_law_1.2,4.202131958007813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,2,1,power_law_1.2,0.1709369659423828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,2,1,power_law_1.01,3.916563720703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,2,1,power_law_1.2,0.23530303955078127
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,2,1,power_law_1.2,0.3028838348388672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,2,1,power_law_1.2,0.437498893737793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,1,power_law_1.2,5.241624755859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,2,1,power_law_1.2,0.4760883331298828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,2,1,power_law_1.2,0.4633382415771484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,2,1,power_law_1.2,0.5103667068481446
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,2,1,power_law_1.2,0.49184127807617184
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,2,1,power_law_1.2,0.5294182586669922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,2,1,power_law_1.01,5.155225830078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,2,1,power_law_1.2,0.5451091384887695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,2,1,power_law_1.2,0.57438720703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,2,1,power_law_1.2,0.5977702331542969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,2,1,power_law_1.2,0.6203200149536132
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,2,1,power_law_1.2,0.6622975921630859
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,2,1,power_law_1.2,0.8547987365722657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,2,1,power_law_1.2,0.75056640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,2,1,power_law_1.2,1.0178675079345703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,2,1,power_law_1.2,1.196654052734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,2,1,power_law_1.2,1.5919244384765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,2,balanced,0.08775424003601075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,2,balanced,0.12269311904907228
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,2,1,power_law_1.2,2.0635353088378907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,2,balanced,0.26058368682861327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,2,balanced,0.809081573486328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,2,balanced,0.46162559509277346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,2,balanced,1.5983488464355469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,2,1,power_law_1.2,3.0157568359374998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,2,balanced,1.6086642456054687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,2,balanced,1.5755775451660157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,2,balanced,1.6088102722167967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,2,balanced,1.6292965698242188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,2,1,power_law_1.2,4.009610290527344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,2,balanced,1.614960632324219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,2,balanced,1.7800755310058594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,2,balanced,1.774159393310547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,2,balanced,1.7931674194335936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,2,balanced,1.8187327575683594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,2,balanced,1.8323532104492188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,2,balanced,1.8080409240722655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,2,balanced,1.908601531982422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,2,balanced,1.9435635375976563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,2,balanced,2.1038053894042967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,2,balanced,2.225059814453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,2,balanced,3.380099792480469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,2,balanced,3.5839385986328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,2,balanced,0.05916287899017334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,2,balanced,0.13027456283569336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,2,balanced,0.23655168533325197
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,2,balanced,0.0806873607635498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,2,balanced,5.15481201171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,2,balanced,0.43207935333251957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,2,balanced,0.8034214019775391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,2,balanced,0.8136972808837891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,2,balanced,6.698064575195312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,2,balanced,0.8073702239990235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,2,balanced,0.8220889282226562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,2,balanced,0.8115737915039063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,2,balanced,0.8137920379638672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,2,balanced,0.8852416229248046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,2,balanced,0.899420166015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,2,balanced,0.8795174407958986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,2,balanced,0.8894156646728515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,2,balanced,0.8939520263671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,2,balanced,11.0134619140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,2,balanced,0.9083238220214843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,2,balanced,0.9280397033691405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,2,balanced,0.9932428741455078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,2,balanced,1.0408396911621094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,2,balanced,1.150730209350586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,2,balanced,1.2696294403076172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,2,balanced,1.381529541015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,2,balanced,2.091980743408203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,2,balanced,2.3403237915039066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,2,balanced,3.339146118164062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,2,balanced,14.348177490234374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,2,balanced,4.064698791503906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,1,2,balanced,0.04617343902587891
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,1,2,balanced,0.48469886779785154
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,1,2,balanced,0.4859852981567383
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,1,2,balanced,0.06178304195404053
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,1,2,balanced,0.48649471282958984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,1,2,balanced,0.48666240692138674
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,1,2,balanced,0.08868351936340332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,1,2,balanced,0.4890393447875977
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,1,2,balanced,0.48989822387695314
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,1,2,balanced,0.48839168548583983
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,1,2,balanced,0.1394700813293457
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,1,2,balanced,0.2531545639038086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,1,2,balanced,0.4828787231445313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,1,2,balanced,0.49469951629638675
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,1,2,balanced,0.49740032196044925
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,1,2,balanced,0.4979878234863281
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,1,2,balanced,0.5013951873779297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,1,2,balanced,0.5193664169311523
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,1,2,balanced,0.547720947265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,1,2,balanced,0.5616064071655273
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,1,2,balanced,0.595994873046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,1,2,balanced,0.627330551147461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,2,power_law_1.01,0.12067584037780761
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,2,power_law_1.01,0.2567654418945312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,2,power_law_1.01,0.3494707107543945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,2,power_law_1.01,0.5088243103027343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,2,power_law_1.01,0.7647974395751953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,2,power_law_1.01,0.899310073852539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,2,power_law_1.01,1.260522232055664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,2,power_law_1.01,1.274798049926758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,2,power_law_1.01,1.323125762939453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,2,power_law_1.01,1.3243930053710937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,2,power_law_1.01,1.3549722290039061
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,2,power_law_1.01,1.5164991760253907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,2,power_law_1.01,1.535509796142578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,2,power_law_1.01,1.6380543518066406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,2,power_law_1.01,1.5924517822265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,2,power_law_1.01,1.5958451843261718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,2,power_law_1.01,1.7056243896484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,1,2,balanced,0.9190092468261719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,1,2,balanced,1.0291212463378907
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,1,2,balanced,1.5469453430175781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,2,power_law_1.01,1.8710643005371093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,1,2,balanced,2.0616998291015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,2,power_law_1.01,2.0669081115722654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,2,power_law_1.01,2.415221710205078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,2,power_law_1.01,2.8659481811523437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,2,power_law_1.01,3.636280212402344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,2,power_law_1.01,4.527710571289062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,2,power_law_1.01,0.07964799880981446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,2,power_law_1.01,0.12880127906799316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,2,power_law_1.01,5.9818701171874995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,2,power_law_1.01,0.19770240783691406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,2,power_law_1.01,0.2772748756408691
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,2,power_law_1.01,0.4220684814453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,2,power_law_1.01,0.49383808135986323
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,2,power_law_1.01,0.6443148803710937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,2,power_law_1.01,7.816080932617187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,2,power_law_1.01,0.6773069000244141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,2,power_law_1.01,0.6751449584960938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,2,power_law_1.01,0.678150405883789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,2,power_law_1.01,0.704052505493164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,2,power_law_1.01,0.7636505889892578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,2,power_law_1.01,0.7852531433105469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,2,power_law_1.01,0.786108169555664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,2,power_law_1.01,0.7891404724121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,2,power_law_1.01,0.8010189056396484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,2,power_law_1.01,11.085888671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,2,power_law_1.01,0.837474594116211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,2,power_law_1.01,0.8664128112792969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,2,power_law_1.01,0.9374092864990236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,2,power_law_1.01,1.0328921508789062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,2,power_law_1.01,1.1545804595947264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,2,power_law_1.01,1.3975538635253906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,2,power_law_1.01,13.671112060546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,2,power_law_1.01,1.6682867431640624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,1,2,power_law_1.01,0.06869760036468506
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,1,2,power_law_1.01,0.10231040000915528
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,2,power_law_1.01,2.1367333984375003
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,1,2,power_law_1.01,0.12446208000183105
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,1,2,power_law_1.01,0.172620792388916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,1,2,power_law_1.01,0.25470848083496095
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,1,2,power_law_1.01,0.28230783462524417
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,1,2,power_law_1.01,0.38822784423828127
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,1,2,power_law_1.01,0.3975116729736328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,1,2,power_law_1.01,0.40179073333740234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,1,2,power_law_1.01,0.4057702255249024
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,1,2,power_law_1.01,0.4196377563476562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,1,2,power_law_1.01,0.4224729537963867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,1,2,power_law_1.01,0.4312102508544922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,1,2,power_law_1.01,0.4453747177124024
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,1,2,power_law_1.01,0.462053108215332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,1,2,power_law_1.01,0.47043838500976565
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,1,2,power_law_1.01,0.48435073852539057
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,2,power_law_1.01,4.708269958496094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,1,2,power_law_1.01,0.5574988937377929
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,1,2,power_law_1.01,0.6502054595947266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,2,power_law_1.01,2.7346572875976562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,2,power_law_1.01,3.5777511596679688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,1,2,power_law_1.01,0.5252300643920899
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,1,2,power_law_1.01,0.7412134552001953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,1,2,power_law_1.01,0.9451187133789063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,2,power_law_1.2,0.2519820785522461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,2,power_law_1.2,0.3265369415283203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,2,power_law_1.2,0.46096000671386717
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,1,2,power_law_1.01,1.1068287658691407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,2,power_law_1.2,0.6763788604736328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,2,power_law_1.2,0.12059007644653322
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,2,power_law_1.2,0.8327040100097657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,2,power_law_1.2,1.2112409973144531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,2,power_law_1.2,1.2517747497558593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,2,power_law_1.2,1.2917068481445313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,2,power_law_1.2,1.2708096313476562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,2,power_law_1.2,1.3459007263183593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,2,power_law_1.2,1.5071090698242187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,2,power_law_1.2,1.5672845458984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,2,power_law_1.2,1.6251213073730468
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,1,2,power_law_1.01,1.8446438598632813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,1,2,power_law_1.01,1.4498240661621093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,2,power_law_1.2,1.8283033752441404
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,2,power_law_1.2,1.4769381713867187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,2,power_law_1.2,1.9471359252929688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,2,power_law_1.2,2.2419750976562502
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,2,power_law_1.2,1.6658277893066404
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,2,power_law_1.2,2.635015563964844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,2,power_law_1.2,2.98790771484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,2,power_law_1.2,3.8828594970703123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,2,power_law_1.2,6.337908325195313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,2,power_law_1.2,0.08011775970458984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,2,power_law_1.2,0.13001472473144532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,2,power_law_1.2,4.670404968261719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,2,power_law_1.2,0.18123264312744142
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,2,power_law_1.2,8.319833374023437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,2,power_law_1.2,0.38697727203369137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,2,power_law_1.2,0.4367820739746094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,2,power_law_1.2,0.6415052795410157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,2,power_law_1.2,0.6648320007324219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,2,power_law_1.2,0.6674752044677734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,2,power_law_1.2,0.7064383697509766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,2,power_law_1.2,12.1927734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,2,power_law_1.2,0.7643135833740234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,2,power_law_1.2,0.7637824249267579
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,2,power_law_1.2,0.7963724517822266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,2,power_law_1.2,0.8177139282226562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,2,power_law_1.2,0.8609037017822265
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,1,2,power_law_1.01,3.6850405883789064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,2,power_law_1.2,14.566346435546876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,2,power_law_1.2,0.9237709045410156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,2,power_law_1.2,1.0025011444091798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,2,power_law_1.2,1.1119667053222657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,2,power_law_1.2,1.2405964660644533
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,1,2,power_law_1.01,2.6226651000976564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,2,power_law_1.2,1.4916339111328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,2,power_law_1.2,0.24355072021484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,2,power_law_1.2,1.7861439514160158
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,1,2,power_law_1.2,0.10032511711120606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,2,power_law_1.2,2.181788177490234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,1,2,power_law_1.2,0.1192563247680664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,1,2,power_law_1.2,0.17840768814086913
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,1,2,power_law_1.2,0.22547584533691406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,2,power_law_1.2,2.7587265014648437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,1,2,power_law_1.2,0.26355968475341796
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,1,2,power_law_1.2,0.3794623947143555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,1,2,power_law_1.2,0.3981990432739258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,1,2,power_law_1.2,0.4004019165039063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,2,power_law_1.2,0.6659763336181641
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,1,2,power_law_1.2,0.39446014404296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,2,power_law_1.2,0.7802816009521484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,1,2,power_law_1.2,0.42066814422607424
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,1,2,power_law_1.2,0.431646728515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,1,2,power_law_1.2,0.43855743408203124
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,1,2,power_law_1.2,0.4663641738891601
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,1,2,power_law_1.2,0.4718668746948242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,2,power_law_1.2,4.677890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,1,2,power_law_1.2,0.5091891098022461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,1,2,power_law_1.2,0.5573235321044921
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,1,2,power_law_1.2,0.6229465484619141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,2,power_law_1.2,3.530325622558594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,1,2,power_law_1.2,0.06707968235015868
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,1,2,power_law_1.2,0.7176614379882812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,1,2,power_law_1.2,0.868880615234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,2,1,balanced,0.09217727661132813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,1,2,power_law_1.2,1.2060288238525392
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,1,2,power_law_1.2,1.4889471435546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,2,1,balanced,0.1356172752380371
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,1,2,power_law_1.2,0.4395865631103516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,2,1,balanced,0.2618911933898926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,2,1,balanced,0.4614636611938477
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,1,2,power_law_1.2,2.161318359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,2,1,balanced,0.8165408325195311
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,2,1,balanced,1.636156768798828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,1,2,power_law_1.2,2.4690328979492184
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,2,1,balanced,1.6398419189453126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,2,1,balanced,1.6521888732910157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,2,1,balanced,1.6307174682617187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,2,1,balanced,1.6959269714355467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,2,1,balanced,1.7107725524902342
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,2,1,balanced,1.6454316711425783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,2,1,balanced,1.7574719238281251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,2,1,balanced,1.8345184326171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,2,1,balanced,1.8338182067871096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,2,1,balanced,1.8465216064453123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,2,1,balanced,1.9257919311523437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,2,1,balanced,1.9360453796386718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,2,1,balanced,2.0840960693359376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,2,1,balanced,2.2052653503417967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,2,1,balanced,3.2908178710937497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,2,1,balanced,3.532559509277344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,2,1,balanced,0.06757567882537842
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,2,1,balanced,1.6168249511718749
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,2,1,balanced,0.08865216255187988
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,2,1,balanced,0.24953535079956052
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,2,1,balanced,5.0674066162109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,2,1,balanced,0.44450622558593744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,2,1,balanced,0.8620249938964843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,2,1,balanced,0.8586329650878906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,2,1,balanced,0.8688563537597656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,2,1,balanced,6.35500244140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,2,1,balanced,0.8712525177001954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,2,1,balanced,0.14423871994018556
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,2,1,balanced,0.8761996459960937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,2,1,balanced,0.9053440093994141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,2,1,balanced,0.8861363220214844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,2,1,balanced,0.8743462371826173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,2,1,balanced,0.9071949005126954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,2,1,balanced,1.001683807373047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,2,1,balanced,1.0046272277832031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,2,1,balanced,1.0016921234130858
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,2,1,balanced,9.194361572265624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,2,1,balanced,1.054677734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,2,1,balanced,1.092824935913086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,2,1,balanced,1.1062387084960936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,2,1,balanced,1.151910400390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,2,1,balanced,1.265433578491211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,2,1,balanced,1.3819833374023438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,1,2,power_law_1.2,3.7016293334960935
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,2,1,balanced,2.1333056640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,2,1,balanced,12.111356201171876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,2,1,balanced,2.2795750427246095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,2,1,balanced,3.298922119140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,2,1,balanced,4.097720336914063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,1,2,power_law_1.2,5.110501098632812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,2,1,balanced,0.07906239986419678
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,2,1,balanced,0.06623167991638183
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,2,1,balanced,0.5232108688354492
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,2,1,balanced,0.1048639965057373
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,2,1,balanced,0.15560128211975097
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,2,1,balanced,0.523809928894043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,2,1,balanced,0.5255801773071289
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,2,1,balanced,0.5275276947021484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,2,1,balanced,0.5314112091064453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,2,1,balanced,0.5328812789916992
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,2,1,balanced,0.27407936096191404
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,2,1,balanced,0.536519660949707
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,2,1,balanced,0.5217817687988282
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,2,1,balanced,0.5481740951538085
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,2,1,balanced,0.5636505508422852
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,2,1,balanced,0.5202265548706054
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,2,1,balanced,0.5763404846191407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,2,1,balanced,0.5407743835449219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,2,1,balanced,0.5193644714355469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,2,1,balanced,0.5998009490966797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,2,1,balanced,0.6328364944458007
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,2,1,balanced,0.689319076538086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,2,1,balanced,0.7561164855957031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,2,1,power_law_1.01,0.13654784202575682
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,2,1,power_law_1.01,0.46478206634521485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,2,1,power_law_1.01,0.6535167694091797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,2,1,power_law_1.01,0.2662483215332031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,2,1,power_law_1.01,0.8399366760253907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,2,1,power_law_1.01,1.1700141143798828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,2,1,power_law_1.01,1.245966110229492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,2,1,power_law_1.01,0.09234368324279785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,2,1,power_law_1.01,1.2797209930419922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,2,1,power_law_1.01,1.295404815673828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,2,1,power_law_1.01,1.3650866699218749
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,2,1,power_law_1.01,1.4745068359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,2,1,power_law_1.01,1.5667327880859374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,2,1,power_law_1.01,1.6912570190429688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,2,1,power_law_1.01,1.6306611633300783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,2,1,power_law_1.01,1.618231658935547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,2,1,power_law_1.01,1.7361421203613283
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,2,1,power_law_1.01,1.8798841857910156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,2,1,power_law_1.01,2.043937225341797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,2,1,power_law_1.01,2.4065830993652346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,2,1,power_law_1.01,2.7756805419921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,2,1,power_law_1.01,3.3981912231445315
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,2,1,power_law_1.01,4.082344360351563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,2,1,power_law_1.01,5.53017822265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,2,1,power_law_1.01,0.06635776042938232
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,2,1,power_law_1.01,6.985223388671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,2,1,power_law_1.01,0.1446636772155762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,2,1,power_law_1.01,0.24979007720947263
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,2,1,balanced,1.1698662567138673
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,2,1,balanced,1.3183520507812498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,2,1,power_law_1.01,0.3648940658569336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,2,1,power_law_1.01,0.08836864471435547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,2,1,power_law_1.01,9.615714721679687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,2,1,power_law_1.01,0.4814035034179687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,2,1,power_law_1.01,0.6415289306640626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,2,1,power_law_1.01,0.675665283203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,2,1,power_law_1.01,0.6894944000244141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,2,1,power_law_1.01,0.6941165161132813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,2,1,balanced,1.9165548706054687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,2,1,power_law_1.01,0.727286376953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,2,1,balanced,2.4592735290527346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,2,1,power_law_1.01,0.7553964996337891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,2,1,power_law_1.01,0.7625228881835937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,2,1,power_law_1.01,0.8770009613037109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,2,1,power_law_1.01,0.8703340911865235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,2,1,power_law_1.01,0.7726866912841797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,2,1,power_law_1.01,0.9169344329833985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,2,1,power_law_1.01,0.9429004669189454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,2,1,power_law_1.01,1.0178854370117187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,2,1,power_law_1.01,1.116974105834961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,2,1,power_law_1.01,1.2867199707031252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,2,1,power_law_1.01,1.4340658569335938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,2,1,power_law_1.01,12.455897216796874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,2,1,power_law_1.01,1.6916050720214844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,2,1,power_law_1.01,0.06835840225219728
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,2,1,power_law_1.01,2.2201869201660154
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,2,1,power_law_1.01,0.0865760040283203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,2,1,power_law_1.01,0.11249407768249511
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,2,1,power_law_1.01,2.617587890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,2,1,power_law_1.01,0.15894783973693846
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,2,1,power_law_1.01,0.2223276710510254
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,2,1,power_law_1.01,0.27727680206298827
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,2,1,power_law_1.01,0.38313472747802735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,2,1,power_law_1.01,3.6069573974609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,2,1,power_law_1.01,0.4041964721679688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,2,1,power_law_1.01,0.41364097595214844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,2,1,power_law_1.01,0.4195904159545899
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,2,1,power_law_1.01,0.44288894653320315
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,2,1,power_law_1.01,4.536808471679687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,2,1,power_law_1.01,0.45589759826660153
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,2,1,power_law_1.01,0.4628031921386719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,2,1,power_law_1.01,0.4762144088745117
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,2,1,power_law_1.01,0.49457534790039065
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,2,1,power_law_1.01,0.503512954711914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,2,1,power_law_1.01,0.5354022216796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,2,1,power_law_1.01,0.6418630218505859
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,2,1,power_law_1.01,0.5885715103149414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,2,1,power_law_1.01,0.7583999633789062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,2,1,power_law_1.01,1.073330535888672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,2,1,power_law_1.01,0.851822738647461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,2,1,power_law_1.2,0.09369152069091798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,2,1,power_law_1.2,0.13582847595214845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,2,1,power_law_1.2,0.2656601524353027
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,2,1,power_law_1.01,1.330641326904297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,2,1,power_law_1.2,0.4108857727050781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,2,1,power_law_1.01,1.8525894165039063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,2,1,power_law_1.2,0.5982092666625977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,2,1,power_law_1.2,0.7859686279296876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,2,1,power_law_1.01,2.3947731018066407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,2,1,power_law_1.2,1.1547135925292968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,2,1,power_law_1.2,1.2136627197265626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,2,1,power_law_1.2,1.2411743927001955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,2,1,power_law_1.2,1.3024205017089845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,2,1,power_law_1.2,1.3751788330078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,2,1,power_law_1.2,1.4574790954589845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,2,1,power_law_1.2,1.5737869262695312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,2,1,power_law_1.2,1.7484057617187498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,2,1,power_law_1.2,1.6354765319824218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,2,1,power_law_1.2,1.6834841918945311
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,2,1,power_law_1.2,1.7664665222167968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,2,1,power_law_1.2,1.951763916015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,2,1,power_law_1.2,2.1553689575195314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,2,1,power_law_1.2,2.5507725524902343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,2,1,power_law_1.2,2.8056103515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,2,1,power_law_1.2,3.5293127441406247
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,2,1,power_law_1.2,0.06817408084869385
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,2,1,power_law_1.2,4.244827575683594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,2,1,power_law_1.2,0.14717375755310058
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,2,1,power_law_1.2,0.08867263793945312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,2,1,power_law_1.2,0.2356972885131836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,2,1,power_law_1.2,0.3363980865478516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,2,1,power_law_1.2,5.56535400390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,2,1,power_law_1.2,0.4539871978759765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,2,1,power_law_1.2,0.6304742431640624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,2,1,power_law_1.2,0.6648902130126954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,2,1,power_law_1.2,0.6794054412841797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,2,1,power_law_1.2,7.011422119140624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,2,1,power_law_1.2,0.6998751831054688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,2,1,power_law_1.2,0.7227763366699219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,2,1,power_law_1.2,0.7460467529296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,2,1,power_law_1.2,0.7600409698486328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,2,1,power_law_1.2,0.7967481231689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,2,1,power_law_1.2,0.8874777221679688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,2,1,power_law_1.2,0.9097254180908203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,2,1,power_law_1.2,0.934618911743164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,2,1,power_law_1.2,1.0045938873291016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,2,1,power_law_1.2,1.0500332641601564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,2,1,power_law_1.2,9.867605590820313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,2,1,power_law_1.2,1.1770611572265626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,2,1,power_law_1.2,1.3065504455566406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,2,1,power_law_1.2,1.5305087280273437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,2,1,power_law_1.2,1.758278350830078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,2,1,power_law_1.2,2.2760198974609374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,2,1,power_law_1.2,2.717953796386719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,2,1,power_law_1.2,12.441953125000001
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,2,1,power_law_1.2,0.06889472007751465
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,2,1,power_law_1.2,0.10955648422241211
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,2,1,power_law_1.2,0.08620287895202636
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,2,1,power_law_1.2,3.663064880371094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,2,1,power_law_1.2,0.15147647857666016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,2,1,power_law_1.2,0.2052275276184082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,2,1,power_law_1.2,0.26298879623413085
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,2,1,power_law_1.01,3.4314285278320313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,2,1,power_law_1.2,0.3764006423950195
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,2,1,power_law_1.2,0.3991129684448242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,2,1,power_law_1.2,4.593117370605468
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,2,1,power_law_1.2,0.4100076675415039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,2,1,power_law_1.2,0.4243590545654297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,2,1,power_law_1.2,0.4385766220092774
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,2,1,power_law_1.2,0.4533062362670899
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,2,1,power_law_1.2,0.4713766479492188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,2,1,power_law_1.01,4.537582702636719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,2,1,power_law_1.2,0.49099071502685543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,2,1,power_law_1.2,0.5126950454711914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,2,1,power_law_1.2,0.5323020935058593
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,2,1,power_law_1.2,0.5670028686523437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,2,1,power_law_1.2,0.6463929748535155
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,2,1,power_law_1.2,0.7110451507568359
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,2,1,power_law_1.2,0.8775987243652343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,2,balanced,0.022339839935302735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,2,balanced,0.032561919689178466
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,2,1,power_law_1.2,1.036376953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,2,1,power_law_1.2,1.4250483703613281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,2,balanced,0.045543680191040034
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,2,balanced,0.12236415863037109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,2,balanced,0.07028223991394042
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,2,balanced,0.12335743904113769
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,2,balanced,0.12371199607849122
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,2,balanced,0.1403238391876221
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,2,1,power_law_1.2,1.8206297302246093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,2,balanced,0.1406719970703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,2,balanced,0.14237183570861817
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,2,balanced,0.14764927864074706
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,2,balanced,0.14621439933776856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,2,balanced,0.15405823707580565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,2,balanced,0.12244480133056641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,2,balanced,0.1560960006713867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,2,balanced,0.16247039794921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,2,balanced,0.17636863708496092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,2,balanced,0.1906265640258789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,2,balanced,0.26334720611572265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,2,balanced,0.296048641204834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,2,balanced,0.14933631896972657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,2,balanced,0.4105920028686524
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,2,1,power_law_1.2,2.6579928588867188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,2,balanced,0.523900146484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,2,balanced,0.0255679988861084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,2,balanced,0.7534822082519531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,2,balanced,0.026023681163787844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,2,balanced,0.03132416009902954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,2,balanced,0.05747968196868897
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,2,balanced,0.08763008117675782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,2,balanced,0.08971648216247559
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,2,balanced,0.08960127830505371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,2,balanced,0.09012736320495605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,2,balanced,0.09775872230529785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,2,balanced,1.4442227172851563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,2,balanced,0.09880191802978515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,2,balanced,0.09922687530517578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,2,balanced,0.10094592094421387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,2,balanced,0.10182271957397462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,2,balanced,0.10325887680053711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,2,balanced,0.10523648262023926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,2,balanced,0.1056332778930664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,2,balanced,0.10858495712280274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,2,balanced,0.11665920257568359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,2,balanced,0.9826995086669921
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,2,balanced,0.129171199798584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,2,balanced,1.9078182983398437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,2,balanced,0.14472448348999023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,2,balanced,0.1693619155883789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,2,balanced,0.24987007141113282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,2,balanced,0.2911372756958008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,2,balanced,0.4110655975341797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,2,balanced,0.5354547119140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,2,balanced,0.7696141052246094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,2,balanced,1.009384994506836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,1,2,balanced,0.02642047882080078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,1,2,balanced,0.05534207820892334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,1,2,balanced,0.0559168004989624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,1,2,balanced,0.03865216016769409
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,1,2,balanced,0.05641600131988526
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,1,2,balanced,0.031169281005859372
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,1,2,balanced,0.0268339204788208
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,1,2,balanced,0.05672959804534912
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,1,2,balanced,0.05749248027801514
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,1,2,balanced,0.057784318923950195
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,1,2,balanced,0.057116160392761225
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,1,2,balanced,0.05837952136993409
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,1,2,balanced,0.059087362289428715
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,1,2,balanced,0.05569024085998535
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,1,2,balanced,0.060093441009521485
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,1,2,balanced,0.06146304130554199
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,1,2,balanced,0.06323071956634521
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,1,2,balanced,0.06650239944458007
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,1,2,balanced,0.07077504158020019
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,1,2,balanced,0.07604608058929443
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,1,2,balanced,0.05643904209136963
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,2,1,power_law_1.2,3.522653503417969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,1,2,balanced,0.11510784149169921
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,1,2,balanced,0.1967078399658203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,2,power_law_1.01,0.02641792058944702
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,2,power_law_1.01,0.04106880187988281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,2,power_law_1.01,0.05201024055480957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,2,power_law_1.01,0.06571135997772218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,2,power_law_1.01,0.07925888061523437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,2,power_law_1.01,0.11426431655883788
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,2,power_law_1.01,0.11656703948974609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,2,power_law_1.01,0.11922304153442383
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,1,2,balanced,0.10512639999389647
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,1,2,balanced,0.15491840362548828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,2,power_law_1.01,0.13599103927612305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,2,power_law_1.01,0.1365772819519043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,1,2,balanced,0.2856332778930664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,2,power_law_1.01,0.13832703590393067
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,2,power_law_1.01,0.1409280014038086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,2,power_law_1.01,0.14350720405578615
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,2,power_law_1.01,0.1488051223754883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,2,power_law_1.01,0.15531264305114748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,2,power_law_1.01,0.16043264389038087
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,2,power_law_1.01,0.17255807876586915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,2,power_law_1.01,0.20594560623168948
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,2,power_law_1.01,0.22851583480834964
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,2,power_law_1.01,0.2938060760498047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,2,power_law_1.01,0.3517657470703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,2,power_law_1.01,0.4641049575805664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,2,power_law_1.01,0.6045593643188477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,2,power_law_1.01,0.8415692901611328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,2,power_law_1.01,0.025998079776763917
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,2,power_law_1.01,0.03164416074752808
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,2,power_law_1.01,0.042548480033874514
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,1,2,balanced,0.3690739059448242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,2,power_law_1.01,1.067557144165039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,2,power_law_1.01,0.05306367874145508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,2,power_law_1.01,0.06223487854003906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,2,power_law_1.01,0.08286720275878906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,2,power_law_1.01,0.09478272438049316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,2,power_law_1.01,0.09642751693725586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,2,power_law_1.01,1.6702362060546876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,2,power_law_1.01,0.0975820827484131
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,2,power_law_1.01,0.09986559867858887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,2,power_law_1.01,0.10013567924499511
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,2,power_law_1.01,0.10319231986999511
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,2,power_law_1.01,0.10509568214416505
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,2,power_law_1.01,0.10704511642456054
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,2,power_law_1.01,2.1433535766601564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,2,power_law_1.01,0.11108223915100097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,2,power_law_1.01,0.12363007545471191
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,2,power_law_1.01,0.14030079841613768
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,2,power_law_1.01,0.17085439682006837
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,2,power_law_1.01,0.2011622428894043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,2,power_law_1.01,0.26270719528198244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,2,power_law_1.01,0.3310745620727539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,2,power_law_1.01,0.08069120407104492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,2,power_law_1.01,0.4584972763061524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,2,power_law_1.01,0.5766771316528321
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,1,2,power_law_1.01,0.030485761165618897
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,1,2,power_law_1.01,0.034434559345245364
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,1,2,power_law_1.01,0.03810431957244873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,2,power_law_1.01,0.8322342681884767
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,1,2,power_law_1.01,0.04213759899139404
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,1,2,power_law_1.01,0.050958080291748045
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,1,2,power_law_1.01,0.05246208190917969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,1,2,power_law_1.01,0.05371520042419433
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,2,power_law_1.01,1.0966143798828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,1,2,power_law_1.01,0.05398399829864502
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,1,2,power_law_1.01,0.05579391956329346
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,1,2,power_law_1.01,0.054963197708129886
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,1,2,power_law_1.01,0.05702015876770019
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,1,2,power_law_1.01,0.057114877700805665
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,1,2,power_law_1.01,0.05966720104217529
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,1,2,power_law_1.01,0.06369279861450196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,2,power_law_1.01,0.08518912315368651
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,1,2,power_law_1.01,0.02724992036819458
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,1,2,power_law_1.01,0.06227071762084961
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,1,2,power_law_1.01,0.06835328102111816
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,1,2,power_law_1.01,0.08359935760498047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,1,2,power_law_1.01,0.08057727813720703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,1,2,power_law_1.01,0.10837120056152343
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,1,2,power_law_1.01,0.13204480171203614
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,1,2,power_law_1.01,0.17027456283569337
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,1,2,power_law_1.01,0.221213436126709
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,2,power_law_1.2,0.0397542405128479
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,1,2,power_law_1.01,0.32074623107910155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,2,power_law_1.2,0.02717695951461792
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,2,power_law_1.2,0.049585919380187984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,2,power_law_1.2,0.055788798332214354
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,2,power_law_1.2,0.07472256183624268
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,2,power_law_1.2,0.11368576049804688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,2,power_law_1.2,0.12108799934387207
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,1,2,power_law_1.01,0.4279296112060547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,2,power_law_1.2,0.13728896141052246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,2,power_law_1.2,0.13870976448059083
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,2,power_law_1.2,0.14106752395629882
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,2,power_law_1.2,0.14441984176635742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,2,power_law_1.2,0.14613375663757325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,2,power_law_1.2,0.151779842376709
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,2,power_law_1.2,0.11801600456237793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,2,power_law_1.2,0.1579916763305664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,2,power_law_1.2,0.16284671783447266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,2,power_law_1.2,0.17361791610717775
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,2,power_law_1.2,0.2072640037536621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,2,power_law_1.2,0.2333363151550293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,2,power_law_1.2,0.29469440460205076
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,2,power_law_1.2,0.3575283050537109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,2,power_law_1.2,0.47320064544677737
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,2,power_law_1.2,0.6174067306518555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,1,2,power_law_1.01,0.6630194854736329
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,2,power_law_1.2,0.02563711881637573
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,2,power_law_1.2,0.03214591979980469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,2,power_law_1.2,0.03891583919525147
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,2,power_law_1.2,0.05179647922515869
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,2,power_law_1.2,0.055000319480895996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,2,power_law_1.2,0.07988863945007324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,2,power_law_1.2,1.090740509033203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,2,power_law_1.2,0.08535039901733399
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,2,power_law_1.2,0.08833151817321777
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,2,power_law_1.2,0.09706496238708497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,2,power_law_1.2,0.09745408058166503
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,2,power_law_1.2,0.09750783920288086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,2,power_law_1.2,0.09962623596191407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,2,power_law_1.2,0.10195455551147461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,2,power_law_1.2,0.1026636791229248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,2,power_law_1.2,0.10657792091369629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,2,power_law_1.2,0.10604288101196288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,2,power_law_1.2,0.11245311737060545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,2,power_law_1.2,0.12827008247375488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,2,power_law_1.2,0.14275456428527833
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,2,power_law_1.2,0.17603071212768556
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,2,power_law_1.2,0.20308864593505857
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,2,power_law_1.2,0.26571136474609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,2,power_law_1.2,1.661817626953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,2,power_law_1.2,0.34268161773681644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,2,power_law_1.2,0.8419443511962891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,2,power_law_1.2,0.5830579376220703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,1,2,power_law_1.2,0.02713344097137451
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,1,2,power_law_1.2,0.02969088077545166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,2,power_law_1.2,0.46952190399169924
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,1,2,power_law_1.2,0.03379456043243408
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,1,2,power_law_1.2,0.03658112049102783
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,1,2,power_law_1.2,0.03984895944595337
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,1,2,power_law_1.2,0.05138432025909424
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,1,2,power_law_1.2,0.053084158897399904
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,1,2,power_law_1.2,0.05353983879089356
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,1,2,power_law_1.2,0.05606143951416016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,1,2,power_law_1.2,0.05555840015411377
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,1,2,power_law_1.2,0.05769216060638428
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,1,2,power_law_1.01,0.7592332458496094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,1,2,power_law_1.2,0.05823872089385986
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,1,2,power_law_1.2,0.0594982385635376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,1,2,power_law_1.2,0.06606592178344727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,2,power_law_1.2,2.202750701904297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,2,power_law_1.2,1.1215526580810546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,1,2,power_law_1.2,0.06114943981170654
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,1,2,power_law_1.2,0.06895232200622559
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,1,2,power_law_1.2,0.07088640213012695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,1,2,power_law_1.2,0.08565376281738282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,2,power_law_1.2,0.8682828521728515
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,1,2,power_law_1.2,0.09979647636413574
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,1,2,power_law_1.2,0.22291072845458984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,1,2,power_law_1.2,0.1451084804534912
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,1,balanced,0.027576959133148192
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,1,balanced,0.05485504150390626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,1,balanced,0.0780736017227173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,1,balanced,0.13303872108459475
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,1,2,power_law_1.2,0.32225025177001954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,1,balanced,0.132805757522583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,1,balanced,0.13472000122070313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,1,balanced,0.1346348762512207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,1,balanced,0.13556672096252442
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,1,balanced,0.13428288459777832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,1,balanced,0.13554176330566406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,1,balanced,0.15225024223327638
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,1,balanced,0.15430399894714356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,1,balanced,0.15431103706359864
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,1,balanced,0.15814592361450194
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,1,balanced,0.15978431701660156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,1,balanced,0.16582847595214845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,1,balanced,0.17781951904296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,1,balanced,0.1914726448059082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,1,balanced,0.24821311950683594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,1,balanced,0.27623104095458983
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,1,balanced,0.37624191284179687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,1,balanced,0.4753900909423828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,1,2,power_law_1.2,0.15398783683776857
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,1,2,power_law_1.2,0.46843521118164066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,1,balanced,0.6803609466552734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,1,balanced,0.03204735994338989
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,1,balanced,0.8930169677734374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,1,balanced,0.035317120552062986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,1,balanced,0.04273536205291748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,1,balanced,0.07196415901184082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,1,balanced,0.10759167671203614
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,1,balanced,0.10892928123474122
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,1,balanced,1.7105619812011716
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,1,balanced,0.10889472007751463
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,1,balanced,0.10975359916687011
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,1,balanced,0.1096735954284668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,1,balanced,0.10991680145263671
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,1,balanced,0.10921792030334472
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,1,balanced,0.12376447677612305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,1,balanced,0.1242636775970459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,1,balanced,0.12596927642822267
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,1,balanced,0.12592320442199706
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,1,balanced,0.1285811233520508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,1,balanced,0.13256704330444335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,1,balanced,0.13848959922790527
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,1,balanced,0.14515328407287598
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,1,balanced,0.1621683120727539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,1,balanced,0.17519487380981444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,1,balanced,0.25851968765258787
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,1,balanced,0.28999168395996094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,1,balanced,0.4001862335205078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,1,balanced,0.5204716873168945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,1,balanced,1.297051544189453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,1,balanced,0.7510963439941406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,1,balanced,0.981945571899414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,1,2,power_law_1.2,0.6190348815917969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,1,balanced,0.036563839912414554
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,2,1,balanced,0.037114241123199464
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,2,1,balanced,0.042906880378723145
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,2,1,balanced,0.03761215925216675
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,2,1,balanced,0.0717248010635376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,2,1,balanced,0.0722976016998291
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,2,1,balanced,0.07235904216766358
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,2,1,balanced,0.05217855930328369
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,2,1,balanced,0.07241280078887939
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,2,1,balanced,0.07274367809295654
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,2,1,balanced,0.07389120101928712
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,2,1,balanced,0.07486591815948487
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,2,1,balanced,0.07187520027160645
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,2,1,balanced,0.0751532793045044
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,2,1,balanced,0.07683263778686523
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,2,1,balanced,0.07783487796783448
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,2,1,balanced,0.0804319953918457
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,2,1,balanced,0.07015935897827148
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,2,1,balanced,0.09004799842834472
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,2,1,balanced,0.0980025577545166
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,2,1,balanced,0.10709440231323242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,2,1,balanced,0.08470272064208985
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,1,2,power_law_1.2,0.953623046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,1,power_law_1.01,0.029548799991607665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,1,power_law_1.01,0.040976638793945315
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,1,power_law_1.01,0.05474112033843994
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,1,power_law_1.01,0.0699942398071289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,1,power_law_1.01,0.08109951972961425
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,1,power_law_1.01,0.12444479942321778
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,1,power_law_1.01,0.13027071952819824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,1,power_law_1.01,0.13151679992675783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,1,power_law_1.01,0.13237055778503418
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,1,2,power_law_1.2,1.213601303100586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,1,power_law_1.01,0.13025152206420898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,1,power_law_1.01,0.1386131191253662
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,1,power_law_1.01,0.14820735931396484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,1,power_law_1.01,0.150316801071167
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,1,power_law_1.01,0.15465472221374513
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,1,power_law_1.01,0.1592140769958496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,1,power_law_1.01,0.16103744506835938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,1,power_law_1.01,0.16848127365112303
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,1,power_law_1.01,0.19923200607299804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,1,power_law_1.01,0.2174143981933594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,1,power_law_1.01,0.2665056037902832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,1,power_law_1.01,0.3129670333862305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,1,power_law_1.01,0.4138924789428711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,1,power_law_1.01,0.5160870361328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,1,power_law_1.01,0.7231436920166016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,1,power_law_1.01,0.9324736022949219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,1,power_law_1.01,1.3523898315429688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,1,power_law_1.01,0.03247488021850586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,1,power_law_1.01,0.03572607994079589
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,1,power_law_1.01,0.04660031795501709
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,1,power_law_1.01,1.7841900634765626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,1,power_law_1.01,0.058674559593200684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,1,power_law_1.01,0.07605631828308106
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,1,power_law_1.01,0.10194239616394044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,1,power_law_1.01,0.09749567985534668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,1,power_law_1.01,0.1039699172973633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,1,power_law_1.01,0.10623488426208497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,1,power_law_1.01,0.10707648277282715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,1,power_law_1.01,0.10691519737243653
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,1,power_law_1.01,0.12257920265197755
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,1,power_law_1.01,0.12243264198303223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,1,power_law_1.01,0.12712703704833986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,1,power_law_1.01,0.12787391662597655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,1,power_law_1.01,0.13203328132629394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,1,power_law_1.01,0.13888192176818848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,1,power_law_1.01,0.14697152137756347
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,1,power_law_1.01,0.16295808792114258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,1,power_law_1.01,0.18401407241821288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,1,power_law_1.01,0.21617664337158202
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,1,power_law_1.01,0.2704755210876465
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,2,1,balanced,0.16939136505126953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,1,power_law_1.01,0.3248390579223633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,2,1,balanced,0.23480768203735353
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,2,1,balanced,0.15097663879394532
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,2,1,balanced,0.3012428855895996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,2,1,balanced,0.43580287933349615
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,1,power_law_1.01,0.5600940704345703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,1,power_law_1.01,0.4420390319824219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,2,1,power_law_1.01,0.03696703910827637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,1,power_law_1.01,0.7952185821533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,2,1,power_law_1.01,0.053004159927368164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,2,1,power_law_1.01,0.04474624156951904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,1,power_law_1.01,1.026929931640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,2,1,balanced,0.5738265609741211
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,2,1,power_law_1.01,0.06737343788146972
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,2,1,power_law_1.01,0.06893631935119629
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,2,1,power_law_1.01,0.07013887882232665
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,2,1,power_law_1.01,0.07068160057067871
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,2,1,power_law_1.01,0.07303296089172363
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,2,1,power_law_1.01,0.07515135765075684
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,2,1,power_law_1.01,0.0761036777496338
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,2,1,power_law_1.01,0.07997312068939208
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,2,1,power_law_1.01,0.0839731216430664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,2,1,power_law_1.01,0.06541823863983154
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,2,1,power_law_1.01,0.03940479993820191
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,2,1,power_law_1.01,0.048751358985900876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,2,1,power_law_1.01,0.10906432151794435
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,2,1,power_law_1.01,0.08720447540283202
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,2,1,power_law_1.01,0.09444479942321778
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,2,1,power_law_1.01,0.12522496223449706
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,2,1,power_law_1.01,0.1586355209350586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,1,power_law_1.2,0.028761599063873287
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,2,1,power_law_1.01,0.1951180839538574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,1,power_law_1.2,0.04185728073120117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,1,power_law_1.2,0.05426496028900146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,1,power_law_1.2,0.06663424015045166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,1,power_law_1.2,0.07766848087310792
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,1,power_law_1.2,0.1267046356201172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,1,power_law_1.2,0.13062848091125487
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,1,power_law_1.2,0.13185343742370606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,1,power_law_1.2,0.13255743980407714
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,1,power_law_1.2,0.13193920135498047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,1,power_law_1.2,0.1404582405090332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,1,power_law_1.2,0.15101311683654786
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,1,power_law_1.2,0.15179136276245117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,1,power_law_1.2,0.15724608421325684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,1,power_law_1.2,0.1597548770904541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,1,power_law_1.2,0.1633657646179199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,1,power_law_1.2,0.17095935821533204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,1,power_law_1.2,0.20244480133056642
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,1,power_law_1.2,0.2198464012145996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,1,power_law_1.2,0.27222015380859377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,1,power_law_1.2,0.3153183937072754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,1,power_law_1.2,0.4161094284057617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,1,power_law_1.2,0.5203801727294921
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,1,power_law_1.2,0.727737579345703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,2,1,power_law_1.01,0.26674495697021483
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,2,1,power_law_1.01,0.3427033615112305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,1,power_law_1.2,0.031624319553375246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,1,power_law_1.2,0.0374726390838623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,1,power_law_1.2,0.046117758750915526
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,1,power_law_1.2,0.054625921249389656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,2,1,power_law_1.01,0.5020902252197266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,1,power_law_1.2,0.0734668779373169
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,1,power_law_1.2,0.10274880409240723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,1,power_law_1.2,0.10501376152038575
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,2,1,power_law_1.01,0.6652998352050782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,1,power_law_1.2,0.10683327674865724
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,2,1,power_law_1.01,1.0006591796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,1,power_law_1.2,0.1073964786529541
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,2,1,power_law_1.01,1.34114501953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,1,power_law_1.2,0.10737792015075684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,1,power_law_1.2,0.12278592109680175
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,1,power_law_1.2,0.12332287788391114
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,1,power_law_1.2,0.13006719589233398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,1,power_law_1.2,0.13313535690307615
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,1,power_law_1.2,0.1401535987854004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,1,power_law_1.2,0.12825535774230956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,1,power_law_1.2,0.15188096046447755
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,1,power_law_1.2,0.9363276672363281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,1,power_law_1.2,0.165229434967041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,1,power_law_1.2,0.1946873664855957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,1,power_law_1.2,0.21763135910034181
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,1,power_law_1.2,0.2714137649536133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,1,power_law_1.2,0.33372161865234373
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,2,1,power_law_1.2,0.03696576118469238
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,1,power_law_1.2,0.09841088294982911
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,2,1,power_law_1.2,0.03996288061141968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,2,1,power_law_1.2,0.04465919971466065
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,2,1,power_law_1.2,0.04747903823852539
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,2,1,power_law_1.2,0.05168896198272706
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,2,1,power_law_1.2,0.06521599769592286
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,1,power_law_1.2,0.44448513031005865
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,2,1,power_law_1.2,0.0678656005859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,2,1,power_law_1.2,0.0687724781036377
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,2,1,power_law_1.2,0.07105216026306152
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,2,1,power_law_1.2,0.07121920108795166
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,2,1,power_law_1.2,0.07377408027648927
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,2,1,power_law_1.2,0.07670144081115723
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,2,1,power_law_1.2,0.07851776123046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,2,1,power_law_1.2,0.0831328010559082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,2,1,power_law_1.2,0.0923795223236084
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,2,1,power_law_1.2,0.10161024093627931
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,1,power_law_1.2,0.5643328094482423
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,1,power_law_1.2,1.3672384643554687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,2,1,power_law_1.2,0.0879916763305664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,2,1,power_law_1.2,0.12478336334228515
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,2,1,power_law_1.2,0.14808639526367187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,1,power_law_1.2,0.7993138885498047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,2,1,power_law_1.2,0.19999616622924804
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,2,1,power_law_1.2,0.25091840744018556
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,2,1,power_law_1.2,0.3628012847900391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,2,balanced,0.05992320060729981
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,2,balanced,0.0794163179397583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,2,balanced,0.13178624153137206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,2,balanced,0.2323801612854004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,2,balanced,0.42903167724609376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,2,balanced,0.4293388748168946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,2,balanced,0.43235328674316403
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,2,1,power_law_1.2,0.4748729705810547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,2,balanced,0.43333889007568355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,2,balanced,0.45865215301513673
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,2,balanced,0.4621235275268555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,2,balanced,0.45868160247802736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,2,balanced,0.4699635314941406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,2,balanced,0.469043197631836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,2,balanced,0.4763699340820312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,2,balanced,0.4872396850585938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,2,balanced,0.48816001892089844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,2,balanced,0.5061248016357422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,2,balanced,0.5345663833618164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,2,balanced,0.5766630554199219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,2,balanced,0.8283071899414063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,2,1,power_law_1.2,0.7006143951416016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,2,balanced,0.9219302368164062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,1,power_law_1.2,1.7649087524414064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,2,balanced,1.3425010681152343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,2,balanced,1.7245478820800781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,2,balanced,2.484083251953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,2,balanced,0.03753472089767456
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,2,balanced,0.05089663982391357
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,2,balanced,0.08332927703857422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,2,balanced,0.13189375877380372
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,2,balanced,3.1644110107421874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,2,balanced,0.22964223861694336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,2,balanced,0.2342336082458496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,2,balanced,0.2350758361816406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,2,balanced,0.23987327575683595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,2,balanced,0.2595353507995605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,2,balanced,0.260897274017334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,2,balanced,0.2598374366760254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,2,balanced,0.26319360733032227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,2,balanced,0.2631820869445801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,2,balanced,0.2655692863464355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,2,balanced,0.26589824676513674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,2,balanced,5.011080932617188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,2,balanced,0.26768768310546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,2,balanced,0.27350271224975586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,2,balanced,0.29516672134399413
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,2,balanced,0.3298803329467773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,2,balanced,0.3458099365234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,2,balanced,0.41066879272460943
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,2,balanced,0.5758156967163086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,2,balanced,0.6504102325439453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,1,power_law_1.2,1.0287206268310547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,2,1,power_law_1.2,0.9300454711914062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,2,balanced,0.9282303619384766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,2,balanced,6.844140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,2,balanced,1.1927308654785156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,2,balanced,1.7505535888671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,2,balanced,2.2661721801757815
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,2,1,power_law_1.2,1.3821101379394531
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,1,2,balanced,0.029710080623626713
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,1,2,balanced,0.14492032051086426
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,1,2,balanced,0.03854847908020019
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,1,2,balanced,0.05476992130279541
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,1,2,balanced,0.14531968116760255
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,1,2,balanced,0.1462054443359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,1,2,balanced,0.14601856231689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,1,2,balanced,0.14646016120910646
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,1,2,balanced,0.14769920349121093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,1,2,balanced,0.1487180805206299
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,1,2,balanced,0.14945023536682128
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,1,2,balanced,0.1522099208831787
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,1,2,balanced,0.15562623977661133
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,1,2,balanced,0.08444160461425781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,1,2,balanced,0.16648191452026367
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,1,2,balanced,0.17673215866088868
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,1,2,balanced,0.18745599746704103
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,1,2,balanced,0.1436953639984131
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,1,2,balanced,0.16072832107543947
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,1,2,balanced,0.14396544456481936
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,1,2,balanced,0.1448153591156006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,2,power_law_1.01,0.07476736068725585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,2,power_law_1.01,0.10718720436096192
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,2,power_law_1.01,0.1689676856994629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,2,power_law_1.01,0.21496192932128907
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,1,2,balanced,0.26705408096313477
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,1,2,balanced,0.2899942398071289
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,1,2,balanced,0.40570625305175784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,2,power_law_1.01,0.26227455139160155
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,1,2,balanced,0.517017593383789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,2,power_law_1.01,0.38213504791259767
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,2,power_law_1.01,0.38612094879150394
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,1,2,balanced,0.7537216186523438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,2,power_law_1.01,0.4404032135009766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,2,power_law_1.01,0.4415232086181641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,2,power_law_1.01,0.45121791839599606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,2,power_law_1.01,0.4562368011474609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,2,power_law_1.01,0.4679232025146485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,2,power_law_1.01,0.36139137268066407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,2,power_law_1.01,0.4999385452270507
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,2,power_law_1.01,0.5212351989746093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,1,2,balanced,1.051878433227539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,2,power_law_1.01,0.5319654464721679
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,2,power_law_1.01,0.5492095947265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,2,power_law_1.01,0.6360332870483398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,2,power_law_1.01,0.7224845123291016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,2,power_law_1.01,0.9437734222412109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,2,power_law_1.01,1.146466522216797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,2,power_law_1.01,0.05066112041473388
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,2,power_law_1.01,1.5668453979492187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,2,power_law_1.01,0.07834496021270751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,2,power_law_1.01,1.997216033935547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,2,power_law_1.01,0.09903231620788575
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,2,power_law_1.01,0.12501503944396972
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,2,power_law_1.01,0.20900352478027343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,2,power_law_1.01,0.21871488571166992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,2,power_law_1.01,2.767174377441406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,2,power_law_1.01,0.21983999252319336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,2,power_law_1.01,0.24173568725585937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,2,power_law_1.01,0.24595968246459962
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,2,power_law_1.01,0.24613632202148436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,2,power_law_1.01,0.24972160339355468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,2,power_law_1.01,0.26492416381835937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,2,power_law_1.01,3.761446533203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,2,power_law_1.01,0.2582156753540039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,2,power_law_1.01,0.26884607315063475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,2,power_law_1.01,0.28314111709594725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,2,power_law_1.01,0.30611839294433596
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,2,power_law_1.01,0.3451558303833008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,2,power_law_1.01,0.40990337371826174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,2,power_law_1.01,0.47651840209960933
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,2,power_law_1.01,0.6205260848999024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,2,power_law_1.01,5.584075317382813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,2,power_law_1.01,0.765000991821289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,2,power_law_1.01,1.0589453125000001
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,1,2,power_law_1.01,0.0412556791305542
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,2,power_law_1.01,1.350145263671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,1,2,power_law_1.01,0.05079040050506591
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,2,power_law_1.01,6.60967041015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,1,2,power_law_1.01,0.07056000232696533
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,1,2,power_law_1.01,0.0732480001449585
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,1,2,power_law_1.01,0.08902015686035156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,1,2,power_law_1.01,0.12324480056762695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,1,2,power_law_1.01,0.13097472190856935
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,1,2,power_law_1.01,0.1331110382080078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,2,power_law_1.01,2.0841445922851562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,1,2,power_law_1.01,0.1354854393005371
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,1,2,power_law_1.01,0.13779711723327637
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,1,2,power_law_1.01,0.14312191963195803
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,1,2,power_law_1.01,0.14633215904235838
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,1,2,power_law_1.01,0.15603584289550781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,2,power_law_1.01,2.577927551269531
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,1,2,power_law_1.01,0.15853568077087404
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,1,2,power_law_1.01,0.14814080238342284
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,2,power_law_1.01,0.14402175903320313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,2,power_law_1.01,0.26178688049316406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,2,1,power_law_1.2,1.834279022216797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,1,2,power_law_1.01,0.1636300849914551
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,1,2,power_law_1.01,0.22278272628784182
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,1,2,power_law_1.01,0.1771801567077637
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,1,2,power_law_1.01,0.19240192413330076
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,1,2,power_law_1.01,0.27511167526245117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,2,power_law_1.2,0.0746073579788208
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,2,power_law_1.2,0.09703807830810547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,2,power_law_1.2,0.15009407997131347
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,2,power_law_1.2,0.19804800033569336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,2,power_law_1.2,0.23457151412963867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,1,2,power_law_1.01,0.33238784790039066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,2,power_law_1.2,0.35012863159179686
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,1,2,power_law_1.01,0.4368064117431641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,2,power_law_1.2,0.3824806213378906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,2,power_law_1.2,0.41356159210205073
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,2,power_law_1.2,0.4467161560058594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,1,2,power_law_1.01,0.5247014236450196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,2,power_law_1.2,0.4539020919799805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,2,power_law_1.2,0.4695091247558594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,2,power_law_1.2,0.5136460876464843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,2,power_law_1.2,0.48391040802001956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,2,power_law_1.2,0.44391937255859376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,2,power_law_1.2,0.5268991851806641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,2,power_law_1.2,0.5368960189819336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,2,power_law_1.2,0.5760153579711914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,2,power_law_1.2,0.6400217437744141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,2,power_law_1.2,0.7548595428466797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,2,power_law_1.2,0.968652801513672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,2,power_law_1.2,1.1813990020751954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,2,power_law_1.2,1.4975680541992187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,2,power_law_1.2,2.0581759643554687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,2,power_law_1.2,2.8744833374023435
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,2,power_law_1.2,0.05171711921691895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,2,power_law_1.2,0.0701632022857666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,2,power_law_1.2,0.0970905590057373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,2,power_law_1.2,0.11911168098449706
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,2,power_law_1.2,3.8801675415039063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,2,power_law_1.2,0.1436684799194336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,2,power_law_1.2,0.20207744598388672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,2,power_law_1.2,0.2232691192626953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,2,power_law_1.2,0.24406911849975588
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,2,power_law_1.2,0.24942592620849607
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,2,power_law_1.2,0.21765888214111326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,2,power_law_1.2,0.24711296081542972
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,2,power_law_1.2,0.24567167282104493
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,1,2,power_law_1.01,0.8075878143310546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,2,power_law_1.2,0.24246400833129883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,2,power_law_1.2,0.2599385643005371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,2,power_law_1.2,0.2629388809204102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,2,power_law_1.2,0.27101184844970705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,2,power_law_1.2,0.2825484848022461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,2,power_law_1.2,0.31505279541015624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,2,power_law_1.2,0.35202686309814457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,2,power_law_1.2,0.4198783874511719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,2,power_law_1.2,0.4840678405761719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,1,2,power_law_1.01,1.0752486419677734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,1,2,power_law_1.01,1.6117184448242188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,2,power_law_1.2,0.6314342498779297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,1,2,power_law_1.01,2.3107481384277344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,2,power_law_1.2,0.7953715515136719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,2,power_law_1.2,1.1080524444580078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,1,2,power_law_1.2,0.040920319557189944
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,1,2,power_law_1.2,0.06692992210388184
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,1,2,power_law_1.2,0.0671129608154297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,1,2,power_law_1.2,0.0841420841217041
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,1,2,power_law_1.2,0.1242623996734619
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,1,2,power_law_1.2,0.13178367614746095
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,1,2,power_law_1.2,0.136878080368042
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,2,power_law_1.2,5.602598266601563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,1,2,power_law_1.2,0.1395456027984619
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,1,2,power_law_1.2,0.1426841640472412
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,1,2,power_law_1.2,0.15074432373046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,1,2,power_law_1.2,0.15063551902770994
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,1,2,power_law_1.2,0.15402239799499512
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,1,2,power_law_1.2,0.1625164794921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,1,2,power_law_1.2,0.16671615600585937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,1,2,power_law_1.2,0.17394432067871093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,2,power_law_1.2,1.4394163513183593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,2,power_law_1.2,1.9667622375488283
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,2,power_law_1.2,2.6250021362304685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,2,power_law_1.2,6.940411987304688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,1,2,power_law_1.2,0.049986557960510256
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,1,2,power_law_1.2,0.2211929512023926
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,1,2,power_law_1.2,0.25038848876953124
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,1,2,power_law_1.2,0.1888115119934082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,1,2,power_law_1.2,0.3276544189453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,1,balanced,0.06422527790069579
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,1,balanced,0.08610239982604981
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,1,balanced,0.13729920387268066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,1,balanced,0.23379199981689452
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,1,balanced,0.43271232604980464
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,1,balanced,0.4343993759155273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,1,balanced,0.43528064727783206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,1,balanced,0.4362598419189453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,1,balanced,0.43816318511962893
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,1,balanced,0.43866497039794916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,1,balanced,0.44290302276611326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,1,balanced,0.4670175933837891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,1,balanced,0.47298110961914064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,1,balanced,0.4783871841430664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,1,balanced,0.48007808685302733
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,1,2,power_law_1.2,0.415863037109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,1,balanced,0.48202175140380865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,1,balanced,0.4932121658325196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,1,balanced,0.5343641662597657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,1,2,power_law_1.2,0.5645708847045898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,1,balanced,0.5696723175048828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,1,balanced,0.8317126464843749
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,1,balanced,0.8862348937988281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,1,balanced,1.2259519958496095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,1,balanced,1.655023956298828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,1,balanced,2.3081350708007813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,1,2,power_law_1.2,0.7037773132324219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,1,balanced,0.04220416069030762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,1,balanced,0.060195841789245606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,1,balanced,0.09564160346984864
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,1,balanced,3.1478118896484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,1,balanced,0.1472275161743164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,1,balanced,0.2678144073486328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,1,balanced,0.2567852783203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,1,balanced,0.2621932792663574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,1,balanced,0.26267648696899415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,1,balanced,0.2647097587585449
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,1,balanced,0.26422719955444335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,1,balanced,0.26506303787231444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,1,balanced,0.2989465522766113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,1,balanced,0.29873023986816405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,1,balanced,4.54852294921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,1,balanced,0.300447998046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,1,balanced,0.30442880630493163
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,1,balanced,0.31164735794067383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,1,balanced,0.3024390411376953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,1,balanced,0.3263884735107422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,1,balanced,0.3446598434448242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,1,balanced,0.3709779357910156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,1,balanced,0.40131649017333987
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,1,2,power_law_1.2,1.0908019256591797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,1,balanced,5.952772216796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,1,balanced,0.591357421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,1,balanced,0.6581715393066406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,1,balanced,0.933154525756836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,1,balanced,1.2068236541748045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,1,balanced,1.7635653686523436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,1,balanced,2.262487030029297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,1,2,power_law_1.2,1.4349440002441407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,1,2,power_law_1.2,2.078873596191406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,2,1,balanced,0.040227842330932614
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,2,1,balanced,0.04911744117736817
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,2,1,balanced,0.15861760139465333
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,2,1,balanced,0.0635641622543335
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,2,1,balanced,0.15818943977355956
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,2,1,balanced,0.16108095169067382
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,2,1,balanced,0.09427264213562012
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,2,1,balanced,0.16133119583129882
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,2,1,balanced,0.16122432708740236
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,2,1,balanced,0.16433855056762697
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,2,1,balanced,0.16554624557495118
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,2,1,balanced,0.15711935997009277
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,2,1,balanced,0.1672172737121582
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,2,1,balanced,0.1696294403076172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,2,1,balanced,0.17526208877563476
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,2,1,balanced,0.18268543243408203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,2,1,balanced,0.19142400741577148
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,2,1,balanced,0.20900480270385743
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,2,1,balanced,0.22550527572631837
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,2,1,balanced,0.1568607997894287
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,2,1,balanced,0.1575494384765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,1,power_law_1.01,0.06250751972198486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,1,power_law_1.01,0.08637824058532714
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,1,power_law_1.01,0.1299244785308838
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,1,power_law_1.01,0.20924800872802735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,1,power_law_1.01,0.23622144699096678
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,1,power_law_1.01,0.351143684387207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,1,power_law_1.01,0.37741695404052733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,1,power_law_1.01,0.3896851348876953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,1,power_law_1.01,0.4108083343505859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,1,power_law_1.01,0.4146527862548828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,1,power_law_1.01,0.44303615570068355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,1,power_law_1.01,0.450300178527832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,1,power_law_1.01,0.45464321136474606
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,2,1,balanced,0.3271308898925781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,1,power_law_1.01,0.4848518371582031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,2,1,balanced,0.5184921646118165
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,2,1,balanced,0.364830093383789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,1,power_law_1.01,0.5125446319580078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,1,power_law_1.01,0.5164179229736329
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,2,1,balanced,0.6976959991455078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,1,power_law_1.01,0.5338387298583984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,2,1,balanced,1.0476595306396486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,1,power_law_1.01,0.6051859283447266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,1,power_law_1.01,0.6570988464355468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,1,power_law_1.01,0.8653702545166017
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,1,power_law_1.01,1.0369593811035156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,1,2,power_law_1.2,3.0948147583007812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,1,power_law_1.01,1.4079615783691406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,1,power_law_1.01,0.04
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,1,power_law_1.01,0.06454016208648682
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,1,power_law_1.01,1.7707731628417966
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,1,power_law_1.01,0.09651328086853027
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,1,power_law_1.01,0.12437248229980469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,1,power_law_1.01,0.16322559356689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,1,power_law_1.01,0.22326271057128907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,1,power_law_1.01,0.23531455993652345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,1,power_law_1.01,2.4644435119628905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,1,power_law_1.01,0.23881408691406253
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,1,power_law_1.01,0.24567167282104493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,1,power_law_1.01,0.24621376037597656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,1,power_law_1.01,0.2620895957946777
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,1,power_law_1.01,0.2825081634521484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,1,power_law_1.01,0.2844038391113281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,1,power_law_1.01,0.2974483108520508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,1,power_law_1.01,0.30833984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,1,power_law_1.01,3.2522964477539062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,1,power_law_1.01,0.3057081604003906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,1,power_law_1.01,0.32111873626708987
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,1,power_law_1.01,0.3384320068359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,1,power_law_1.01,0.36711296081542966
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,1,power_law_1.01,0.4298591995239258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,1,power_law_1.01,0.4970355224609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,1,power_law_1.01,0.6178086471557618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,1,power_law_1.01,0.74932861328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,1,power_law_1.01,4.671916809082031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,1,power_law_1.01,1.003601303100586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,2,1,power_law_1.01,0.03942271947860718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,1,power_law_1.01,1.2945741271972655
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,2,1,power_law_1.01,0.05308608055114746
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,2,1,power_law_1.01,0.06421631813049317
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,1,power_law_1.01,1.8307225036621095
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,2,1,power_law_1.01,0.07973120212554932
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,2,1,power_law_1.01,0.0942841625213623
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,2,1,power_law_1.01,0.13380672454833983
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,2,1,power_law_1.01,0.14254015922546387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,1,power_law_1.01,6.200320434570313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,2,1,power_law_1.01,0.14640959739685058
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,2,1,power_law_1.01,0.15172224044799804
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,2,1,power_law_1.01,0.15198911666870116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,1,power_law_1.01,2.36911865234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,2,1,power_law_1.01,0.15841983795166015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,2,1,power_law_1.01,0.16219711303710938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,2,1,power_law_1.01,0.1648134422302246
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,2,1,power_law_1.01,0.17412992477416994
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,2,1,power_law_1.01,0.20608896255493164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,2,1,balanced,1.3768818664550782
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,2,1,power_law_1.01,0.18467840194702148
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,2,1,power_law_1.01,0.19089216232299805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,2,1,power_law_1.01,0.27415744781494145
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,2,1,power_law_1.01,0.23904384613037108
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,2,1,power_law_1.01,0.3515545654296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,2,1,power_law_1.01,0.43880577087402345
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,2,1,power_law_1.01,0.6047571182250977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,1,power_law_1.2,0.0642739200592041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,1,power_law_1.2,0.08460672378540039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,1,power_law_1.2,0.12568960189819336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,1,power_law_1.2,0.19066816329956054
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,1,power_law_1.2,0.22934080123901368
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,1,power_law_1.2,0.35739200592041015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,2,1,power_law_1.01,0.7783257293701171
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,1,power_law_1.2,0.3888896179199219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,1,power_law_1.2,0.40203582763671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,1,power_law_1.2,0.4180582427978516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,1,power_law_1.2,0.41765182495117187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,1,power_law_1.2,0.447880973815918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,1,power_law_1.2,0.45918399810791016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,1,power_law_1.2,0.46215167999267576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,1,power_law_1.2,0.503191032409668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,1,power_law_1.2,0.5190892791748046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,1,power_law_1.2,0.5207513427734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,1,power_law_1.2,0.5391065597534179
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,1,power_law_1.2,0.6151859283447265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,1,power_law_1.2,0.6848531341552734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,1,power_law_1.2,0.8403961944580078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,1,power_law_1.2,1.0377875518798827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,1,power_law_1.2,1.4095123291015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,1,power_law_1.2,1.7544723510742188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,1,power_law_1.2,2.4993043518066407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,1,power_law_1.2,0.04207935810089111
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,1,power_law_1.2,0.060176000595092774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,1,power_law_1.2,0.0929094409942627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,1,power_law_1.2,0.11218111991882325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,1,power_law_1.2,0.15126015663146972
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,1,power_law_1.2,0.22114303588867185
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,1,power_law_1.2,0.23535295486450197
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,1,power_law_1.2,0.24160383224487303
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,1,power_law_1.2,0.24680639266967774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,1,power_law_1.2,0.247524471282959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,1,power_law_1.2,0.2535219192504883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,1,power_law_1.2,0.2880064010620117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,1,power_law_1.2,3.171478271484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,1,power_law_1.2,0.2903750419616699
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,1,power_law_1.2,0.3079840087890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,2,1,power_law_1.01,1.1335033416748046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,1,power_law_1.2,0.30874752044677733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,1,power_law_1.2,0.3143833541870117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,1,power_law_1.2,0.3235129547119141
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,2,1,power_law_1.01,1.5079513549804688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,1,power_law_1.2,0.3480537414550781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,1,power_law_1.2,0.3736640167236328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,1,power_law_1.2,0.4382131195068359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,1,power_law_1.2,0.5037120056152344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,1,power_law_1.2,0.6212185668945313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,2,1,power_law_1.01,2.249468231201172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,1,power_law_1.2,0.75334716796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,2,1,power_law_1.01,3.0187564086914063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,2,1,power_law_1.2,0.04024256229400634
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,2,1,power_law_1.2,0.053567361831665036
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,2,1,power_law_1.2,0.06346111774444581
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,2,1,power_law_1.2,0.07569344043731689
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,2,1,power_law_1.2,0.08976127624511718
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,2,1,power_law_1.2,0.14403200149536133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,1,power_law_1.2,1.0214419555664063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,2,1,power_law_1.2,0.1484806442260742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,2,1,power_law_1.2,0.13533184051513672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,2,1,power_law_1.2,0.15365247726440429
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,2,1,power_law_1.2,0.15383423805236815
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,2,1,power_law_1.2,0.15897664070129394
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,2,1,power_law_1.2,0.16661567687988282
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,2,1,power_law_1.2,0.17111488342285158
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,2,1,power_law_1.2,0.19380352020263672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,2,1,power_law_1.2,0.20096960067749023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,1,power_law_1.2,1.2853138732910157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,1,power_law_1.2,4.664618835449219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,2,1,power_law_1.2,0.18305471420288086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,2,1,power_law_1.2,0.2252889633178711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,2,1,power_law_1.2,0.2728166389465332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,1,power_law_1.2,1.8203514099121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,2,1,power_law_1.2,0.32609535217285157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,2,1,power_law_1.2,0.43645057678222654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,1,power_law_1.2,6.20509521484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,2,1,power_law_1.2,0.553138542175293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,1,2,balanced,0.09560832023620605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,1,2,balanced,0.147007999420166
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,2,1,power_law_1.2,0.7942867279052734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,1,2,balanced,0.3151308822631836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,1,2,balanced,0.5208217620849609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,1,2,balanced,0.999658203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,2,1,power_law_1.2,1.0400051116943358
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,1,2,balanced,1.2588365173339844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,1,2,balanced,1.277053451538086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,1,2,balanced,1.2645362854003905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,1,2,balanced,1.278924789428711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,1,2,balanced,1.3886553955078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,1,2,balanced,1.3864128112792968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,1,2,balanced,1.4056973266601562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,1,power_law_1.2,2.3589984130859376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,1,2,balanced,1.428423614501953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,1,2,balanced,1.4257971191406251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,1,2,balanced,1.4570828247070313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,1,2,balanced,1.4506982421874999
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,1,2,balanced,1.5018406677246094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,1,2,balanced,1.5578355407714843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,1,2,balanced,1.6336090087890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,2,1,power_law_1.2,1.5405754089355468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,1,2,balanced,2.614569091796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,1,2,balanced,2.6225765991210936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,1,2,balanced,3.7529138183593753
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,1,2,balanced,4.689144287109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,1,2,balanced,0.06454400062561036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,1,2,balanced,0.10469375610351564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,1,2,balanced,0.16261375427246094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,1,2,balanced,0.28272256851196287
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,2,1,power_law_1.2,2.0607513427734374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,1,2,balanced,0.5100134277343751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,1,2,balanced,6.65889892578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,1,2,balanced,0.6309503936767579
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,1,2,balanced,0.6342015838623046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,1,2,balanced,0.6366207885742188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,1,2,balanced,0.6356159973144531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,1,2,balanced,0.6821376037597656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,1,2,balanced,0.7204338836669921
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,1,2,balanced,0.6940787506103516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,1,2,balanced,0.7036672210693359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,1,2,balanced,0.7027699279785156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,1,2,balanced,9.560745239257812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,1,2,balanced,0.7062348937988281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,1,2,balanced,0.7173363494873046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,1,2,balanced,0.7193881225585937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,1,2,balanced,0.7653836822509765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,1,2,balanced,0.826055679321289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,1,2,balanced,0.8429491424560547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,1,2,balanced,0.9629119873046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,1,2,balanced,1.4670169067382812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,1,2,balanced,1.604259796142578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,1,2,balanced,2.2093606567382813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,1,2,balanced,2.866766052246094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,1,2,balanced,14.2444873046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,1,2,balanced,3.734119567871094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,1,2,balanced,4.895409851074218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,1,2,balanced,19.991716308593748
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,2,1,power_law_1.2,3.1182272338867185
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,1,2,balanced,0.04870272159576416
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,1,2,balanced,0.06479360103607178
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,1,2,balanced,0.37532032012939454
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,1,2,balanced,0.375280647277832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,1,2,balanced,0.0977894401550293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,1,2,balanced,0.3800998306274414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,1,2,balanced,0.37936256408691404
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,1,2,balanced,0.38070785522460937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,1,2,balanced,0.3764966583251953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,1,2,balanced,0.3861734390258789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,1,2,balanced,0.16396160125732423
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,1,2,balanced,0.38811775207519533
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,1,2,balanced,0.3877811050415039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,1,2,balanced,0.39605632781982425
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,1,2,balanced,0.4043008041381836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,1,2,balanced,0.414202880859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,1,2,balanced,0.43444351196289066
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,1,2,balanced,0.45671806335449217
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,1,2,balanced,0.3054540824890137
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,1,2,balanced,0.3739763259887695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,1,2,balanced,0.37407104492187504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,1,2,power_law_1.01,0.14477824211120605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,1,2,power_law_1.01,0.3174054336547852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,1,2,power_law_1.01,0.369246711730957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,1,2,power_law_1.01,0.5589286422729491
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,1,2,power_law_1.01,0.6919398498535156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,1,2,power_law_1.01,1.0689100646972656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,1,2,power_law_1.01,1.1065599822998047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,2,1,power_law_1.2,4.173086853027344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,1,2,power_law_1.01,1.1477196502685547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,1,2,power_law_1.01,1.1467660522460936
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,1,2,balanced,0.6768089294433594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,1,2,power_law_1.01,1.2692006683349608
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,1,2,balanced,0.734682846069336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,1,2,balanced,1.0796173095703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,1,2,power_law_1.01,1.2830041503906249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,1,2,power_law_1.01,1.3304447937011719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,1,2,balanced,1.4292927551269532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,1,2,power_law_1.01,1.3508172607421876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,1,2,power_law_1.01,1.3762240600585938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,1,2,balanced,1.8665766906738281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,1,2,power_law_1.01,1.4751449584960938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,1,2,power_law_1.01,1.5306623840332032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,1,2,power_law_1.01,1.5838783264160157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,1,2,power_law_1.01,1.8029542541503907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,1,2,power_law_1.01,2.1034764099121093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,1,2,power_law_1.01,2.5268902587890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,1,2,power_law_1.01,3.070082702636719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,1,2,power_law_1.01,0.10674943923950195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,1,2,power_law_1.01,4.107702941894532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,1,2,power_law_1.01,0.14765824317932127
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,1,2,balanced,2.6301336669921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,1,2,power_law_1.01,0.19904767990112304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,1,2,power_law_1.01,0.26633983612060547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,1,2,power_law_1.01,0.5059622573852539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,1,2,power_law_1.01,0.5290636825561523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,1,2,power_law_1.01,5.190293579101562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,1,2,power_law_1.01,0.5611187362670899
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,1,2,power_law_1.01,0.368276481628418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,1,2,power_law_1.01,0.624711685180664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,1,2,power_law_1.01,0.6279423904418946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,1,2,power_law_1.01,0.642529296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,1,2,power_law_1.01,0.6642047882080078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,1,2,power_law_1.01,7.32685302734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,1,2,power_law_1.01,0.6817330932617187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,1,2,power_law_1.01,0.7080998229980469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,1,2,power_law_1.01,0.7102963256835937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,1,2,power_law_1.01,0.7761510467529297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,1,2,power_law_1.01,0.8465612792968751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,1,2,power_law_1.01,8.834146728515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,1,2,power_law_1.01,0.9798041534423827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,1,2,power_law_1.01,1.1018022155761718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,1,2,power_law_1.01,1.423205108642578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,1,2,power_law_1.01,1.7444429016113283
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,1,2,power_law_1.01,2.267710723876953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,1,2,power_law_1.01,0.5483750534057618
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,1,2,power_law_1.01,0.07367936134338379
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,1,2,power_law_1.01,0.11639936447143555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,1,2,power_law_1.01,2.9668389892578126
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,1,2,power_law_1.01,0.16579200744628905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,1,2,power_law_1.01,13.034718017578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,1,2,power_law_1.01,0.30112127304077146
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,1,2,power_law_1.01,0.3311564636230469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,1,2,power_law_1.01,0.3109542465209961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,1,2,power_law_1.01,4.110176086425781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,1,2,power_law_1.01,0.32275840759277347
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,1,2,power_law_1.01,0.20618879318237307
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,1,2,power_law_1.01,0.3277350234985351
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,1,2,power_law_1.01,0.3488025665283203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,1,2,power_law_1.01,0.3474406433105469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,1,2,power_law_1.01,0.3481996917724609
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,1,2,power_law_1.01,0.3778099060058594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,1,2,power_law_1.01,0.6689907073974609
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,1,2,power_law_1.01,0.38577022552490237
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,1,2,power_law_1.01,0.39322624206542967
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,1,2,power_law_1.01,0.5113267135620118
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,1,2,power_law_1.01,5.436279296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,1,2,power_law_1.01,18.062177734375002
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,1,2,power_law_1.01,0.09474944114685059
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,1,2,power_law_1.01,0.40689151763916015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,1,2,power_law_1.01,0.45228416442871094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,1,2,power_law_1.01,0.6119948959350585
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,1,2,power_law_1.01,0.7064230346679687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,1,2,power_law_1.2,0.14391807556152342
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,1,2,power_law_1.2,0.23413759231567383
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,1,2,power_law_1.01,0.9168447875976563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,1,2,power_law_1.2,0.3840857696533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,1,2,power_law_1.01,1.157770233154297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,1,2,power_law_1.2,0.5181862258911133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,1,2,power_law_1.2,0.5874560165405274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,1,2,power_law_1.2,1.027915496826172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,1,2,power_law_1.2,1.1228787231445314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,1,2,power_law_1.2,1.2650252532958983
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,1,2,power_law_1.01,1.9568882751464844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,1,2,power_law_1.2,1.3195890808105468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,1,2,power_law_1.2,1.311550750732422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,1,2,power_law_1.2,1.0912934112548829
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,1,2,power_law_1.2,1.3853375244140627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,1,2,power_law_1.2,1.1373632049560547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,1,2,power_law_1.2,1.4284594726562498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,1,2,power_law_1.2,1.4823680114746094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,1,2,power_law_1.2,1.5172557067871093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,1,2,power_law_1.2,1.6779046630859376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,1,2,power_law_1.2,1.8584013366699217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,1,2,power_law_1.2,2.1186956787109374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,1,2,power_law_1.2,2.6759988403320314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,1,2,power_law_1.2,3.22727294921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,1,2,power_law_1.2,4.354151611328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,1,2,power_law_1.2,5.513328857421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,1,2,power_law_1.2,0.10674688339233399
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,1,2,power_law_1.2,0.12698623657226563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,1,2,power_law_1.2,0.1969203186035156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,1,2,power_law_1.2,7.383450927734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,1,2,power_law_1.2,0.235731201171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,1,2,power_law_1.2,0.344714241027832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,1,2,power_law_1.2,0.5133491134643554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,1,2,power_law_1.2,0.5363737487792969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,1,2,power_law_1.2,0.558023681640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,1,2,power_law_1.2,0.5737395095825195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,1,2,power_law_1.2,0.6225126266479493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,1,2,power_law_1.2,9.133694458007813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,1,2,power_law_1.2,0.6488025665283204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,1,2,power_law_1.2,0.6492454528808593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,1,2,power_law_1.2,0.6830502319335937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,1,2,power_law_1.2,0.6779417419433594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,1,2,power_law_1.2,0.7059903717041015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,1,2,power_law_1.2,0.702455062866211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,1,2,power_law_1.2,0.8099584197998047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,1,2,power_law_1.2,0.8479424285888673
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,1,2,power_law_1.2,1.0076748657226564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,1,2,power_law_1.2,1.1379443359374999
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,1,2,power_law_1.01,2.2166105651855466
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,1,2,power_law_1.2,1.524586181640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,1,2,power_law_1.2,1.8342361450195312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,1,2,power_law_1.2,0.6682956695556641
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,1,2,power_law_1.01,2.9497229003906247
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,1,2,power_law_1.2,0.07031295776367188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,1,2,power_law_1.2,0.08072319984436035
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,1,2,power_law_1.2,0.11919487953186034
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,1,2,power_law_1.01,4.002744445800781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,1,2,power_law_1.2,0.14808704376220702
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,1,2,power_law_1.2,2.9652301025390626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,1,2,power_law_1.2,0.28709632873535157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,1,2,power_law_1.2,0.18902528762817383
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,1,2,power_law_1.2,0.31931007385253907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,1,2,power_law_1.2,2.3626010131835935
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,1,2,power_law_1.2,0.31657728195190427
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,1,2,power_law_1.2,0.32760704040527344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,1,2,power_law_1.2,0.3374425506591797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,1,2,power_law_1.2,0.3693606567382813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,1,2,power_law_1.2,0.35956993103027346
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,1,2,power_law_1.2,0.37107967376708983
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,1,2,power_law_1.2,0.3844224166870117
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,1,2,power_law_1.2,0.410693130493164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,1,2,power_law_1.2,0.44947456359863286
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,1,2,power_law_1.2,4.260110168457031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,1,2,power_law_1.2,13.9738037109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,1,2,power_law_1.2,19.076329345703122
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,1,2,power_law_1.2,5.653848266601562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,1,2,power_law_1.2,0.3884505462646485
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,1,2,power_law_1.2,0.5068313598632812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,1,2,power_law_1.2,0.5484620666503905
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,1,2,power_law_1.2,0.7365888214111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,2,1,balanced,0.10493375778198241
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,2,1,balanced,0.15876928329467774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,2,1,balanced,0.3260831832885742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,1,2,power_law_1.2,0.8518169403076172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,2,1,balanced,0.5186956787109376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,1,2,power_law_1.2,1.1267481231689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,2,1,balanced,1.0028787231445313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,2,1,balanced,1.2792428588867186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,2,1,balanced,1.2724813079833983
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,2,1,balanced,1.2893228149414062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,2,1,balanced,1.308023681640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,2,1,balanced,1.2873568725585938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,2,1,balanced,1.3185618591308592
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,2,1,balanced,1.3594950866699218
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,1,2,power_law_1.2,1.494430694580078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,2,1,balanced,1.423098907470703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,2,1,balanced,1.4322131347656248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,2,1,balanced,1.4437516784667968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,2,1,balanced,1.4270841979980469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,2,1,balanced,1.4857171630859374
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,1,2,power_law_1.2,2.3028326416015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,2,1,balanced,1.5453823852539064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,2,1,balanced,1.6368888854980468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,2,1,balanced,2.5311212158203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,2,1,balanced,2.5749850463867188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,2,1,balanced,3.618413391113281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,2,1,balanced,4.592040405273438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,2,1,balanced,0.07832255840301514
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,2,1,balanced,0.11627712249755859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,2,1,balanced,0.17938432693481446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,2,1,balanced,0.29953727722167967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,2,1,balanced,6.038240356445312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,2,1,balanced,0.5385696029663085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,2,1,balanced,0.678611831665039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,2,1,balanced,0.6668236541748047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,2,1,balanced,0.6758643341064453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,2,1,balanced,0.6738617706298828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,2,1,balanced,0.680505599975586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,2,1,balanced,8.063255615234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,2,1,balanced,0.6856012725830078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,2,1,balanced,0.694172134399414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,2,1,balanced,0.7621382141113282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,2,1,balanced,0.7757740783691406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,2,1,balanced,0.7819808197021485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,2,1,balanced,0.7933190155029297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,2,1,balanced,0.805317153930664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,2,1,balanced,0.8478099060058593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,2,1,balanced,0.8499302673339845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,2,1,balanced,0.919021453857422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,2,1,balanced,11.297901611328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,2,1,balanced,0.9722284698486329
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,1,2,power_law_1.2,2.64548095703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,2,1,balanced,1.4972268676757812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,2,1,balanced,1.6008370971679688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,2,1,balanced,2.300576629638672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,1,2,power_law_1.2,4.072300720214844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,2,1,balanced,2.9094451904296874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,2,1,balanced,3.8088729858398436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,2,1,balanced,15.47906982421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,2,1,balanced,5.071888732910156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,1,2,power_law_1.2,5.249784545898438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,2,1,balanced,0.05757503986358643
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,2,1,balanced,0.07306047916412353
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,2,1,balanced,0.39866943359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,2,1,balanced,0.3998451232910156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,2,1,balanced,0.10670911788940429
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,2,1,balanced,0.4020991897583007
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,2,1,balanced,0.4038528060913086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,2,1,balanced,0.41143424987792965
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,2,1,balanced,0.40514560699462887
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,2,1,balanced,0.17522687911987306
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,2,1,balanced,0.3233561706542969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,2,1,balanced,0.4168864059448242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,2,1,balanced,0.41303424835205077
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,2,1,balanced,0.42461566925048827
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,2,1,balanced,0.39768062591552733
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,2,1,balanced,0.44180862426757816
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,2,1,balanced,0.45373374938964844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,2,1,balanced,0.3973286437988281
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,2,1,balanced,0.48431617736816407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,2,1,balanced,0.5123795318603516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,2,1,balanced,0.3976031875610352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,2,1,power_law_1.01,0.10585663795471192
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,2,1,power_law_1.01,0.15869888305664062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,2,1,power_law_1.01,0.32464065551757815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,2,1,power_law_1.01,0.4737158584594726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,2,1,power_law_1.01,0.631641616821289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,2,1,power_law_1.01,0.9867865753173828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,2,1,power_law_1.01,1.0502098846435548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,2,1,power_law_1.01,1.087923812866211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,2,1,power_law_1.01,1.1354003143310547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,2,1,power_law_1.01,1.1731533050537108
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,2,1,power_law_1.01,1.3158668518066405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,2,1,power_law_1.01,1.3900697326660156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,2,1,power_law_1.01,1.3401081848144532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,2,1,power_law_1.01,1.3830015563964844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,2,1,power_law_1.01,1.424833984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,2,1,power_law_1.01,1.4494444274902345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,2,1,power_law_1.01,1.5494969177246094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,2,1,power_law_1.01,1.77180419921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,2,1,power_law_1.01,1.9692153930664062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,2,1,power_law_1.01,2.3634655761718752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,2,1,power_law_1.01,2.841034240722656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,2,1,power_law_1.01,3.613970031738281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,2,1,power_law_1.01,4.514497985839844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,2,1,balanced,0.8329625701904296
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,2,1,balanced,0.900967025756836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,2,1,balanced,1.2620435333251954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,2,1,power_law_1.01,0.07729983806610108
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,2,1,balanced,1.649357452392578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,2,1,power_law_1.01,0.11515839576721192
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,2,1,power_law_1.01,0.1812473678588867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,2,1,balanced,2.2380589294433593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,2,1,power_law_1.01,0.25610559463500976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,2,1,power_law_1.01,0.3505318450927734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,2,1,balanced,3.011205749511719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,2,1,power_law_1.01,0.5357478332519532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,2,1,power_law_1.01,0.5882252883911133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,2,1,power_law_1.01,0.6042521667480469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,2,1,power_law_1.01,6.221103515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,2,1,power_law_1.01,0.6341228866577149
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,2,1,power_law_1.01,0.5666361618041992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,2,1,power_law_1.01,0.6392531204223633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,2,1,power_law_1.01,0.6097670364379882
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,2,1,power_law_1.01,0.7155430603027344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,2,1,power_law_1.01,0.7384825897216797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,2,1,power_law_1.01,0.7374406433105468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,2,1,power_law_1.01,0.7409388732910156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,2,1,power_law_1.01,0.7642675018310546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,2,1,power_law_1.01,0.821048355102539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,2,1,power_law_1.01,0.8920390319824218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,2,1,power_law_1.01,0.993658905029297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,2,1,power_law_1.01,1.1483558654785155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,2,1,power_law_1.01,1.4175814819335937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,2,1,power_law_1.01,7.991239624023438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,2,1,power_law_1.01,1.7137049865722656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,2,1,power_law_1.01,2.2865087890624998
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,2,1,power_law_1.01,0.06484992027282714
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,2,1,power_law_1.01,0.08111680030822754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,2,1,power_law_1.01,2.85763916015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,2,1,power_law_1.01,0.10913215637207033
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,2,1,power_law_1.01,0.15448512077331542
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,2,1,power_law_1.01,0.20165760040283204
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,2,1,power_law_1.01,0.3330131149291992
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,2,1,power_law_1.01,0.31325056076049806
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,2,1,power_law_1.01,0.3466969680786133
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,2,1,power_law_1.01,0.35858558654785155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,2,1,power_law_1.01,3.903152770996094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,2,1,power_law_1.01,11.44615478515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,2,1,power_law_1.01,0.3645260620117187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,2,1,power_law_1.01,0.3764031982421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,2,1,power_law_1.01,0.38703102111816406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,2,1,power_law_1.01,0.3915411376953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,2,1,power_law_1.01,0.41268543243408207
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,2,1,power_law_1.01,0.4234265518188477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,2,1,power_law_1.01,5.067239074707031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,2,1,power_law_1.01,0.4278694534301758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,2,1,power_law_1.01,14.816036376953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,2,1,power_law_1.01,0.46141311645507815
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,2,1,power_law_1.01,0.5248320007324219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,2,1,power_law_1.01,0.7032633972167969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,2,1,power_law_1.01,0.5798284912109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,2,1,power_law_1.01,0.8567475128173829
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,2,1,power_law_1.01,1.1333074951171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,2,1,power_law_1.2,0.15875136375427246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,2,1,power_law_1.2,0.3170265579223633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,2,1,power_law_1.2,0.42519359588623046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,2,1,power_law_1.2,0.5659263992309571
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,2,1,power_law_1.2,0.9804006195068359
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,2,1,power_law_1.01,1.4242098999023436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,2,1,power_law_1.2,1.0386681365966797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,2,1,power_law_1.01,2.0332019042968748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,2,1,power_law_1.2,0.10400192260742187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,2,1,power_law_1.2,1.1138182067871094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,2,1,power_law_1.2,1.1605741119384765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,2,1,power_law_1.2,1.2259117126464845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,2,1,power_law_1.2,1.4047103881835938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,2,1,power_law_1.2,1.386742401123047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,2,1,power_law_1.2,1.4107519531250001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,2,1,power_law_1.2,1.2911117553710938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,2,1,power_law_1.2,1.4937088012695312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,2,1,power_law_1.2,1.5023269653320312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,2,1,power_law_1.2,1.6065478515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,2,1,power_law_1.2,1.8258023071289062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,2,1,power_law_1.2,1.966374969482422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,2,1,power_law_1.2,2.478435821533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,2,1,power_law_1.2,2.848327026367188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,2,1,power_law_1.2,3.74956787109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,2,1,power_law_1.2,4.616448059082031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,2,1,power_law_1.2,0.08041343688964844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,2,1,power_law_1.2,0.11671999931335449
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,2,1,power_law_1.2,0.17313407897949218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,2,1,power_law_1.2,0.2384806442260742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,2,1,power_law_1.2,6.278627319335937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,2,1,power_law_1.2,0.32708736419677736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,2,1,power_law_1.2,0.5329478454589844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,2,1,power_law_1.2,0.5587865447998046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,2,1,power_law_1.2,0.5901772689819336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,2,1,power_law_1.2,0.6102323150634765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,2,1,power_law_1.2,0.6260358428955077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,2,1,power_law_1.2,8.110819702148437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,2,1,power_law_1.2,0.6410355377197265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,2,1,power_law_1.2,0.6549747467041016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,2,1,power_law_1.2,0.7494169616699219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,2,1,power_law_1.2,0.7447001647949218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,2,1,power_law_1.2,0.7620038604736328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,2,1,power_law_1.2,0.7935558319091797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,2,1,power_law_1.2,0.7955987548828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,2,1,power_law_1.2,0.8409343719482422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,2,1,power_law_1.2,0.8945913696289063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,2,1,power_law_1.2,1.0547398376464845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,2,1,power_law_1.2,1.1768876647949218
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,2,1,power_law_1.01,2.6603744506835936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,2,1,power_law_1.2,1.5055731201171876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,2,1,power_law_1.2,1.7502847290039063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,2,1,power_law_1.01,3.9100094604492184
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,2,1,power_law_1.2,2.286456298828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,2,1,power_law_1.2,0.06582848072052003
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,2,1,power_law_1.2,11.345270996093749
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,2,1,power_law_1.01,5.21806640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,2,1,power_law_1.2,0.07796031951904296
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,2,1,power_law_1.2,0.10660032272338867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,2,1,power_law_1.2,2.8122906494140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,2,1,power_law_1.2,0.14173503875732424
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,2,1,power_law_1.2,0.31076416015624997
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,2,1,power_law_1.2,0.18844287872314452
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,2,1,power_law_1.2,0.33185726165771484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,2,1,power_law_1.2,0.3513625717163086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,2,1,power_law_1.2,0.36325889587402344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,2,1,power_law_1.2,0.3755257415771484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,2,1,power_law_1.2,0.38891712188720706
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,2,1,power_law_1.2,0.3971775817871094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,2,1,power_law_1.2,0.409815673828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,2,1,power_law_1.2,0.4239174270629883
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,2,1,power_law_1.2,0.44713535308837893
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,2,1,power_law_1.2,0.501228141784668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,2,1,power_law_1.2,3.919146728515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,2,1,power_law_1.2,14.977630615234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,2,1,power_law_1.2,5.054707946777343
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,2,1,power_law_1.2,0.4655852890014648
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,2,1,power_law_1.2,0.5787987136840821
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,2,1,power_law_1.2,0.6585279846191406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,2,1,power_law_1.2,0.8485088348388672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,2,1,power_law_1.2,1.0530643463134766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,2,balanced,0.09643391609191895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,2,balanced,0.1469324779510498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,2,balanced,0.29600383758544924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,2,balanced,0.5224256134033203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,2,balanced,0.9471820831298828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,2,balanced,1.8914073181152342
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,2,1,power_law_1.2,1.4895610046386718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,2,balanced,2.8180096435546877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,2,balanced,2.8694107055664064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,2,balanced,2.8376782226562502
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,2,balanced,2.9031668090820313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,2,balanced,2.9179174804687498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,2,balanced,2.970415344238281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,2,balanced,2.9249700927734374
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,2,1,power_law_1.2,1.9496794128417967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,2,balanced,3.069135437011719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,2,balanced,3.081553955078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,2,balanced,3.055566101074219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,2,balanced,3.09454833984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,2,balanced,3.1804376220703126
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,2,1,power_law_1.2,2.9379339599609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,2,balanced,3.2513702392578123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,2,balanced,3.394940185546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,2,balanced,3.567906494140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,2,balanced,3.8573989868164062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,2,balanced,5.807893676757812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,2,balanced,6.244625854492187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,2,balanced,0.06372223854064941
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,2,balanced,0.14927743911743163
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,2,balanced,0.08716544151306152
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,2,balanced,0.2706841659545899
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,2,balanced,0.500682258605957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,2,balanced,0.9410688018798827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,2,balanced,1.4006092834472657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,2,balanced,9.128952026367188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,2,balanced,1.3965414428710938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,2,balanced,1.3920870971679686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,2,balanced,1.4095513916015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,2,1,power_law_1.2,3.910331420898437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,2,balanced,1.4136140441894531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,2,balanced,1.421624298095703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,2,balanced,1.4291366577148437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,2,balanced,1.515924530029297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,2,balanced,1.5589479064941405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,2,balanced,13.319353027343748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,2,balanced,1.5660581970214844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,2,balanced,1.5887667846679687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,2,balanced,1.6041510009765623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,2,balanced,1.6661888122558595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,2,balanced,1.738464050292969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,2,balanced,1.8251596069335938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,2,balanced,1.960966339111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,2,balanced,2.151585235595703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,2,balanced,2.412602844238281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,2,balanced,3.441509094238281
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,2,1,power_law_1.2,5.8950476074218745
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,2,balanced,3.9435366821289066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,2,balanced,19.985057373046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,2,balanced,5.292034301757813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,2,1,power_law_1.2,7.958756103515626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,1,2,balanced,0.05871103763580322
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,1,2,balanced,0.07541759967803954
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,1,2,balanced,0.843737564086914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,1,2,balanced,0.8434214019775391
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,1,2,balanced,0.8536268615722656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,1,2,balanced,0.1047987174987793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,1,2,balanced,0.16573312759399414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,1,2,balanced,0.8489766693115234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,1,2,balanced,0.2955801582336426
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,1,2,balanced,0.857149429321289
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,1,2,balanced,0.8479232025146484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,1,2,balanced,0.5663948822021484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,1,2,balanced,0.8551731109619141
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,1,2,balanced,0.861185302734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,1,2,balanced,0.875551986694336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,1,2,balanced,0.8785549163818359
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,1,2,balanced,0.8928345489501954
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,1,2,balanced,0.9197721862792969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,1,2,balanced,0.9311027526855469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,1,2,balanced,0.8375654602050782
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,1,2,balanced,0.9734015655517577
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,1,2,balanced,1.0193267059326172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,1,2,balanced,1.1219058990478517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,2,power_law_1.01,0.14776063919067384
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,2,power_law_1.01,0.29804800033569334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,2,power_law_1.01,0.5188377761840821
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,2,power_law_1.01,0.6614924621582031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,2,power_law_1.01,0.987347183227539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,2,power_law_1.01,1.36050048828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,2,power_law_1.01,1.4496231079101562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,2,power_law_1.01,1.9819340515136719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,2,power_law_1.01,2.0519564819335936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,2,power_law_1.01,2.098859558105469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,2,power_law_1.01,2.0832485961914062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,2,power_law_1.01,2.4400767517089843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,2,power_law_1.01,2.5487577819824216
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,2,power_law_1.01,2.5958682250976564
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,1,2,balanced,0.8391308593750001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,2,power_law_1.01,2.116732177734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,2,power_law_1.01,2.6402880859374998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,2,power_law_1.01,2.6805902099609376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,2,power_law_1.01,2.8228466796875002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,2,power_law_1.01,3.130415344238281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,2,power_law_1.01,3.5968423461914063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,2,power_law_1.01,4.033917541503906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,2,power_law_1.01,4.911499633789062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,2,power_law_1.01,6.096400756835938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,2,power_law_1.01,0.08641152381896973
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,2,power_law_1.01,0.14753664016723633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,2,power_law_1.01,7.446196289062501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,2,power_law_1.01,0.2609868812561035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,2,power_law_1.01,0.3392230224609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,2,power_law_1.01,0.5240678405761718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,2,power_law_1.01,0.7084941101074219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,2,power_law_1.01,9.562552490234376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,2,power_law_1.01,0.7905152130126953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,2,power_law_1.01,0.9943756866455079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,2,power_law_1.01,1.0739981079101564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,2,power_law_1.01,1.092010269165039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,2,power_law_1.01,12.953400878906251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,2,power_law_1.01,1.1479385375976563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,2,power_law_1.01,1.1263475036621093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,2,power_law_1.01,1.2724198150634767
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,2,power_law_1.01,1.2975578308105469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,2,power_law_1.01,1.271115493774414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,1,2,balanced,1.736747589111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,2,power_law_1.01,1.3279257202148438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,1,2,balanced,1.9299533081054687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,2,power_law_1.01,1.4063487243652344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,2,power_law_1.01,1.5066610717773439
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,1,2,balanced,2.6221325683593752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,2,power_law_1.01,16.563717041015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,2,power_law_1.01,1.0392012786865235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,2,power_law_1.01,1.6309785461425783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,2,power_law_1.01,1.7464997863769534
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,1,2,power_law_1.01,0.08970239639282226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,2,power_law_1.01,1.9571315002441405
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,1,2,power_law_1.01,0.12057600021362305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,1,2,power_law_1.01,0.17948287963867188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,2,power_law_1.01,2.300537567138672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,1,2,power_law_1.01,0.21271551132202146
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,1,2,power_law_1.01,0.32625919342041015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,1,2,power_law_1.01,0.44761726379394534
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,2,power_law_1.01,2.8952691650390627
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,1,2,power_law_1.01,0.469306869506836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,1,2,power_law_1.01,0.5827161788940429
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,1,2,power_law_1.01,0.603509750366211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,2,power_law_1.01,3.4855422973632812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,1,2,power_law_1.01,0.6140979385375976
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,1,2,power_law_1.01,0.6460415649414062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,1,2,power_law_1.01,0.6425804901123047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,1,2,power_law_1.01,0.7000908660888672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,1,2,power_law_1.01,0.7155993652343751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,2,power_law_1.01,4.43799560546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,1,2,power_law_1.01,0.7416729736328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,1,2,power_law_1.01,0.7512486267089844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,1,2,power_law_1.01,0.7686566162109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,1,2,power_law_1.01,0.8529548645019531
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,1,2,power_law_1.01,0.8822156524658202
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,2,power_law_1.01,5.786732788085938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,1,2,power_law_1.01,0.931415023803711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,1,2,power_law_1.01,1.0358643341064453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,2,power_law_1.2,0.14767871856689455
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,1,2,power_law_1.01,1.3145945739746092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,2,power_law_1.2,0.29691904067993163
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,1,2,power_law_1.01,1.4491123962402344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,2,power_law_1.2,0.4057740783691406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,2,power_law_1.2,0.6736422729492187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,1,2,power_law_1.01,1.9754290771484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,1,2,power_law_1.01,2.3226739501953126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,2,power_law_1.2,0.931825942993164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,2,power_law_1.2,1.2949964904785156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,2,power_law_1.2,1.3552000427246094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,1,2,power_law_1.01,3.2085940551757814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,2,power_law_1.2,1.9507020568847657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,2,power_law_1.2,2.0238195800781247
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,2,power_law_1.2,2.144595184326172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,1,2,power_law_1.01,4.2186892700195315
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,2,power_law_1.2,2.26184326171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,2,power_law_1.2,2.351152648925781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,2,power_law_1.2,2.4657728576660154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,2,power_law_1.2,1.9029171752929688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,2,power_law_1.2,2.5367141723632813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,2,power_law_1.2,2.652998352050781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,2,power_law_1.2,2.823157653808594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,2,power_law_1.2,2.920929260253906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,2,power_law_1.2,3.1154995727539063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,2,power_law_1.2,3.7269708251953126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,2,power_law_1.2,0.08576640129089355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,2,power_law_1.2,4.314202880859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,2,power_law_1.2,0.14766847610473632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,2,power_law_1.2,0.22060287475585935
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,2,power_law_1.2,0.32619647979736327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,2,power_law_1.2,5.219015502929688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,2,power_law_1.2,0.4969728088378906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,2,power_law_1.2,0.6215001678466796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,2,power_law_1.2,6.270086669921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,2,power_law_1.2,0.7270835113525391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,2,power_law_1.2,1.0125376129150392
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,2,power_law_1.2,7.700038452148438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,2,power_law_1.2,1.015594253540039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,2,power_law_1.2,1.0350220489501953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,2,power_law_1.2,1.0818470764160157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,2,power_law_1.2,1.0451801300048829
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,2,power_law_1.2,1.1205951690673828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,2,power_law_1.2,1.269612808227539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,2,power_law_1.2,1.289608917236328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,2,power_law_1.2,10.119354858398438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,2,power_law_1.2,1.3342437744140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,2,power_law_1.2,1.2566719818115235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,2,power_law_1.2,1.4384780883789063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,2,power_law_1.2,1.5269287109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,2,power_law_1.2,1.673017578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,2,power_law_1.2,1.857987823486328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,2,power_law_1.2,2.1594995117187503
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,2,power_law_1.2,2.482704620361328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,2,power_law_1.2,13.453369140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,1,2,power_law_1.2,0.08791040420532227
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,1,2,power_law_1.2,0.11694975852966309
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,2,power_law_1.2,3.0284646606445316
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,1,2,power_law_1.2,0.15210623741149903
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,1,2,power_law_1.2,0.21984256744384764
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,1,2,power_law_1.2,0.3008793640136719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,1,2,power_law_1.2,0.364158706665039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,1,2,power_law_1.2,0.4429401779174805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,2,power_law_1.2,3.711192321777344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,1,2,power_law_1.2,0.5793305587768555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,1,2,power_law_1.2,0.5881868743896485
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,1,2,power_law_1.2,0.626539535522461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,1,2,power_law_1.2,0.6292083358764649
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,1,2,power_law_1.2,0.6490009307861329
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,1,2,power_law_1.2,0.6850624084472656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,1,2,power_law_1.2,0.6998758697509766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,1,2,power_law_1.2,0.7289715576171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,2,power_law_1.2,4.792797546386718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,2,power_law_1.2,17.0120849609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,1,2,power_law_1.2,0.7620159912109374
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,1,2,power_law_1.2,0.802881317138672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,1,2,power_law_1.2,0.8810816192626953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,2,power_law_1.2,6.108272705078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,1,2,power_law_1.2,0.9454962921142578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,1,2,power_law_1.2,1.0132838439941405
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,1,2,power_law_1.2,1.1967526245117188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,1,balanced,0.11229248046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,1,balanced,0.15422847747802734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,1,2,power_law_1.2,1.451646728515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,1,2,power_law_1.2,1.7743641662597658
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,1,balanced,0.3071225547790527
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,1,balanced,0.5335667037963867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,1,balanced,0.9679673767089844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,1,2,power_law_1.2,2.388716735839844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,1,2,power_law_1.2,3.2417535400390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,1,balanced,1.8942105102539064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,1,balanced,2.8634732055664065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,1,balanced,2.8952313232421876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,1,2,power_law_1.2,4.3950567626953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,1,balanced,2.953238525390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,1,balanced,2.886555480957031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,1,balanced,2.9556192016601566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,1,balanced,2.893987731933594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,1,balanced,2.9779870605468752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,1,balanced,3.0001715087890624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,1,balanced,2.9383917236328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,1,balanced,3.054859619140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,1,balanced,3.13214599609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,1,balanced,3.177757568359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,1,balanced,3.2625778198242186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,1,balanced,3.3819699096679683
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,1,balanced,3.540204772949219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,1,balanced,3.8635424804687504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,1,balanced,0.08578432083129883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,1,balanced,0.10427007675170899
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,1,balanced,0.16327743530273436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,1,balanced,0.28682111740112304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,1,balanced,0.5131257629394531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,1,balanced,5.763173828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,1,balanced,0.9979058837890624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,1,balanced,6.163255615234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,1,balanced,1.503190460205078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,1,balanced,1.5190296936035157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,1,balanced,1.5121017456054688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,1,balanced,1.5259327697753906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,1,balanced,1.5314515686035155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,1,balanced,8.492389526367187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,1,balanced,1.5342239379882812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,1,balanced,1.5516741943359373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,1,balanced,1.5515660095214845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,1,balanced,1.56718017578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,1,balanced,1.5587423706054688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,1,balanced,1.779651794433594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,1,balanced,1.7943968200683593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,1,balanced,11.38852294921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,1,balanced,1.8354086303710937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,1,balanced,1.8914207458496093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,1,balanced,1.9172262573242187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,1,balanced,2.096708526611328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,1,balanced,2.1914770507812498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,1,balanced,2.5135821533203124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,1,balanced,3.5941976928710937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,1,balanced,4.172294921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,1,2,power_law_1.2,5.962952880859374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,1,balanced,5.518220214843749
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,1,balanced,17.911204833984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,2,1,balanced,0.0895206356048584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,2,1,balanced,0.08899328231811524
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,2,1,balanced,0.09113663673400879
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,2,1,balanced,0.1478726387023926
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,2,1,balanced,0.159617919921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,2,1,balanced,0.09478528022766113
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,2,1,balanced,0.18339456558227538
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,2,1,balanced,0.10017791748046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,2,1,balanced,0.12400832176208496
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,2,1,balanced,0.11157440185546874
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,2,1,balanced,0.13662400245666503
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,2,1,balanced,0.21893951416015628
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,2,1,balanced,0.24240831375122068
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,2,1,balanced,0.29139520645141603
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,2,1,balanced,0.35307136535644534
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,2,1,balanced,0.4021036911010742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,2,1,balanced,0.5118239974975586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,2,1,balanced,0.7395750427246094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,2,1,balanced,0.9614937591552735
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,2,1,balanced,1.396942138671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,1,power_law_1.01,0.10896767616271971
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,2,1,balanced,1.8342880249023437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,2,1,balanced,2.7208569335937502
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,1,power_law_1.01,0.15085951805114745
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,1,power_law_1.01,0.30952768325805663
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,1,power_law_1.01,0.5327737426757813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,1,power_law_1.01,0.9009907531738282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,1,power_law_1.01,1.2676927947998047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,2,1,balanced,3.5912274169921874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,1,power_law_1.01,1.4615744018554688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,1,power_law_1.01,1.9160768127441408
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,2,1,balanced,5.378711547851562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,1,power_law_1.01,2.0025567626953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,1,power_law_1.01,2.0809837341308595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,1,power_law_1.01,2.1750944519042967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,1,power_law_1.01,2.254451904296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,1,power_law_1.01,2.330243835449219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,1,power_law_1.01,2.547962951660156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,1,power_law_1.01,2.7396807861328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,1,power_law_1.01,2.9511621093750002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,1,power_law_1.01,2.6897268676757813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,1,power_law_1.01,2.8754144287109376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,1,power_law_1.01,3.0907745361328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,1,power_law_1.01,3.5221746826171874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,1,power_law_1.01,3.897885437011719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,1,power_law_1.01,4.788574829101562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,1,power_law_1.01,5.609237060546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,1,power_law_1.01,7.1770129394531255
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,1,power_law_1.01,0.07263167858123779
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,1,power_law_1.01,0.10166591644287108
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,1,power_law_1.01,8.841860351562499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,1,power_law_1.01,0.28501951217651367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,1,power_law_1.01,0.16326719284057617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,1,power_law_1.01,0.4865011215209961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,1,power_law_1.01,0.6869721221923828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,1,power_law_1.01,0.8150374603271484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,1,power_law_1.01,1.0252992248535155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,1,power_law_1.01,1.0796832275390624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,1,power_law_1.01,1.1176582336425782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,1,power_law_1.01,1.1553126525878905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,1,power_law_1.01,1.1692947387695312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,1,power_law_1.01,12.039366455078126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,1,power_law_1.01,1.2152601623535157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,1,power_law_1.01,1.2598604583740234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,1,power_law_1.01,1.2614015960693359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,1,power_law_1.01,1.3121292114257812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,1,power_law_1.01,1.4637919616699218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,1,power_law_1.01,1.5259992980957031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,1,power_law_1.01,1.584172821044922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,1,power_law_1.01,1.7265536499023437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,1,power_law_1.01,15.448818359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,1,power_law_1.01,1.8615065002441404
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,1,power_law_1.01,2.145635223388672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,1,power_law_1.01,2.3961203002929685
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,2,1,power_law_1.01,0.08894399642944337
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,1,power_law_1.01,3.5068255615234376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,1,power_law_1.01,2.9420947265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,2,1,power_law_1.01,0.08909952163696289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,1,power_law_1.01,4.558469848632813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,1,power_law_1.01,5.686192626953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,2,1,power_law_1.01,0.09075455665588379
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,2,1,power_law_1.01,0.09413248062133789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,2,1,power_law_1.01,0.09934911727905274
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,2,1,power_law_1.01,0.11212544441223145
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,2,1,power_law_1.01,0.1234438419342041
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,2,1,power_law_1.01,0.13624640464782717
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,2,1,power_law_1.01,0.18418176651000978
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,2,1,power_law_1.01,0.14841856002807616
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,2,1,power_law_1.01,0.15944704055786132
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,2,1,power_law_1.01,0.24285120010375977
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,2,1,power_law_1.01,0.21895360946655273
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,2,1,power_law_1.01,0.3525350570678711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,2,1,power_law_1.01,0.29203071594238283
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,2,1,power_law_1.01,0.401399040222168
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,2,1,power_law_1.01,0.5094899368286133
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,2,1,power_law_1.01,1.83384765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,2,1,power_law_1.01,0.7392018890380859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,1,power_law_1.2,0.15194496154785156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,2,1,power_law_1.01,0.9604434967041016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,1,power_law_1.2,0.30607999801635744
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,2,1,power_law_1.01,1.4036096191406249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,1,power_law_1.2,0.5311891174316405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,1,power_law_1.2,0.803201904296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,1,power_law_1.2,1.1432921600341797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,1,power_law_1.2,1.3442579650878907
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,2,1,power_law_1.01,2.71982666015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,1,power_law_1.2,1.8581459045410156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,1,power_law_1.2,1.9539794921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,1,power_law_1.2,2.0226963806152343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,1,power_law_1.2,2.0739923095703126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,1,power_law_1.2,2.245430450439453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,2,1,power_law_1.01,3.58253173828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,1,power_law_1.2,2.3237702941894534
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,1,power_law_1.2,2.546810302734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,1,power_law_1.2,2.7494305419921874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,1,power_law_1.2,2.891748352050781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,1,power_law_1.2,2.699388122558594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,1,power_law_1.2,2.969407958984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,1,power_law_1.2,3.2101324462890624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,1,power_law_1.2,3.6422296142578126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,1,power_law_1.2,4.080015258789063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,2,1,power_law_1.01,5.3532177734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,1,power_law_1.2,4.969774169921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,1,power_law_1.2,0.07548736095428467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,1,power_law_1.2,0.10048319816589354
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,1,power_law_1.2,0.16228864669799806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,1,power_law_1.2,5.8055181884765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,1,power_law_1.2,0.2844812774658203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,1,power_law_1.2,0.44571262359619135
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,1,power_law_1.2,7.438622436523437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,1,power_law_1.2,0.6341491317749023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,1,power_law_1.2,0.7612620544433594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,1,power_law_1.2,1.0033042907714844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,1,power_law_1.2,1.041119384765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,1,power_law_1.2,1.08316162109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,1,power_law_1.2,9.078695678710938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,1,power_law_1.2,1.1029676818847656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,1,power_law_1.2,1.165266571044922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,1,power_law_1.2,1.204738540649414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,1,power_law_1.2,12.23550537109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,1,power_law_1.2,1.2646886444091796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,1,power_law_1.2,1.2281088256835937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,1,power_law_1.2,1.2875328063964844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,1,power_law_1.2,1.4593606567382813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,1,power_law_1.2,1.5517926025390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,1,power_law_1.2,1.625463104248047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,1,power_law_1.2,1.7847602844238282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,1,power_law_1.2,1.9279743957519533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,1,power_law_1.2,2.241261444091797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,1,power_law_1.2,15.47100341796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,1,power_law_1.2,2.527953338623047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,1,power_law_1.2,3.070748291015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,1,power_law_1.2,3.647189636230469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,2,1,power_law_1.2,0.08933759689331054
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,1,power_law_1.2,4.7016293334960935
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,2,1,power_law_1.2,0.0916256046295166
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,2,1,power_law_1.2,0.0928281593322754
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,2,1,power_law_1.2,0.09982848167419434
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,2,1,power_law_1.2,0.111463041305542
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,2,1,power_law_1.2,0.12372991561889649
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,1,power_law_1.2,5.894326171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,2,1,power_law_1.2,0.1361081600189209
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,2,1,power_law_1.2,0.14795136451721191
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,2,1,power_law_1.2,0.18324928283691405
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,2,1,power_law_1.2,0.15924991607666014
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,2,1,power_law_1.2,0.21930879592895508
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,2,1,power_law_1.2,0.24281343460083007
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,2,1,power_law_1.2,0.2909574317932129
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,2,1,power_law_1.2,0.3516672134399414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,2,1,power_law_1.2,0.3998515319824219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,2,1,power_law_1.2,0.9594611358642577
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,2,1,power_law_1.2,1.8339967346191408
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,2,1,power_law_1.2,2.7198040771484377
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,2,1,power_law_1.2,3.582665100097656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,2,1,power_law_1.2,0.08769920349121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,2,1,power_law_1.2,0.5151923370361329
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,2,balanced,0.049603838920593266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,2,balanced,0.06104959964752197
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,2,balanced,0.09246335983276367
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,2,1,power_law_1.2,0.7392902374267578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,2,balanced,0.17789823532104493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,2,balanced,0.33168128967285154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,2,balanced,0.5946611022949219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,2,balanced,0.6002239990234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,2,balanced,0.6014003372192382
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,2,balanced,0.606556167602539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,2,balanced,0.6089382553100586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,2,1,power_law_1.2,1.395779266357422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,2,balanced,0.6148313522338867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,2,balanced,0.6907123565673828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,2,balanced,0.7006169891357421
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,2,balanced,0.7026483154296874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,2,balanced,0.6893593597412109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,2,balanced,0.7097996520996094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,2,balanced,0.7209011077880859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,2,balanced,0.7473318481445312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,2,balanced,0.7762278747558594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,2,balanced,0.8870835113525392
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,2,balanced,1.2971609497070313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,2,balanced,1.986713562011719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,2,balanced,2.5965863037109376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,2,balanced,0.03058687925338745
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,2,balanced,0.042928638458251955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,2,balanced,0.07095680236816407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,2,balanced,0.10693375587463379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,2,balanced,0.19121280670166016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,2,balanced,4.143274230957031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,2,balanced,0.32588672637939453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,2,balanced,0.33616897583007815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,2,balanced,0.33043071746826175
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,2,balanced,0.3353561782836914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,2,balanced,0.3327257537841797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,2,balanced,0.3356748962402344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,2,balanced,0.37521408081054686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,2,balanced,0.3709439849853516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,2,balanced,5.464554443359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,2,balanced,1.411735076904297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,2,balanced,0.37200126647949217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,2,balanced,0.37342334747314454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,2,balanced,0.3859404754638672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,2,balanced,0.37904895782470704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,2,balanced,0.4386048126220703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,2,1,power_law_1.2,5.355256347656249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,2,balanced,0.37139583587646485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,2,balanced,0.439758071899414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,2,balanced,0.5649958419799804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,2,balanced,0.48846462249755856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,2,balanced,0.9046860504150391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,2,balanced,0.6381926345825195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,2,balanced,1.0422386932373047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,2,balanced,1.5033740234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,2,balanced,1.9155941772460938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,1,2,balanced,0.04482304096221924
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,1,2,balanced,0.2039423942565918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,1,2,balanced,0.2045017623901367
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,1,2,balanced,0.20498687744140626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,1,2,balanced,0.20492160797119138
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,1,2,balanced,0.20492799758911132
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,1,2,balanced,0.20511104583740236
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,1,2,balanced,0.20630527496337892
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,1,2,balanced,0.20685951232910157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,1,2,balanced,0.037487359046936036
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,1,2,balanced,0.0748095989227295
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,1,2,balanced,0.20834047317504883
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,1,2,balanced,0.054712319374084474
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,1,2,balanced,0.20968063354492186
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,1,2,balanced,0.21381759643554688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,1,2,balanced,0.21820159912109377
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,1,2,balanced,0.22375808715820314
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,1,2,balanced,0.23261951446533202
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,1,2,balanced,0.2421721649169922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,1,2,balanced,0.2615872001647949
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,1,2,balanced,0.279182071685791
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,1,2,balanced,0.20891775131225584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,1,2,balanced,0.11710335731506347
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,2,power_law_1.01,0.06091135978698731
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,2,power_law_1.01,0.09279616355895996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,2,power_law_1.01,0.1721049690246582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,2,power_law_1.01,0.22909183502197267
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,1,2,balanced,0.6483097839355468
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,1,2,balanced,0.45675647735595704
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,1,2,balanced,0.41345664978027347
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,2,power_law_1.01,0.3105036735534668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,2,power_law_1.01,0.3560704040527344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,1,2,balanced,0.8609881591796874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,2,power_law_1.01,0.44353790283203126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,2,power_law_1.01,0.4583232116699219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,2,power_law_1.01,0.5077747344970703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,2,power_law_1.01,0.5130854415893554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,2,power_law_1.01,0.5904780960083007
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,2,power_law_1.01,0.6104947280883789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,2,power_law_1.01,0.6258943939208985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,2,power_law_1.01,0.6268044662475586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,2,power_law_1.01,0.6562751770019531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,2,power_law_1.01,0.6795161437988282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,2,power_law_1.01,0.7490483093261718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,2,power_law_1.01,0.8088102722167969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,2,power_law_1.01,0.9437055969238282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,2,power_law_1.01,1.0797913360595703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,2,power_law_1.01,1.375650634765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,2,power_law_1.01,1.6145458984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,2,power_law_1.01,0.5321894454956054
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,2,power_law_1.01,0.04325247764587402
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,2,power_law_1.01,2.349971160888672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,2,power_law_1.01,3.1042880249023437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,2,power_law_1.01,0.0702233600616455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,2,power_law_1.01,0.0986457633972168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,2,power_law_1.01,0.13218048095703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,2,power_law_1.01,0.1700057601928711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,2,power_law_1.01,0.19914880752563477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,2,power_law_1.01,0.25190784454345705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,2,power_law_1.01,4.166645812988281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,2,power_law_1.01,0.2625996780395508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,2,power_law_1.01,0.27289472579956053
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,2,power_law_1.01,0.2852697563171387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,2,power_law_1.01,0.3224383926391602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,2,power_law_1.01,0.29711103439331055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,2,power_law_1.01,0.32713344573974606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,2,power_law_1.01,0.3335513687133789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,2,power_law_1.01,0.3366873550415039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,2,power_law_1.01,0.3373183822631836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,2,power_law_1.01,0.35131904602050784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,2,power_law_1.01,5.25343994140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,2,power_law_1.01,0.37196414947509765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,2,power_law_1.01,0.3971916961669922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,2,power_law_1.01,0.45055488586425785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,2,power_law_1.01,0.5099673461914063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,2,power_law_1.01,0.6259673690795898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,2,power_law_1.01,0.7331520080566406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,2,power_law_1.01,0.9670451354980468
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,1,2,power_law_1.01,0.04565120220184326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,2,power_law_1.01,1.2915303039550783
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,1,2,power_law_1.01,0.05883264064788818
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,1,2,power_law_1.01,0.07165440082550048
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,1,2,power_law_1.01,0.08723711967468262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,2,power_law_1.01,1.690416564941406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,1,2,power_law_1.01,0.10666496276855468
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,1,2,power_law_1.01,0.12516223907470703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,1,2,power_law_1.01,0.16155008316040037
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,1,2,power_law_1.01,0.1631987190246582
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,1,2,power_law_1.01,0.16752256393432616
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,1,2,power_law_1.01,0.17646976470947268
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,1,2,power_law_1.01,0.17741056442260744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,2,power_law_1.01,2.2342066955566406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,1,2,power_law_1.01,0.18114175796508789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,1,2,power_law_1.01,0.18986623764038085
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,1,2,power_law_1.01,0.19374080657958986
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,1,2,power_law_1.01,0.2009663963317871
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,1,2,power_law_1.01,0.20249216079711915
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,1,2,power_law_1.01,0.2071232032775879
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,1,2,power_law_1.01,0.22863744735717773
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,1,2,power_law_1.01,0.29308544158935546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,1,2,power_law_1.01,0.337918701171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,1,2,power_law_1.01,0.4344934463500977
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,1,2,power_law_1.01,0.5219071960449219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,1,2,power_law_1.01,0.25473024368286135
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,2,power_law_1.2,0.061401600837707515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,2,power_law_1.2,0.09286784172058106
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,2,power_law_1.2,0.1655513572692871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,2,power_law_1.2,0.21303808212280276
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,2,power_law_1.2,0.27257471084594725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,2,power_law_1.2,0.35617534637451176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,2,power_law_1.2,0.4872639846801758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,2,power_law_1.2,0.49861503601074214
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,2,power_law_1.2,0.503322868347168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,2,power_law_1.2,0.590489616394043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,2,power_law_1.2,0.5953484725952148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,2,power_law_1.2,0.6269247817993164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,2,power_law_1.2,0.5327321624755859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,2,power_law_1.2,0.6685517120361328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,2,power_law_1.2,0.6941120147705078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,2,power_law_1.2,0.7746406555175781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,2,power_law_1.2,0.8472665405273437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,2,power_law_1.2,0.9882240295410156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,2,power_law_1.2,1.1279872131347655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,2,power_law_1.2,1.5048806762695313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,2,power_law_1.2,0.459185905456543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,2,power_law_1.2,1.8139199829101564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,2,power_law_1.2,2.5061080932617186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,2,power_law_1.2,3.049280090332031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,2,power_law_1.2,0.6415180969238281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,2,power_law_1.2,4.597564086914063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,2,power_law_1.2,0.09733504295349121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,2,power_law_1.2,0.11106431961059571
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,2,power_law_1.2,0.15526016235351564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,2,power_law_1.2,0.18407936096191407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,2,power_law_1.2,5.605198364257812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,2,power_law_1.2,0.24778879165649412
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,2,power_law_1.2,0.27537151336669924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,2,power_law_1.2,0.26784000396728513
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,2,power_law_1.2,0.2846054458618164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,2,power_law_1.2,0.2925350379943848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,2,power_law_1.2,0.3262182235717773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,2,power_law_1.2,0.33122432708740235
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,1,2,power_law_1.01,0.9179840087890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,2,power_law_1.2,0.3337510299682617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,2,power_law_1.2,0.3379430389404297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,2,power_law_1.2,0.35140094757080076
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,2,power_law_1.2,0.3615590286254883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,2,power_law_1.2,0.3819276809692383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,2,power_law_1.2,0.0399833607673645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,2,power_law_1.2,0.42162303924560546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,2,power_law_1.2,0.4739724731445312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,1,2,power_law_1.01,1.3973362731933594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,2,power_law_1.2,0.06718592166900635
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,2,power_law_1.2,0.5344384002685547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,2,power_law_1.2,0.6506867218017578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,2,power_law_1.2,0.770728988647461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,2,power_law_1.2,0.9915570831298828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,1,2,power_law_1.2,0.06089856147766113
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,1,2,power_law_1.2,0.06671616077423095
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,1,2,power_law_1.2,0.08391936302185059
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,2,power_law_1.2,1.2327334594726562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,1,2,power_law_1.2,0.10034687995910645
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,1,2,power_law_1.2,0.12017279624938966
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,1,2,power_law_1.2,0.15475199699401856
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,1,2,power_law_1.2,0.1659008026123047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,1,2,power_law_1.2,0.1706879997253418
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,1,2,power_law_1.2,0.17073408126831055
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,1,2,power_law_1.2,0.17830400466918944
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,1,2,power_law_1.2,0.1836761665344238
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,1,2,power_law_1.01,0.7773145294189453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,2,power_law_1.2,2.321139221191406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,1,2,power_law_1.2,0.18715648651123046
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,1,2,power_law_1.2,0.19332223892211914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,1,2,power_law_1.2,0.21029247283935545
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,1,2,power_law_1.2,0.2209894371032715
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,1,2,power_law_1.01,1.6921343994140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,2,power_law_1.2,1.7313087463378907
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,1,2,power_law_1.2,0.20086656570434572
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,1,2,power_law_1.2,0.24613759994506834
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,1,2,power_law_1.2,0.36067584991455076
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,1,2,power_law_1.2,0.41070079803466797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,1,2,power_law_1.2,0.2899609565734863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,2,1,balanced,0.05320703983306885
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,1,2,power_law_1.2,0.55491455078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,2,1,balanced,0.06915328025817871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,2,1,balanced,0.10390015602111817
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,2,1,balanced,0.18540224075317382
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,1,2,power_law_1.2,0.043797760009765624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,2,1,balanced,0.3346783828735352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,2,1,balanced,0.6131167984008788
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,2,1,balanced,0.6068428802490234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,2,1,balanced,0.6192992019653321
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,2,1,balanced,0.615865592956543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,2,1,balanced,0.6149708938598633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,2,1,balanced,0.625103988647461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,2,1,balanced,0.6261695861816406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,2,1,balanced,0.6286624145507813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,2,1,balanced,0.6410662078857422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,2,1,balanced,0.695827865600586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,2,1,balanced,0.6971302032470703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,2,1,balanced,0.7126175689697265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,2,1,balanced,0.7391078186035156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,2,1,balanced,0.7575839996337891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,2,1,balanced,0.8404434967041017
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,2,1,balanced,0.8968077087402344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,2,1,balanced,1.2271456146240234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,2,1,balanced,1.4152178955078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,2,1,balanced,1.9717823791503908
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,1,2,power_law_1.2,0.7271743774414062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,2,1,balanced,2.4458988952636718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,2,1,balanced,3.623843688964844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,2,1,balanced,0.08468928337097167
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,2,1,balanced,0.12326399803161621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,2,1,balanced,0.2068729591369629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,2,1,balanced,0.36157760620117185
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,2,1,balanced,0.037674241065979004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,2,1,balanced,0.36556095123291016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,2,1,balanced,4.773195495605469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,2,1,balanced,0.36540927886962893
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,2,1,balanced,0.3674995040893555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,2,1,balanced,0.37812545776367185
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,2,1,balanced,0.3727302551269531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,2,1,balanced,0.3762809753417969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,2,1,balanced,0.3799961471557617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,2,1,balanced,0.38178688049316406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,2,1,balanced,0.4292755126953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,2,1,balanced,0.052480640411376955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,2,1,balanced,0.4336684799194336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,2,1,balanced,0.4365478515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,2,1,balanced,0.4473408126831055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,2,1,balanced,0.4613407897949219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,2,1,balanced,0.484637451171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,2,1,balanced,0.5064633560180665
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,1,2,power_law_1.2,1.056189422607422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,1,2,power_law_1.2,1.3681919860839844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,2,1,balanced,0.5620608139038086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,2,1,balanced,0.6216831970214843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,2,1,balanced,0.9333529663085937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,2,1,balanced,1.0490476989746094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,2,1,balanced,1.9311398315429689
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,1,2,power_law_1.2,2.047587890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,1,2,power_law_1.2,3.0585534667968752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,2,1,balanced,1.4775302124023437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,2,1,balanced,0.058026242256164554
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,2,1,balanced,0.06403647899627686
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,2,1,balanced,0.09080831527709961
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,2,1,balanced,0.07281599998474121
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,2,1,balanced,0.22550336837768556
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,2,1,balanced,0.22823936462402344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,2,1,balanced,0.13305279731750488
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,2,1,balanced,0.22686912536621096
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,2,1,balanced,0.23041343688964844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,2,1,balanced,0.228591365814209
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,2,1,balanced,0.22985919952392578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,2,1,balanced,0.2259129524230957
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,2,1,balanced,0.23268735885620115
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,2,1,balanced,0.2360211181640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,2,1,balanced,0.23968000411987306
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,2,1,balanced,0.2459539222717285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,2,1,balanced,0.2534227180480957
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,2,1,balanced,0.2682732772827149
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,2,1,balanced,0.2838265609741211
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,2,1,balanced,0.3086649513244629
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,2,1,balanced,0.2256947135925293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,2,1,balanced,0.3371091079711914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,2,1,power_law_1.01,0.0519923210144043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,2,1,balanced,0.22451967239379883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,2,1,power_law_1.01,0.06881855964660645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,2,1,power_law_1.01,0.10392191886901855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,2,1,power_law_1.01,0.2640006446838379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,2,1,power_law_1.01,0.33537151336669924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,2,1,power_law_1.01,0.4470624160766602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,2,1,power_law_1.01,0.47838912963867186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,2,1,power_law_1.01,0.49126846313476563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,2,1,power_law_1.01,0.49435585021972656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,2,1,power_law_1.01,0.515475845336914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,2,1,power_law_1.01,0.5495347213745118
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,2,1,power_law_1.01,0.5693996810913087
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,2,1,power_law_1.01,0.6095731353759766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,2,1,power_law_1.01,0.6147654342651367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,2,1,power_law_1.01,0.6271500778198242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,2,1,power_law_1.01,0.6633932495117187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,2,1,power_law_1.01,0.708532485961914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,2,1,power_law_1.01,0.7773529815673828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,2,1,power_law_1.01,0.8927852630615234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,2,1,power_law_1.01,1.030911331176758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,2,1,power_law_1.01,1.3061970520019532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,2,1,power_law_1.01,1.6131341552734377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,2,1,power_law_1.01,2.1796083068847656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,2,1,power_law_1.01,0.1850579261779785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,2,1,power_law_1.01,0.03819839954376221
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,2,1,power_law_1.01,2.751614685058594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,2,1,power_law_1.01,0.08284223556518555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,2,1,power_law_1.01,0.12334015846252441
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,2,1,power_law_1.01,0.16945024490356447
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,2,1,power_law_1.01,3.8554266357421874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,2,1,power_law_1.01,0.22355199813842774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,2,1,power_law_1.01,0.2890905570983887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,2,1,power_law_1.01,0.2962361526489258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,2,1,power_law_1.01,0.30087039947509764
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,2,1,power_law_1.01,0.3054182434082031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,2,1,power_law_1.01,0.319879035949707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,2,1,power_law_1.01,0.32765377044677735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,2,1,power_law_1.01,4.958559875488281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,2,1,power_law_1.01,0.056431999206542974
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,2,1,power_law_1.01,0.33147712707519533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,2,1,power_law_1.01,0.33783870697021484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,2,1,power_law_1.01,0.3871360015869141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,2,1,power_law_1.01,0.40563838958740234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,2,1,power_law_1.01,0.420574722290039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,2,1,power_law_1.01,0.447715835571289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,2,1,power_law_1.01,0.49050559997558596
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,2,1,power_law_1.01,0.39272895812988284
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,2,1,balanced,0.5138668823242187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,2,1,power_law_1.01,0.5360505676269531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,2,1,power_law_1.01,0.6327673721313476
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,2,1,balanced,0.5784390258789063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,2,1,power_law_1.01,0.7428217315673828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,2,1,power_law_1.01,0.9517926025390626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,2,1,power_law_1.01,1.1872051239013672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,2,1,power_law_1.01,0.057392640113830565
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,2,1,balanced,0.8177728271484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,2,1,balanced,1.1022509002685548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,2,1,power_law_1.01,1.672678985595703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,2,1,power_law_1.01,0.06353087902069092
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,2,1,power_law_1.01,0.07770944118499756
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,2,1,power_law_1.01,2.0480108642578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,2,1,power_law_1.01,0.11366080284118651
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,2,1,power_law_1.01,0.13266176223754883
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,2,1,power_law_1.01,0.1732480049133301
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,2,1,power_law_1.01,0.18034368515014648
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,2,1,power_law_1.01,0.09368063926696776
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,2,1,power_law_1.01,0.18448511123657227
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,2,1,power_law_1.01,0.18877376556396483
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,2,1,power_law_1.01,0.20415615081787108
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,2,1,power_law_1.01,0.20712127685546874
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,2,1,power_law_1.01,0.2115705680847168
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,2,1,power_law_1.01,0.22123584747314454
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,2,1,power_law_1.01,0.22681983947753906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,2,1,power_law_1.01,0.2420908737182617
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,2,1,power_law_1.01,0.19751232147216796
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,2,1,power_law_1.01,0.29643520355224606
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,2,1,power_law_1.01,0.42076351165771486
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,2,1,power_law_1.01,0.3578278350830078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,2,1,power_law_1.01,0.26777919769287106
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,2,1,power_law_1.01,0.6891693115234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,2,1,power_law_1.2,0.055478401184082024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,2,1,power_law_1.2,0.06918591976165771
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,2,1,power_law_1.01,0.5490572738647461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,2,1,power_law_1.2,0.1032313632965088
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,2,1,power_law_1.2,0.1812646484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,2,1,power_law_1.2,0.25229568481445314
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,2,1,power_law_1.01,0.9723564910888672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,2,1,power_law_1.2,0.3135807991027832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,2,1,power_law_1.2,0.4366342544555664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,2,1,power_law_1.2,0.46403648376464846
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,2,1,power_law_1.01,1.2829887390136718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,2,1,power_law_1.2,0.4782681655883789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,2,1,power_law_1.2,0.49595520019531253
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,2,1,power_law_1.2,0.5138028717041015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,2,1,power_law_1.2,0.5422323226928711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,2,1,power_law_1.2,0.5709497451782226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,2,1,power_law_1.2,0.6280582427978516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,2,1,power_law_1.2,0.636819190979004
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,2,1,power_law_1.01,1.87666748046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,2,1,power_law_1.2,0.6523481750488281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,2,1,power_law_1.2,0.6741075134277343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,2,1,power_law_1.2,0.7388082885742188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,2,1,power_law_1.2,0.7909721374511719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,2,1,power_law_1.2,0.9262329864501954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,2,1,power_law_1.2,1.069359359741211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,2,1,power_law_1.2,1.2875077819824219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,2,1,power_law_1.2,0.040628480911254886
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,2,1,power_law_1.2,1.627921905517578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,2,1,power_law_1.2,0.05642623901367187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,2,1,power_law_1.2,0.08337856292724609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,2,1,power_law_1.2,2.1940454101562503
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,2,1,power_law_1.2,0.15954431533813476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,2,1,power_law_1.2,0.21345600128173828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,2,1,power_law_1.2,0.2853683280944824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,2,1,power_law_1.2,2.779151306152344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,2,1,power_law_1.2,0.29203903198242187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,2,1,power_law_1.2,0.2994047927856445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,2,1,power_law_1.2,0.30911231994628907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,2,1,power_law_1.2,0.3223014450073242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,2,1,power_law_1.2,0.32674816131591794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,2,1,power_law_1.2,0.3336281585693359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,2,1,power_law_1.2,0.3426015853881836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,2,1,power_law_1.2,0.39249534606933595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,2,1,power_law_1.2,0.4041388702392578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,2,1,power_law_1.2,3.8632754516601566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,2,1,power_law_1.2,0.40928833007812504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,2,1,power_law_1.2,0.4376838302612304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,2,1,power_law_1.2,0.4628313446044922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,2,1,power_law_1.2,0.513012466430664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,2,1,power_law_1.2,0.5635417556762695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,2,1,power_law_1.2,0.6655078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,2,1,power_law_1.2,5.019626770019531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,2,1,power_law_1.2,0.7769977569580078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,2,1,power_law_1.2,0.9824761962890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,2,1,power_law_1.2,1.2107443237304687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,2,1,power_law_1.2,0.11706815719604494
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,2,1,power_law_1.2,0.057217278480529786
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,2,1,power_law_1.2,0.06360127925872802
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,2,1,power_law_1.2,1.651111755371094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,2,1,power_law_1.2,0.07670080184936523
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,2,1,power_law_1.2,0.09016768455505371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,2,1,power_law_1.2,2.109607696533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,2,1,power_law_1.2,0.107959041595459
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,2,1,power_law_1.2,0.12812671661376954
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,2,1,power_law_1.2,0.17852287292480468
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,2,1,power_law_1.2,0.16987327575683592
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,2,1,power_law_1.2,0.18192447662353514
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,2,1,power_law_1.2,0.18941120147705076
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,2,1,power_law_1.2,0.1969043159484863
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,2,1,power_law_1.2,0.20337152481079102
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,2,1,power_law_1.2,0.2209446334838867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,2,1,power_law_1.2,0.2116115188598633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,2,1,power_law_1.2,0.23162559509277342
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,2,1,power_law_1.2,0.24263999938964842
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,2,1,power_law_1.2,0.26129024505615234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,2,1,power_law_1.01,2.5015359497070313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,2,1,power_law_1.2,0.3475743865966797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,2,1,power_law_1.2,0.4466259384155274
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,2,1,power_law_1.2,0.30525888442993165
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,2,1,power_law_1.2,0.7668621063232421
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,2,1,power_law_1.2,0.545359992980957
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,2,1,power_law_1.2,1.005984649658203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,2,1,power_law_1.2,1.5115199279785156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,2,balanced,0.03990528106689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,2,balanced,0.08179455757141113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,2,balanced,0.054999041557312014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,2,balanced,0.1377459239959717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,2,balanced,0.31057024002075195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,2,balanced,0.3120652770996094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,2,balanced,0.22708480834960937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,2,balanced,0.31591167449951174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,2,balanced,0.3187635231018066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,2,balanced,0.320711669921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,2,1,power_law_1.2,2.062957458496094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,2,balanced,0.30769920349121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,2,balanced,0.30282112121582033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,2,balanced,0.30940799713134765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,2,balanced,0.34912254333496096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,2,balanced,0.34452865600585936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,2,balanced,0.53644287109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,2,balanced,0.5594931030273438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,2,balanced,0.3487807846069336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,2,balanced,0.6408895874023438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,2,balanced,1.0519385528564453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,2,balanced,1.1318975830078126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,2,balanced,0.3536127853393555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,2,balanced,0.36919551849365234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,2,balanced,1.6386904907226563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,2,balanced,1.8116493225097656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,2,balanced,0.4043942260742187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,2,balanced,0.6318771362304687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,2,1,power_law_1.2,3.2416839599609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,2,1,power_law_1.2,4.516748657226563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,1,2,balanced,0.03302016019821167
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,1,2,balanced,0.02929919958114624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,1,2,balanced,0.04375679969787598
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,1,2,balanced,0.12039039611816407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,1,2,balanced,0.12080512046813965
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,1,2,balanced,0.12182527542114259
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,1,2,balanced,0.12201215744018554
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,1,2,balanced,0.12208000183105469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,1,2,balanced,0.12418432235717773
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,1,2,balanced,0.12410880088806153
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,1,2,balanced,0.12522239685058595
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,1,2,balanced,0.12764672279357908
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,1,2,balanced,0.13184639930725098
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,1,2,balanced,0.06222847938537598
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,1,2,balanced,0.13484416007995606
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,1,2,balanced,0.14295040130615236
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,1,2,balanced,0.0949452781677246
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,1,2,balanced,0.1505497646331787
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,1,2,balanced,0.12030719757080077
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,1,2,balanced,0.11996671676635742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,1,2,balanced,0.1204595184326172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,2,power_law_1.01,0.0736575984954834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,2,power_law_1.01,0.0471833610534668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,2,power_law_1.01,0.0969382381439209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,2,power_law_1.01,0.13660799980163574
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,1,2,balanced,0.2376908874511719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,1,2,balanced,0.3356044769287109
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,1,2,balanced,0.22227712631225588
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,1,2,balanced,0.3695116806030273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,2,power_law_1.01,0.2768448066711426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,2,power_law_1.01,0.2917299270629883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,2,power_law_1.01,0.17050752639770508
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,1,2,balanced,0.5715840148925782
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,1,2,balanced,0.7074649810791016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,2,power_law_1.01,0.3595942306518555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,2,power_law_1.01,0.26864511489868165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,2,power_law_1.01,0.27600255966186527
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,2,power_law_1.01,0.28313087463378905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,2,power_law_1.01,0.3006131172180176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,2,power_law_1.01,0.5686528015136718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,2,power_law_1.01,0.3090444755554199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,2,power_law_1.01,0.6214630508422851
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,2,power_law_1.01,0.8111347198486328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,2,power_law_1.01,0.24508544921875003
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,2,power_law_1.01,0.35425281524658203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,2,power_law_1.01,1.0642022705078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,2,power_law_1.01,1.801799621582031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,1,2,power_law_1.01,0.03651583909988403
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,1,2,power_law_1.01,0.04663680076599121
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,1,2,power_law_1.01,0.05240320205688477
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,1,2,power_law_1.01,0.06598400115966797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,1,2,power_law_1.01,0.07860352039337158
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,1,2,power_law_1.01,0.10407423973083496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,2,power_law_1.01,0.42116737365722656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,2,power_law_1.01,0.326506233215332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,1,2,power_law_1.01,0.10673791885375976
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,1,2,power_law_1.01,0.10951295852661133
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,1,2,power_law_1.01,0.10735615730285644
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,1,2,power_law_1.01,0.11133695602416993
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,1,2,power_law_1.01,0.1130457592010498
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,1,2,power_law_1.01,0.11742848396301268
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,1,2,power_law_1.01,0.11902079582214356
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,1,2,power_law_1.01,0.12205568313598633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,1,2,power_law_1.01,0.12554752349853515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,2,power_law_1.01,0.35628543853759764
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,2,power_law_1.01,0.3903142547607422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,2,power_law_1.01,0.6039910507202149
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,2,power_law_1.01,1.2378521728515626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,2,power_law_1.01,2.0362765502929685
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,1,2,power_law_1.01,0.13023488044738768
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,1,2,power_law_1.01,0.1337676811218262
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,1,2,power_law_1.01,0.15183615684509277
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,1,2,power_law_1.01,0.1672921562194824
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,1,2,power_law_1.01,0.19769344329833985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,2,power_law_1.2,0.04679296016693115
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,1,2,power_law_1.01,0.23337472915649413
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,2,power_law_1.2,0.06531455993652344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,2,power_law_1.2,0.09742207527160644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,2,power_law_1.2,0.1634316825866699
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,1,2,power_law_1.01,0.32397567749023437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,2,power_law_1.2,0.26864639282226566
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,1,2,power_law_1.01,0.39410430908203126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,2,power_law_1.2,0.12799872398376463
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,2,power_law_1.2,0.2676902389526367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,2,power_law_1.2,0.2834880065917969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,2,power_law_1.2,0.30297088623046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,2,power_law_1.2,0.3164787292480469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,2,power_law_1.2,0.25162752151489254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,2,power_law_1.2,0.28506496429443356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,2,power_law_1.2,0.3381452941894531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,2,power_law_1.2,0.3549529647827149
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,2,power_law_1.2,0.4029401779174805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,1,2,power_law_1.01,0.5412275314331054
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,2,power_law_1.2,0.29387264251708983
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,2,power_law_1.2,0.5835878372192382
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,1,2,power_law_1.01,0.7596377563476563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,2,power_law_1.2,0.6400563049316406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,2,power_law_1.2,0.6816998291015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,1,2,power_law_1.01,1.0841433715820312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,2,power_law_1.2,0.3704934310913086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,2,power_law_1.2,0.8625804901123046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,2,power_law_1.2,1.109075164794922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,1,2,power_law_1.2,0.03585536003112793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,1,2,power_law_1.2,0.04264448165893554
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,1,2,power_law_1.2,0.05248000144958496
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,1,2,power_law_1.2,0.06360960006713867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,1,2,power_law_1.2,0.07499648094177246
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,1,2,power_law_1.2,0.10081791877746582
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,1,2,power_law_1.01,1.4749440002441407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,1,2,power_law_1.2,0.10869248390197754
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,1,2,power_law_1.2,0.11281536102294923
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,1,2,power_law_1.2,0.10478976249694824
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,1,2,power_law_1.2,0.11462783813476562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,1,2,power_law_1.2,0.11444479942321777
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,1,2,power_law_1.2,0.12026495933532715
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,1,2,power_law_1.2,0.11951104164123535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,2,power_law_1.2,1.32276611328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,1,2,power_law_1.2,0.12845567703247068
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,1,2,power_law_1.2,0.13758079528808592
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,2,power_law_1.2,1.775089874267578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,1,2,power_law_1.2,0.14981760025024413
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,1,2,power_law_1.2,0.19333248138427733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,2,power_law_1.2,2.2617958068847654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,2,power_law_1.2,0.3639756774902344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,2,power_law_1.2,0.42781311035156244
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,1,2,power_law_1.2,0.12766592025756834
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,1,2,power_law_1.2,0.16652544021606447
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,1,2,power_law_1.2,0.2514252853393555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,1,2,power_law_1.2,0.4143807983398437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,1,2,power_law_1.2,0.300447998046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,1,2,power_law_1.2,0.5235225677490234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,1,balanced,0.08482048034667969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,1,balanced,0.05575295925140381
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,1,balanced,0.13849023818969725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,1,balanced,0.3169171142578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,1,2,power_law_1.2,0.7569766235351563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,1,balanced,0.31781375885009766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,1,balanced,0.3198297691345215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,1,balanced,0.32544319152832035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,1,balanced,0.3276927947998047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,1,balanced,0.32916736602783203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,1,balanced,0.23572223663330077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,1,balanced,0.3140768051147461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,1,balanced,0.3332255935668945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,1,balanced,0.42634048461914065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,1,balanced,0.037351040840148925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,1,balanced,0.3219065475463867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,1,balanced,0.3291417694091797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,1,2,power_law_1.2,1.1221644592285156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,1,balanced,0.36485824584960935
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,1,balanced,0.48178047180175787
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,1,balanced,0.44180927276611326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,1,balanced,0.3858899307250977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,1,balanced,0.4616672134399414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,1,balanced,0.6127443313598633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,1,2,power_law_1.2,1.4569996643066407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,1,balanced,0.6361772918701172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,1,2,power_law_1.2,2.037671661376953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,1,balanced,1.0974655914306641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,1,balanced,1.1732198333740234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,1,balanced,1.7053555297851564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,1,balanced,1.8223333740234373
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,2,1,balanced,0.04439743995666504
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,2,1,balanced,0.038621439933776855
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,2,1,balanced,0.05419712066650391
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,2,1,balanced,0.07353407859802245
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,2,1,balanced,0.13814016342163088
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,2,1,balanced,0.13786815643310546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,2,1,balanced,0.13747712135314943
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,2,1,balanced,0.13890496253967285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,2,1,balanced,0.10991935729980469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,2,1,balanced,0.1405075168609619
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,2,1,balanced,0.14093824386596682
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,2,1,balanced,0.14144895553588868
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,2,1,balanced,0.143822078704834
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,2,1,balanced,0.14794367790222168
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,2,1,balanced,0.15317824363708496
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,2,1,balanced,0.15979583740234377
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,2,1,balanced,0.1691059112548828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,2,1,balanced,0.13616767883300782
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,2,1,balanced,0.1809075164794922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,2,1,balanced,0.1365881633758545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,1,power_law_1.01,0.03570048093795776
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,2,1,balanced,0.137194881439209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,1,power_law_1.01,0.05452095985412597
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,1,power_law_1.01,0.08477312088012695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,1,power_law_1.01,0.16932416915893556
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,1,power_law_1.01,0.12858688354492187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,1,power_law_1.01,0.25885183334350587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,1,power_law_1.01,0.24435264587402344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,1,power_law_1.01,0.27421056747436523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,1,power_law_1.01,0.28425472259521484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,1,power_law_1.01,0.29769407272338866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,1,power_law_1.01,0.31099456787109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,1,power_law_1.01,0.3220902252197265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,1,power_law_1.01,0.34443584442138675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,1,power_law_1.01,0.3660569763183594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,1,power_law_1.01,0.39312191009521485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,1,power_law_1.01,0.2892518424987793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,1,power_law_1.01,0.3874726486206055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,1,power_law_1.01,0.45056896209716796
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,2,1,balanced,0.30554176330566407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,2,1,balanced,0.28072256088256836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,2,1,balanced,0.4830118560791016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,2,1,balanced,0.4377990341186523
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,2,1,balanced,0.7733747100830077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,1,power_law_1.01,0.7234419250488282
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,2,1,balanced,0.969952621459961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,1,power_law_1.01,0.848927993774414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,1,power_law_1.01,1.0676678466796876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,1,power_law_1.01,1.2578182220458984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,2,1,power_law_1.01,0.03959167957305908
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,2,1,power_law_1.01,0.045483517646789554
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,2,1,power_law_1.01,0.055395197868347165
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,2,1,power_law_1.01,0.0695091199874878
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,1,power_law_1.01,2.0775602722167967
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,2,1,power_law_1.01,0.08229056358337403
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,2,1,power_law_1.01,0.11212800025939942
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,2,1,power_law_1.01,0.11484416007995606
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,2,1,power_law_1.01,0.12208383560180665
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,2,1,power_law_1.01,0.12458111763000487
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,1,power_law_1.01,0.4245292663574219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,2,1,power_law_1.01,0.12552384376525877
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,2,1,power_law_1.01,0.12868351936340333
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,2,1,power_law_1.01,0.13283455848693848
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,2,1,power_law_1.01,0.1359436798095703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,2,1,power_law_1.01,0.139999361038208
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,2,1,power_law_1.01,0.1451219177246094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,2,1,power_law_1.01,0.14927871704101564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,1,power_law_1.01,0.4832230377197265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,1,power_law_1.01,0.5399168014526368
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,1,power_law_1.01,1.6615347290039062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,2,1,power_law_1.01,0.15956607818603516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,2,1,power_law_1.01,0.1813100814819336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,2,1,power_law_1.01,0.3034796714782715
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,2,1,power_law_1.01,0.20390207290649415
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,2,1,power_law_1.01,0.25129791259765627
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,2,1,power_law_1.01,0.4052703857421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,2,1,power_law_1.01,0.521789436340332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,1,power_law_1.2,0.03696127891540527
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,1,power_law_1.2,0.1597785568237305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,2,1,power_law_1.01,0.745617904663086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,1,power_law_1.2,0.24311616897583005
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,1,power_law_1.2,0.26314176559448244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,1,power_law_1.2,0.27801792144775395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,1,power_law_1.2,0.05558784008026123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,1,power_law_1.2,0.2846419143676758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,1,power_law_1.2,0.30460351943969727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,1,power_law_1.2,0.31940799713134765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,1,power_law_1.2,0.33263359069824217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,1,power_law_1.2,0.08227392196655273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,1,power_law_1.2,0.3582284927368164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,1,power_law_1.2,0.3765651321411133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,1,power_law_1.2,0.4005177688598633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,1,power_law_1.2,0.1196019172668457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,1,power_law_1.2,0.42936767578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,1,power_law_1.2,0.5044089508056641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,1,power_law_1.2,0.5545401763916016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,1,power_law_1.2,0.29372159957885746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,1,power_law_1.2,0.854163818359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,1,power_law_1.2,1.0813683319091798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,1,power_law_1.2,0.3971974563598633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,1,power_law_1.2,0.46694976806640626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,2,1,power_law_1.2,0.03900160074234009
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,1,power_law_1.2,1.2900991821289063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,2,1,power_law_1.2,0.04599167823791504
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,2,1,power_law_1.2,0.055753598213195796
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,2,1,power_law_1.2,0.06511168003082275
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,2,1,power_law_1.2,0.07896448135375976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,1,power_law_1.2,0.7626035308837891
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,2,1,power_law_1.2,0.1115392017364502
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,2,1,power_law_1.2,0.12224255561828615
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,2,1,power_law_1.2,0.12512895584106445
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,2,1,power_law_1.2,0.12822912216186524
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,2,1,power_law_1.2,0.11536959648132324
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,2,1,power_law_1.2,0.13171775817871095
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,2,1,power_law_1.2,0.1365824031829834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,1,power_law_1.2,1.6941215515136718
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,2,1,power_law_1.2,0.1394547176361084
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,2,1,power_law_1.2,0.14581503868103027
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,2,1,power_law_1.2,0.15225343704223632
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,2,1,power_law_1.01,0.9827315521240234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,2,1,power_law_1.01,1.4558636474609377
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,2,1,power_law_1.01,1.9668019104003907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,1,power_law_1.2,2.126024932861328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,2,1,power_law_1.2,0.17298688888549804
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,2,1,power_law_1.2,0.1603750419616699
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,2,1,power_law_1.2,0.2364339256286621
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,2,1,power_law_1.2,0.20437887191772458
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,2,1,power_law_1.2,0.3098796844482422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,2,1,power_law_1.2,0.3895347213745117
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,2,1,power_law_1.2,0.5598751831054687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,2,1,power_law_1.2,0.7401363372802734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,2,balanced,0.15639552116394043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,2,balanced,0.2483148765563965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,2,balanced,0.4525363159179688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,2,balanced,0.8772608184814453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,2,balanced,1.6690725708007812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,2,balanced,2.4065318298339844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,2,balanced,2.4270745849609376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,2,balanced,2.4500825500488284
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,2,balanced,2.477907257080078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,2,balanced,2.518524169921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,2,1,power_law_1.2,1.105704345703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,2,balanced,2.486968383789063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,2,balanced,2.5387059020996094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,2,1,power_law_1.2,1.4909030151367186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,2,balanced,2.4989888000488283
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,2,1,power_law_1.2,2.2942291259765626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,2,balanced,2.651866760253906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,2,balanced,2.632862854003906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,2,balanced,3.1364517211914062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,2,balanced,3.288620910644531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,2,balanced,2.666983642578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,2,balanced,2.868096008300781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,2,balanced,2.9561126708984373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,2,balanced,4.536537475585938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,2,1,power_law_1.2,3.1069650268554687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,2,balanced,0.16000255584716797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,2,balanced,0.0940492820739746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,2,balanced,5.519984741210938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,2,balanced,0.27477632522583006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,2,balanced,0.5208652877807618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,2,balanced,0.9594380950927734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,2,balanced,1.3620390319824218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,2,balanced,7.757861328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,2,balanced,1.36682373046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,2,balanced,8.291309814453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,2,balanced,1.3760166931152344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,2,balanced,1.4023551940917969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,2,balanced,1.40748291015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,2,balanced,1.411420135498047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,2,balanced,1.4355059814453124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,2,balanced,1.4516543579101562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,2,balanced,1.4899623107910156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,2,balanced,1.5783488464355468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,2,balanced,1.638592071533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,2,balanced,1.6490406799316406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,2,balanced,1.778056945800781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,2,balanced,1.7988864135742186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,2,balanced,12.679034423828124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,2,balanced,2.00171142578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,2,balanced,2.080821685791016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,2,balanced,4.611434326171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,2,balanced,3.2566708374023436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,2,balanced,3.5024664306640623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,2,balanced,15.363092041015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,2,balanced,4.933943176269532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,2,balanced,5.423777465820312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,2,balanced,8.347179565429688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,2,balanced,10.022654418945312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,1,2,balanced,0.08614912033081054
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,1,2,balanced,0.11680640220642088
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,1,2,balanced,0.8629747009277343
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,1,2,balanced,0.1818739128112793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,1,2,balanced,0.8611392211914062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,1,2,balanced,0.8696192169189454
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,1,2,balanced,0.864990692138672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,1,2,balanced,0.876336669921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,1,2,balanced,0.8956594848632813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,1,2,balanced,0.8902783966064453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,1,2,balanced,0.32348545074462887
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,1,2,balanced,0.5923724746704101
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,1,2,balanced,0.9009203338623047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,1,2,balanced,0.9100172424316406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,1,2,balanced,0.9388070678710937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,1,2,balanced,1.0009254455566405
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,1,2,balanced,1.0441356658935548
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,1,2,balanced,0.8456985473632812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,1,2,balanced,0.9302361297607422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,1,2,balanced,0.843852767944336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,1,2,balanced,0.8458073425292969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,2,power_law_1.01,0.2540019226074219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,2,power_law_1.01,0.5870169448852539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,2,power_law_1.01,0.44210689544677734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,2,power_law_1.01,1.225432357788086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,2,power_law_1.01,0.8877734374999999
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,2,power_law_1.01,1.965813751220703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,2,power_law_1.01,2.0029939270019534
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,2,power_law_1.01,2.0743128967285154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,2,power_law_1.01,2.1851776123046873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,2,power_law_1.01,2.2764633178710936
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,1,2,balanced,1.7006182861328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,1,2,balanced,1.7869836425781251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,2,power_law_1.01,2.1503141784667967
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,1,2,balanced,2.8047015380859373
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,1,2,balanced,2.4904039001464846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,2,power_law_1.01,2.3170930480957033
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,1,2,balanced,4.241993103027344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,2,power_law_1.01,2.84824462890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,1,2,balanced,5.282374267578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,2,power_law_1.01,3.280176696777344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,2,power_law_1.01,2.3652210998535157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,2,power_law_1.01,2.5857254028320313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,2,power_law_1.01,3.7938714599609376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,2,power_law_1.01,2.6719500732421873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,2,power_law_1.01,4.808622131347656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,2,power_law_1.01,2.8712857055664065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,2,power_law_1.01,3.081793212890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,2,power_law_1.01,0.15809151649475098
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,2,power_law_1.01,5.4888281249999995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,2,power_law_1.01,0.2665190315246582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,2,power_law_1.01,0.34115455627441404
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,2,power_law_1.01,5.796852416992188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,2,power_law_1.01,0.7137203216552734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,2,power_law_1.01,1.080890884399414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,2,power_law_1.01,1.131366424560547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,2,power_law_1.01,1.150804443359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,2,power_law_1.01,7.466024780273438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,2,power_law_1.01,1.217966079711914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,2,power_law_1.01,1.2027161407470703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,2,power_law_1.01,9.527630004882813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,2,power_law_1.01,1.2798258972167968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,2,power_law_1.01,1.3097689819335936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,2,power_law_1.01,1.3554240417480468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,2,power_law_1.01,1.4577714538574218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,2,power_law_1.01,1.5097343444824218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,2,power_law_1.01,1.5637081909179689
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,2,power_law_1.01,1.7028352355957033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,2,power_law_1.01,1.8180134582519532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,2,power_law_1.01,12.798299560546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,2,power_law_1.01,2.132917785644531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,2,power_law_1.01,2.4905215454101564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,2,power_law_1.01,0.5248908615112304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,2,power_law_1.01,1.2606540679931642
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,1,2,power_law_1.01,0.12943103790283203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,2,power_law_1.01,3.061790771484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,2,power_law_1.01,16.100770263671876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,1,2,power_law_1.01,0.19569280624389648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,2,power_law_1.01,3.6286962890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,1,2,power_law_1.01,0.23111551284790038
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,1,2,power_law_1.01,0.33006336212158205
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,1,2,power_law_1.01,0.4533990478515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,1,2,power_law_1.01,0.6989068603515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,1,2,power_law_1.01,0.6755801391601562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,2,power_law_1.01,4.718381958007813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,1,2,power_law_1.01,0.7413337707519532
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,1,2,power_law_1.01,0.758814697265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,1,2,power_law_1.01,0.776710433959961
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,1,2,power_law_1.01,0.7934796905517578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,1,2,power_law_1.01,0.7978112030029296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,2,power_law_1.01,6.014072265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,1,2,power_law_1.01,0.8440857696533204
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,1,2,power_law_1.01,0.8391053009033204
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,1,2,power_law_1.01,0.8678988647460937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,1,2,power_law_1.01,0.882713623046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,2,power_law_1.01,8.408076782226562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,2,power_law_1.01,11.142535400390624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,1,2,power_law_1.01,1.099768295288086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,1,2,power_law_1.01,0.9085311889648438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,1,2,power_law_1.01,1.022339859008789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,1,2,power_law_1.01,1.3785778808593752
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,1,2,power_law_1.01,1.5885017395019532
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,1,2,power_law_1.01,2.023989715576172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,2,power_law_1.2,0.39006591796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,2,power_law_1.2,0.581454086303711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,1,2,power_law_1.01,2.547271728515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,2,power_law_1.2,0.254524154663086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,2,power_law_1.2,1.8669107055664065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,2,power_law_1.2,2.0210687255859376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,2,power_law_1.2,0.861923828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,1,2,power_law_1.01,3.3260632324218746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,2,power_law_1.2,1.12429443359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,2,power_law_1.2,2.088889617919922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,2,power_law_1.2,2.1392767333984373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,2,power_law_1.2,2.1749977111816405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,2,power_law_1.2,2.304706573486328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,2,power_law_1.2,2.337937927246094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,2,power_law_1.2,2.498981170654297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,2,power_law_1.2,2.679129638671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,2,power_law_1.2,2.848596496582031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,2,power_law_1.2,3.2593920898437503
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,2,power_law_1.2,3.315072021484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,2,power_law_1.2,3.8918298339843753
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,2,power_law_1.2,2.857068786621094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,2,power_law_1.2,2.8743206787109377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,2,power_law_1.2,5.046342468261718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,2,power_law_1.2,5.776218872070312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,2,power_law_1.2,6.102594604492188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,2,power_law_1.2,7.532033081054688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,2,power_law_1.2,0.22154367446899412
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,1,2,power_law_1.01,4.4769189453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,2,power_law_1.2,0.3424281692504883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,2,power_law_1.2,9.777719116210937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,2,power_law_1.2,0.4909823989868164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,2,power_law_1.2,0.6658163452148438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,1,2,power_law_1.01,6.1833563232421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,2,power_law_1.2,1.0823168182373046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,2,power_law_1.2,1.1315968322753907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,2,power_law_1.2,1.1576166534423828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,2,power_law_1.2,1.2149696350097656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,2,power_law_1.2,1.221496353149414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,2,power_law_1.2,1.2620172882080078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,1,2,power_law_1.01,8.428159790039063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,2,power_law_1.2,1.3150169372558593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,2,power_law_1.2,1.3600320434570312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,2,power_law_1.2,1.4237414550781249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,2,power_law_1.2,12.999522705078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,2,power_law_1.2,1.4903347778320313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,2,power_law_1.2,1.5798323059082031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,2,power_law_1.2,1.6249984741210937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,2,power_law_1.2,0.15577856063842774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,2,power_law_1.2,1.7643724060058594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,2,power_law_1.2,16.262467041015626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,2,power_law_1.2,1.959228210449219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,2,power_law_1.2,2.2293927001953127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,2,power_law_1.2,2.6752728271484374
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,1,2,power_law_1.2,0.13002880096435548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,2,power_law_1.2,3.1264382934570314
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,1,2,power_law_1.2,0.1716223907470703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,2,power_law_1.2,3.636979064941406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,1,2,power_law_1.2,0.22924160003662108
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,1,2,power_law_1.2,0.32309761047363283
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,1,2,power_law_1.2,0.4030976104736328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,1,2,power_law_1.2,0.6629055786132813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,1,2,power_law_1.2,0.7107698822021484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,2,power_law_1.2,4.944590148925782
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,1,2,power_law_1.2,0.7427532958984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,1,2,power_law_1.2,0.75106689453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,1,2,power_law_1.2,0.7561740875244141
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,1,2,power_law_1.2,0.7992396545410156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,1,2,power_law_1.2,0.8097305297851562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,2,power_law_1.2,6.197230224609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,1,2,power_law_1.2,0.8175782775878906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,1,2,power_law_1.2,0.8770611572265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,1,2,power_law_1.2,0.8900249481201172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,2,power_law_1.2,8.663073120117186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,2,power_law_1.2,11.64583740234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,1,2,power_law_1.2,0.9215462493896485
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,1,2,power_law_1.2,0.990176010131836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,1,2,power_law_1.2,1.2122637176513673
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,1,2,power_law_1.2,1.1232371520996094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,1,2,power_law_1.2,1.8441587829589843
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,1,2,power_law_1.2,1.5163673400878905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,1,balanced,0.1663609504699707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,1,balanced,0.27346879959106446
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,1,2,power_law_1.2,2.4778341674804687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,1,2,power_law_1.2,3.167088623046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,1,balanced,0.4827705764770508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,1,balanced,0.875315170288086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,1,balanced,1.7469459533691407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,1,balanced,2.4369139099121093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,1,balanced,2.4817657470703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,1,balanced,2.4914508056640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,1,balanced,2.5060768127441406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,1,balanced,2.585089416503906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,1,balanced,2.602788391113281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,1,balanced,2.515774688720703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,1,balanced,2.614220275878906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,1,2,power_law_1.2,4.548133239746094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,1,balanced,2.6085714721679687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,1,balanced,2.5915200805664065
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,1,2,power_law_1.2,5.662847900390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,1,balanced,3.0124627685546876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,1,2,power_law_1.2,8.371687622070313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,1,balanced,2.7629498291015624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,1,balanced,3.4545779418945317
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,1,balanced,3.5109298706054686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,1,balanced,3.9543276977539064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,1,balanced,4.8637939453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,1,balanced,3.0584231567382814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,1,balanced,4.948954162597657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,1,balanced,0.111976957321167
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,1,balanced,8.491465454101562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,1,balanced,0.29664512634277346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,1,balanced,9.082364501953126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,1,balanced,0.1773196792602539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,1,balanced,0.5222905731201172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,1,balanced,1.4530738830566405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,1,balanced,1.484691162109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,1,balanced,12.826895751953126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,1,balanced,0.9654783630371094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,1,balanced,1.4817926025390624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,1,balanced,1.4889132690429687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,1,balanced,13.541694335937501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,1,balanced,1.496973419189453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,1,balanced,1.5221784973144532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,1,balanced,1.5659896850585937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,1,balanced,1.5585568237304688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,1,balanced,1.6107020568847656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,1,balanced,1.533399658203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,1,balanced,1.537175750732422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,1,balanced,1.5763725280761718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,1,balanced,1.8923493957519533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,1,balanced,1.9470399475097657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,1,balanced,2.088776245117187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,1,balanced,2.195924530029297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,1,balanced,3.4627981567382813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,1,balanced,3.64520263671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,1,balanced,5.80506103515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,1,balanced,5.266226196289063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,1,balanced,8.812523803710937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,1,balanced,10.767495117187499
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,2,1,balanced,0.13081600189208983
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,2,1,balanced,0.16875648498535156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,2,1,balanced,0.13046208381652832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,2,1,balanced,0.17860671997070313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,2,1,balanced,0.18807296752929686
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,2,1,balanced,0.13159040451049803
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,2,1,balanced,0.20873279571533204
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,2,1,balanced,0.132423677444458
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,2,1,balanced,0.13850048065185547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,2,1,balanced,0.14846015930175782
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,2,1,balanced,0.15808192253112793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,2,1,balanced,0.24847488403320311
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,2,1,balanced,0.30836351394653316
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,2,1,balanced,0.4097695922851562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,2,1,balanced,0.37067840576171873
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,2,1,balanced,0.2687520027160645
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,2,1,balanced,0.5152870559692383
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,2,1,balanced,0.7241766357421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,2,1,balanced,0.9363040161132812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,2,1,balanced,1.3538893127441407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,2,1,balanced,1.7726060485839845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,1,power_law_1.01,0.16531839370727539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,1,power_law_1.01,0.26933631896972654
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,2,1,balanced,2.6100991821289066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,1,power_law_1.01,0.48558593749999995
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,2,1,balanced,3.447877197265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,2,1,balanced,5.136815795898437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,1,power_law_1.01,0.7909606170654297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,1,power_law_1.01,1.957737579345703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,1,power_law_1.01,1.1541478729248047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,1,power_law_1.01,2.044275817871094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,1,power_law_1.01,2.1127174377441404
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,1,power_law_1.01,2.299883575439453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,1,power_law_1.01,1.8110552978515624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,1,power_law_1.01,2.463564758300781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,2,1,balanced,6.807212524414062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,1,power_law_1.01,2.711949462890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,1,power_law_1.01,2.1560499572753904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,1,power_law_1.01,2.9187808227539063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,1,power_law_1.01,3.151036682128906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,1,power_law_1.01,3.3856402587890626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,1,power_law_1.01,2.339698486328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,1,power_law_1.01,3.071381225585937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,1,power_law_1.01,4.1699328613281255
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,1,power_law_1.01,3.752981872558594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,1,power_law_1.01,3.678055114746094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,1,power_law_1.01,5.652951049804687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,1,power_law_1.01,0.11248191833496093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,1,power_law_1.01,0.29218368530273436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,1,power_law_1.01,0.4729868698120117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,1,power_law_1.01,6.377293701171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,1,power_law_1.01,7.7777252197265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,1,power_law_1.01,9.138746948242188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,1,power_law_1.01,0.1767967987060547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,1,power_law_1.01,1.0882367706298828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,1,power_law_1.01,1.171937255859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,1,power_law_1.01,0.6734483337402344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,1,power_law_1.01,1.2434912109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,1,power_law_1.01,1.267080307006836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,1,power_law_1.01,1.2734848022460938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,1,power_law_1.01,1.3254508972167969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,1,power_law_1.01,1.3417344665527344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,1,power_law_1.01,12.151300048828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,1,power_law_1.01,1.3812255859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,1,power_law_1.01,1.4500723266601563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,1,power_law_1.01,1.4603424072265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,1,power_law_1.01,1.517071990966797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,1,power_law_1.01,1.579710693359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,1,power_law_1.01,1.8826272583007815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,1,power_law_1.01,1.9907072448730467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,1,power_law_1.01,14.9145556640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,1,power_law_1.01,2.2575718688964845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,1,power_law_1.01,2.6191937255859377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,1,power_law_1.01,3.177872619628906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,1,power_law_1.01,3.7527667236328126
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,2,1,power_law_1.01,0.13142144203186035
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,2,1,power_law_1.01,0.13128064155578614
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,1,power_law_1.01,4.9716729736328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,2,1,power_law_1.01,0.13167807579040527
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,2,1,power_law_1.01,0.1329817581176758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,1,power_law_1.01,6.11872314453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,2,1,power_law_1.01,0.13821503639221192
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,2,1,power_law_1.01,0.15859711647033692
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,2,1,power_law_1.01,0.14819392204284668
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,2,1,power_law_1.01,0.16909887313842772
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,2,1,power_law_1.01,0.17870016098022462
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,2,1,power_law_1.01,0.1888844871520996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,1,power_law_1.01,8.601287841796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,2,1,power_law_1.01,0.20803007125854492
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,2,1,power_law_1.01,0.30735424041748044
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,2,1,power_law_1.01,0.371391372680664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,1,power_law_1.01,10.877622070312501
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,2,1,power_law_1.01,0.24887935638427736
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,2,1,power_law_1.01,0.2688985633850097
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,2,1,power_law_1.01,0.4115167999267578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,2,1,power_law_1.01,0.5160185623168945
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,2,1,power_law_1.01,0.7246195220947266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,2,1,power_law_1.01,1.7701286315917968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,2,1,power_law_1.01,0.9346157073974609
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,2,1,power_law_1.01,1.3521011352539063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,1,power_law_1.2,0.1657811164855957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,1,power_law_1.2,0.27386495590209964
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,1,power_law_1.2,0.4886291122436523
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,2,1,power_law_1.01,6.808058471679688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,1,power_law_1.2,1.0302054595947265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,1,power_law_1.2,0.7322374725341797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,1,power_law_1.2,1.775194854736328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,1,power_law_1.2,1.9512205505371092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,1,power_law_1.2,2.026980438232422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,1,power_law_1.2,2.1144972229003907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,1,power_law_1.2,2.227774658203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,1,power_law_1.2,2.3501184082031252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,1,power_law_1.2,2.438356475830078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,1,power_law_1.2,2.5798995971679686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,1,power_law_1.2,2.7375173950195313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,2,1,power_law_1.01,2.609819641113281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,1,power_law_1.2,3.2340518188476564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,1,power_law_1.2,3.2040069580078123
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,2,1,power_law_1.01,5.140563354492188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,1,power_law_1.2,2.952298278808594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,2,1,power_law_1.01,3.444197692871094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,1,power_law_1.2,3.509248046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,1,power_law_1.2,3.899980773925781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,1,power_law_1.2,4.255896911621094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,1,power_law_1.2,3.828153686523437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,1,power_law_1.2,5.865548706054687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,1,power_law_1.2,8.132988891601562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,1,power_law_1.2,0.11235520362854004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,1,power_law_1.2,0.17894079208374022
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,1,power_law_1.2,0.2953887939453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,1,power_law_1.2,6.562484130859374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,1,power_law_1.2,0.4381401443481446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,1,power_law_1.2,1.1851360321044921
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,1,power_law_1.2,1.2146514892578124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,1,power_law_1.2,0.6169292831420898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,1,power_law_1.2,9.352541503906249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,1,power_law_1.2,1.2687833404541016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,1,power_law_1.2,1.3009088134765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,1,power_law_1.2,1.3619059753417968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,1,power_law_1.2,1.376934356689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,1,power_law_1.2,1.4042822265625001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,1,power_law_1.2,1.5078668212890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,1,power_law_1.2,1.5221093749999999
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,1,power_law_1.2,1.5657772827148437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,1,power_law_1.2,1.0753874969482422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,1,power_law_1.2,12.291966552734376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,1,power_law_1.2,1.6987135314941404
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,1,power_law_1.2,1.9627033996582033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,1,power_law_1.2,2.095244750976563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,1,power_law_1.2,2.686981201171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,1,power_law_1.2,3.299028625488281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,1,power_law_1.2,2.4233074951171876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,1,power_law_1.2,3.8279290771484376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,1,power_law_1.2,5.025508422851563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,2,1,power_law_1.2,0.13079551696777342
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,2,1,power_law_1.2,0.13100031852722166
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,2,1,power_law_1.2,0.13131584167480467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,1,power_law_1.2,6.158657836914062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,1,power_law_1.2,15.0295068359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,2,1,power_law_1.2,0.13238783836364748
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,2,1,power_law_1.2,0.13817919731140135
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,2,1,power_law_1.2,0.1473516845703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,2,1,power_law_1.2,0.15868608474731444
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,2,1,power_law_1.2,0.16852415084838868
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,2,1,power_law_1.2,0.18741695404052733
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,2,1,power_law_1.2,0.17836671829223633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,2,1,power_law_1.2,0.20773439407348632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,1,power_law_1.2,8.636658935546874
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,2,1,power_law_1.2,0.2688531112670899
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,1,power_law_1.2,10.99706298828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,2,1,power_law_1.2,0.24856639862060548
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,2,1,power_law_1.2,0.3078534317016602
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,2,1,power_law_1.2,0.41035583496093747
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,2,1,power_law_1.2,0.37109439849853515
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,2,1,power_law_1.2,0.9343711853027343
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,2,1,power_law_1.2,0.5158003234863282
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,2,1,power_law_1.2,2.6097747802734377
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,2,1,power_law_1.2,5.142670288085937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,2,1,power_law_1.2,0.7237267303466797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,1,4,balanced,0.1582387161254883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,1,4,balanced,0.2129689598083496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,1,4,balanced,0.3454963302612305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,1,4,balanced,0.3487526321411133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,1,4,balanced,0.34875648498535156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,1,4,balanced,0.3496755218505859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,1,4,balanced,0.35115135192871094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,1,4,balanced,0.352938232421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,1,4,balanced,0.15888511657714843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,1,4,balanced,0.3533350372314453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,1,4,balanced,0.3628492736816406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,1,4,balanced,0.3676505661010742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,2,1,power_law_1.2,1.3531846618652343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,1,4,balanced,0.37607166290283206
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,2,1,power_law_1.2,1.768975372314453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,1,4,balanced,0.38892288208007814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,1,4,balanced,0.4924518585205078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,1,4,balanced,0.4070131301879883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,1,4,balanced,0.5303897476196289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,1,4,balanced,0.7318822479248046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,1,4,balanced,0.8938713836669923
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,2,1,power_law_1.2,3.4493753051757814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,1,4,balanced,1.3296435546874998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,1,4,balanced,0.3584819030761719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,1,4,balanced,1.7219635009765626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,1,4,balanced,0.38902271270751954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,1,4,balanced,2.6437811279296874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,1,4,balanced,0.09357184410095215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,1,4,balanced,0.11921792030334473
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,1,4,balanced,0.1265113639831543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,1,4,balanced,3.8606030273437497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,1,4,balanced,0.12917375564575195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,1,4,balanced,0.1297049617767334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,1,4,balanced,0.1300761604309082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,1,4,balanced,0.13071871757507325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,1,4,balanced,0.1323840045928955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,1,4,balanced,0.1324339199066162
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,1,4,balanced,0.1350374412536621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,1,4,balanced,0.13729408264160156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,1,4,balanced,0.13961215972900393
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,1,4,balanced,6.244071655273437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,1,4,balanced,0.14260095596313477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,1,4,balanced,0.1493887996673584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,1,4,balanced,0.15671680450439454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,1,4,balanced,0.18775039672851562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,1,4,balanced,0.18862976074218749
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,1,4,balanced,0.22918912887573245
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,1,4,balanced,0.2750937652587891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,1,4,balanced,0.3949798583984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,1,4,balanced,0.4853555297851562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,1,4,balanced,0.7126207733154297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,1,4,balanced,0.09045503616333009
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,1,4,balanced,0.9745088195800781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,1,4,balanced,8.715804443359374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,1,4,balanced,1.465665283203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,2,1,power_law_1.2,6.812680053710937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,1,4,balanced,0.1461299228668213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,1,4,balanced,1.9132762145996094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,1,4,balanced,0.04047232151031494
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,1,4,balanced,0.05893248081207275
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,1,4,balanced,0.05885824203491211
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,1,4,balanced,0.05788288116455078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,1,4,balanced,0.05910783767700195
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,1,4,balanced,0.059850239753723146
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,1,4,balanced,0.06005375862121583
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,1,4,balanced,0.06077439785003662
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,1,4,balanced,0.06090752124786377
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,1,4,balanced,0.061387519836425784
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,1,4,balanced,0.06234496116638184
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,1,4,balanced,0.06355455875396729
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,1,4,balanced,0.06018432140350342
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,1,4,balanced,0.040999679565429686
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,1,4,balanced,0.0739302396774292
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,1,4,balanced,0.05940864086151123
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,1,4,balanced,0.059773440361022945
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,1,4,balanced,0.08131839752197265
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,1,4,balanced,0.07892223834991455
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,1,4,balanced,0.11379839897155761
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,1,4,balanced,0.13408512115478516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,1,4,balanced,0.18809728622436522
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,1,4,power_law_1.01,0.16008960723876953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,1,4,power_law_1.01,0.16174463272094725
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,1,4,balanced,0.24263168334960938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,1,4,power_law_1.01,0.3487270355224609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,1,4,power_law_1.01,0.3505369567871094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,1,4,power_law_1.01,0.3601305770874023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,1,4,power_law_1.01,0.3658163070678711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,1,4,power_law_1.01,0.3676697540283203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,1,4,power_law_1.01,0.35181953430175783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,1,4,power_law_1.01,0.37359359741210935
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,1,4,power_law_1.01,0.3105356788635254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,1,4,power_law_1.01,0.39567615509033205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,1,4,power_law_1.01,0.41025150299072266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,1,4,power_law_1.01,0.4553638458251953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,1,4,power_law_1.01,0.47664127349853513
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,1,4,power_law_1.01,0.5111526489257813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,1,4,power_law_1.01,0.5652735900878907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,1,4,power_law_1.01,0.797998046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,1,4,power_law_1.01,0.35464576721191404
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,1,4,power_law_1.01,1.055024642944336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,1,4,power_law_1.01,1.4204121398925782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,1,4,power_law_1.01,1.988371124267578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,1,4,power_law_1.01,3.802051696777344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,1,4,power_law_1.01,2.8901412963867186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,1,4,power_law_1.01,7.103627319335938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,1,4,power_law_1.01,0.09184127807617187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,1,4,power_law_1.01,0.10886143684387206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,1,4,power_law_1.01,0.11627648353576661
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,1,4,power_law_1.01,0.12668031692504883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,1,4,power_law_1.01,0.12928640365600585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,1,4,power_law_1.01,0.12971136093139649
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,1,4,power_law_1.01,0.1318924808502197
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,1,4,power_law_1.01,0.13339648246765137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,1,4,power_law_1.01,0.13570560455322264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,1,4,power_law_1.01,0.13682432174682618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,1,4,power_law_1.01,0.13994879722595216
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,1,4,power_law_1.01,0.14688384056091308
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,1,4,power_law_1.01,0.15452416419982912
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,1,4,power_law_1.01,0.16162048339843751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,1,4,power_law_1.01,0.17201919555664064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,1,4,power_law_1.01,0.1857868766784668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,1,4,power_law_1.01,0.20411775588989256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,1,4,power_law_1.01,0.2350592041015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,1,4,power_law_1.01,0.30559616088867186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,1,4,power_law_1.01,0.3814028930664063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,1,4,power_law_1.01,0.5023731231689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,1,4,power_law_1.01,15.216607666015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,1,4,power_law_1.01,10.316513671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,1,4,balanced,0.3438848114013672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,1,4,power_law_1.01,0.7247309112548829
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,1,4,balanced,0.4539430236816406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,1,4,power_law_1.01,1.9553549194335935
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,1,4,balanced,0.6820697784423828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,1,4,power_law_1.01,0.04775680065155029
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,1,4,power_law_1.01,0.053009920120239265
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,1,4,power_law_1.01,0.05978879928588867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,1,4,power_law_1.01,0.05883903980255127
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,1,4,power_law_1.01,0.0592409610748291
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,1,4,power_law_1.01,0.05933184146881103
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,1,4,power_law_1.01,0.05956992149353028
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,1,4,power_law_1.01,0.0594214391708374
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,1,4,power_law_1.01,0.060002560615539546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,1,4,power_law_1.01,0.06030079841613769
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,1,4,power_law_1.01,0.06302207946777344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,1,4,balanced,0.9309862518310548
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,1,4,power_law_1.01,0.06677631855010986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,1,4,power_law_1.01,0.9520256042480468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,1,4,power_law_1.01,1.4306866455078127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,1,4,power_law_1.01,22.57568359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,1,4,power_law_1.01,3.804104919433594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,1,4,power_law_1.01,4.382970886230469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,1,4,power_law_1.01,0.04152448177337646
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,1,4,power_law_1.01,0.07931519985198975
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,1,4,power_law_1.01,0.07133823871612549
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,1,4,power_law_1.01,0.3865164947509766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,1,4,power_law_1.01,0.0851084804534912
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,1,4,power_law_1.01,0.09704447746276855
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,1,4,power_law_1.01,0.5679334259033203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,1,4,power_law_1.01,0.12595071792602539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,1,4,power_law_1.2,0.1598796844482422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,1,4,power_law_1.2,0.16191360473632813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,1,4,power_law_1.2,0.30081663131713865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,1,4,power_law_1.2,0.3492812728881836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,1,4,power_law_1.2,0.34923137664794923
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,1,4,power_law_1.2,0.3509695816040039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,1,4,power_law_1.2,0.35530113220214843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,1,4,power_law_1.2,0.3605926513671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,1,4,power_law_1.2,0.3655923080444336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,1,4,power_law_1.2,0.3692582321166992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,1,4,power_law_1.2,0.37490177154541016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,1,4,power_law_1.2,0.3975513458251953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,1,4,power_law_1.2,0.41036033630371094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,1,4,power_law_1.2,0.45448833465576166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,1,4,power_law_1.2,0.48672126770019536
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,1,4,power_law_1.2,0.5289830398559571
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,1,4,power_law_1.01,0.15755904197692872
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,1,4,power_law_1.2,0.5630361557006835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,1,4,power_law_1.2,0.8162432098388672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,1,4,power_law_1.01,0.19682304382324217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,1,4,power_law_1.2,1.039758071899414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,1,4,power_law_1.2,1.401318359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,1,4,power_law_1.2,1.9474957275390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,1,4,power_law_1.2,2.9669619750976564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,1,4,power_law_1.2,3.8562661743164064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,1,4,power_law_1.2,7.156375122070313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,1,4,power_law_1.01,0.2748633575439453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,1,4,power_law_1.2,0.08924287796020508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,1,4,power_law_1.2,0.10299903869628906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,1,4,power_law_1.2,0.11086848258972168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,1,4,power_law_1.2,0.12599295616149903
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,1,4,power_law_1.2,0.12796671867370607
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,1,4,power_law_1.2,0.13034879684448242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,1,4,power_law_1.2,0.13193216323852539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,1,4,power_law_1.2,0.13442432403564453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,1,4,power_law_1.2,0.13590271949768068
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,1,4,power_law_1.2,0.13813119888305664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,1,4,power_law_1.2,0.14036992073059082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,1,4,power_law_1.2,0.15285375595092773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,1,4,power_law_1.2,0.15714943885803223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,1,4,power_law_1.2,0.16347007751464843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,1,4,power_law_1.2,0.1792780876159668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,1,4,power_law_1.2,0.1864089584350586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,1,4,power_law_1.2,0.19852800369262696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,1,4,power_law_1.2,0.24417535781860353
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,1,4,power_law_1.2,14.12008056640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,1,4,power_law_1.2,0.30123903274536135
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,1,4,power_law_1.2,0.4056332778930664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,1,4,power_law_1.01,0.791357421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,1,4,power_law_1.2,0.6953472137451172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,1,4,power_law_1.2,10.672489013671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,1,4,power_law_1.2,0.5260492706298828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,1,4,power_law_1.2,1.053521957397461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,1,4,power_law_1.01,1.5236044311523438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,1,4,power_law_1.01,1.1997503662109374
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,1,4,power_law_1.2,0.0624729585647583
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,1,4,power_law_1.2,0.05591423988342285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,1,4,power_law_1.2,0.060083198547363284
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,1,4,power_law_1.2,0.05998079776763916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,1,4,power_law_1.2,0.059900159835815425
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,1,4,power_law_1.2,0.0598796796798706
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,1,4,power_law_1.2,0.06018176078796387
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,1,4,power_law_1.2,0.06025087833404541
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,1,4,power_law_1.2,0.06070911884307861
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,1,4,power_law_1.2,0.06122879981994629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,1,4,power_law_1.2,3.5136767578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,1,4,power_law_1.01,2.151068115234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,1,4,power_law_1.2,0.07564799785614014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,1,4,power_law_1.2,1.734291229248047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,1,4,power_law_1.2,0.08693887710571288
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,1,4,power_law_1.2,0.08327679634094239
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,1,4,power_law_1.2,24.014108886718752
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,1,4,power_law_1.2,0.1506163215637207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,1,4,power_law_1.2,2.349623107910156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,1,4,power_law_1.2,0.19961599349975584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,1,4,power_law_1.2,3.98877197265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,1,4,power_law_1.2,0.041797118186950685
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,1,4,power_law_1.2,0.06548992156982422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,1,4,power_law_1.2,0.06876800060272217
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,1,4,power_law_1.2,0.1043289566040039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,1,4,power_law_1.2,1.0662258911132814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,2,2,balanced,0.10827903747558594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,2,2,balanced,0.1299238395690918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,2,2,balanced,0.17740032196044922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,2,2,balanced,0.2488972854614258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,2,2,balanced,0.24550783157348635
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,2,2,balanced,0.25340160369873044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,2,2,balanced,0.25854591369628904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,2,2,balanced,0.25971839904785154
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,1,4,power_law_1.2,0.12929408073425291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,2,2,balanced,0.2626380729675293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,2,2,balanced,0.2646912002563476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,2,2,balanced,0.26611328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,2,2,balanced,0.2732134437561035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,2,2,balanced,0.2757913589477539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,2,2,balanced,0.28305152893066404
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,2,2,balanced,0.3040102386474609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,2,2,balanced,0.3074188804626465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,2,2,balanced,0.29951103210449215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,2,2,balanced,0.39803009033203124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,2,2,balanced,0.5032563018798828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,2,2,balanced,0.6858252716064454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,2,2,balanced,0.8461478424072265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,2,2,balanced,1.3280383300781249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,2,2,balanced,1.7448562622070312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,2,2,balanced,2.587310180664063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,2,2,balanced,3.663084716796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,2,2,balanced,5.924807739257813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,2,2,balanced,0.05220479965209961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,2,2,balanced,0.06925055980682374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,2,2,balanced,0.09637632369995117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,2,2,balanced,0.10186367988586427
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,2,2,balanced,0.10286591529846192
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,2,2,balanced,0.1045081615447998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,2,2,balanced,0.10758272171020508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,2,2,balanced,0.10752384185791015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,2,2,balanced,0.1098419189453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,2,2,balanced,0.1106175994873047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,2,2,balanced,0.11270527839660645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,2,2,balanced,0.11450240135192871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,2,2,balanced,0.11678336143493653
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,2,2,balanced,0.12310400009155273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,2,2,balanced,0.12744959831237793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,2,2,balanced,0.1323622417449951
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,2,2,balanced,8.56247314453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,2,2,balanced,0.16277120590209962
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,1,4,power_law_1.2,0.278337287902832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,2,2,balanced,0.17641855239868165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,2,2,balanced,0.23740928649902343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,2,2,balanced,0.284399356842041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,2,2,balanced,0.1283520030975342
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,2,2,balanced,0.40971904754638666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,2,2,balanced,0.534727668762207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,2,2,balanced,0.7762035369873047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,1,4,power_law_1.2,0.47245441436767577
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,2,2,balanced,1.0352793884277343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,2,2,balanced,1.608179168701172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,2,2,balanced,2.018871002197266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,1,4,power_law_1.2,0.501351661682129
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,1,4,power_law_1.2,2.18281982421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,1,4,power_law_1.2,0.7605043029785156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,2,2,balanced,0.023111679553985597
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,2,2,balanced,0.05623040199279785
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,2,2,balanced,0.0561510419845581
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,2,2,balanced,0.056732158660888675
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,2,2,balanced,0.056720638275146486
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,2,2,balanced,0.05623167991638184
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,2,2,balanced,0.05738880157470703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,2,2,balanced,0.057657599449157715
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,2,2,balanced,0.057580800056457514
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,2,2,balanced,0.03551487922668457
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,2,2,balanced,0.057873921394348146
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,2,2,balanced,0.06080128192901611
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,2,2,balanced,0.06074880123138428
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,2,2,balanced,0.05600255966186524
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,2,2,balanced,0.056144638061523436
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,2,2,balanced,0.05705215930938721
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,1,4,power_law_1.2,1.7301158142089843
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,2,2,balanced,0.05724927902221679
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,2,2,balanced,0.07847167968749999
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,2,2,balanced,0.10421631813049317
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,2,2,balanced,0.08062463760375976
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,2,2,balanced,0.12971136093139649
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,2,2,power_law_1.01,0.11129088401794433
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,2,2,power_law_1.01,0.10761856079101562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,2,2,power_law_1.01,0.243374080657959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,2,2,power_law_1.01,0.2340275192260742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,2,2,power_law_1.01,0.2451456069946289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,2,2,power_law_1.01,0.25671295166015623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,2,2,power_law_1.01,0.09797760009765624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,2,2,power_law_1.01,0.2615756797790527
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,2,2,power_law_1.01,0.2643622398376465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,2,2,power_law_1.01,0.2663475227355957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,2,2,power_law_1.01,0.2688960075378418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,2,2,power_law_1.01,0.2734886360168457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,2,2,power_law_1.01,0.28139904022216794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,2,2,power_law_1.01,0.30310527801513676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,2,2,power_law_1.01,0.32320255279541016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,2,2,power_law_1.01,0.35959678649902344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,2,2,power_law_1.01,0.41361408233642577
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,2,2,power_law_1.01,0.49670398712158204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,2,2,power_law_1.01,0.6110489654541016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,2,2,power_law_1.01,0.9639692687988282
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,2,2,balanced,0.18180095672607421
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,2,2,balanced,0.2344371223449707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,2,2,power_law_1.01,1.0522470092773437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,2,2,power_law_1.01,1.8049714660644534
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,2,2,balanced,0.33407871246337895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,2,2,power_law_1.01,2.534324493408203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,2,2,power_law_1.01,0.06278528213500977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,2,2,power_law_1.01,0.0708684778213501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,2,2,power_law_1.01,0.0806783962249756
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,2,2,power_law_1.01,0.10135168075561522
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,2,2,power_law_1.01,0.10116479873657228
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,2,2,power_law_1.01,0.1024345588684082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,2,2,power_law_1.01,0.10515328407287598
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,2,2,power_law_1.01,0.10687487602233887
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,2,2,balanced,0.430621452331543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,2,2,power_law_1.01,0.10616448402404785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,2,2,power_law_1.01,0.10772735595703124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,2,2,power_law_1.01,0.11172863960266113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,2,2,power_law_1.01,3.9747250366210936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,2,2,power_law_1.01,0.11721471786499023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,2,2,power_law_1.01,0.11737728118896484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,2,2,power_law_1.01,0.12684927940368654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,2,2,power_law_1.01,0.14399999618530274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,2,2,power_law_1.01,0.13164544105529785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,2,2,power_law_1.01,0.15662464141845703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,2,2,power_law_1.01,0.20956544876098632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,2,2,power_law_1.01,0.1736524772644043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,2,2,power_law_1.01,0.46048126220703123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,2,2,power_law_1.01,0.35486080169677736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,2,2,power_law_1.01,0.26114559173583984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,2,2,power_law_1.01,7.708919067382813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,2,2,power_law_1.01,0.5647667312622071
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,2,2,power_law_1.01,5.628840942382813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,2,2,power_law_1.01,0.034612479209899905
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,2,2,power_law_1.01,0.04311168193817139
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,2,2,power_law_1.01,0.042120962142944335
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,2,2,power_law_1.01,0.0552729606628418
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,2,2,power_law_1.01,0.0537830400466919
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,2,2,power_law_1.01,0.0559987211227417
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,2,2,power_law_1.01,0.055743999481201165
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,2,2,balanced,0.657096939086914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,2,2,power_law_1.01,0.05576064109802246
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,2,2,power_law_1.01,0.06204031944274903
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,2,2,power_law_1.01,1.952290496826172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,2,2,power_law_1.01,0.0564083194732666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,2,2,balanced,0.872833251953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,2,2,power_law_1.01,0.05991680145263671
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,2,2,power_law_1.01,0.06440576076507568
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,2,2,power_law_1.01,1.0090035247802736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,2,2,power_law_1.01,1.2596915435791014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,2,2,power_law_1.01,2.6057394409179686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,2,2,power_law_1.01,9.891900024414062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,2,2,power_law_1.01,0.16296192169189455
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,2,2,power_law_1.01,0.07142399787902833
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,2,2,power_law_1.01,0.07336448192596436
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,2,2,power_law_1.01,0.06099071979522706
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,2,2,power_law_1.01,0.28400127410888676
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,2,2,power_law_1.01,0.06569471836090088
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,2,2,power_law_1.01,0.08005760192871095
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,2,2,power_law_1.01,0.09298303604125976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,2,2,power_law_1.2,0.10224512100219726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,2,2,power_law_1.2,0.10330112457275391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,2,2,power_law_1.2,0.10739968299865724
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,2,2,power_law_1.2,0.24569728851318356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,2,2,power_law_1.2,0.2411814308166504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,2,2,power_law_1.2,0.24809087753295897
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,2,2,power_law_1.2,0.25780736923217773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,2,2,power_law_1.2,0.26019968032836915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,2,2,power_law_1.2,0.26312320709228515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,2,2,power_law_1.2,0.2630873680114746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,2,2,power_law_1.2,0.26649728775024417
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,2,2,power_law_1.01,0.10684800148010254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,2,2,power_law_1.2,0.273628158569336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,2,2,power_law_1.2,0.27888511657714843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,2,2,power_law_1.2,0.30045568466186523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,2,2,power_law_1.2,0.33349887847900395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,2,2,power_law_1.2,0.3578112030029297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,2,2,power_law_1.2,0.41317760467529296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,2,2,power_law_1.2,0.5147878265380859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,2,2,power_law_1.2,0.6156390380859376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,2,2,power_law_1.2,1.0058201599121095
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,2,2,power_law_1.01,0.18509311676025392
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,2,2,power_law_1.2,1.067430419921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,2,2,power_law_1.2,1.7551104736328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,2,2,power_law_1.2,2.3735731506347655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,2,2,power_law_1.2,3.318233642578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,2,2,power_law_1.01,1.0725708770751954
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,2,2,power_law_1.01,0.34281471252441403
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,2,2,power_law_1.2,0.060707840919494625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,2,2,power_law_1.2,0.07721343994140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,2,2,power_law_1.2,0.07663104057312012
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,2,2,power_law_1.2,0.10105728149414062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,2,2,power_law_1.2,0.10429311752319337
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,2,2,power_law_1.2,0.10322688102722169
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,2,2,power_law_1.2,0.10737279891967773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,2,2,power_law_1.2,0.10691712379455567
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,2,2,power_law_1.2,0.10832639694213868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,2,2,power_law_1.2,0.11043583869934082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,2,2,power_law_1.2,0.11305855751037597
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,2,2,power_law_1.2,0.11494144439697265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,2,2,power_law_1.2,0.11868672370910645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,2,2,power_law_1.2,0.12326656341552736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,2,2,power_law_1.2,0.13737088203430176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,2,2,power_law_1.2,0.14369152069091798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,2,2,power_law_1.2,0.14614399909973144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,2,2,power_law_1.2,0.17800575256347656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,2,2,power_law_1.2,0.21335935592651367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,2,2,power_law_1.2,0.25922048568725586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,2,2,power_law_1.2,7.613817749023437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,2,2,power_law_1.01,0.523370246887207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,2,2,power_law_1.2,0.48062976837158206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,2,2,power_law_1.2,0.36155391693115235
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,2,2,power_law_1.01,0.7388467407226562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,2,2,power_law_1.2,0.6659839630126954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,2,2,power_law_1.2,5.348866577148438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,2,2,power_law_1.2,0.03312256097793579
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,2,2,power_law_1.2,0.04294655799865722
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,2,2,power_law_1.2,0.056062722206115724
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,2,2,power_law_1.2,0.05629695892333984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,2,2,power_law_1.2,0.05646975994110107
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,2,2,power_law_1.2,0.05685760021209717
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,2,2,power_law_1.2,0.056605439186096194
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,2,2,power_law_1.2,0.05707263946533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,2,2,power_law_1.2,0.057194237709045415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,2,2,power_law_1.2,0.9971123504638673
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,2,2,power_law_1.2,0.057944321632385255
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,2,2,power_law_1.2,0.06357759952545165
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,2,2,power_law_1.2,0.06653312206268311
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,2,2,power_law_1.2,2.126929931640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,2,2,power_law_1.01,1.31359619140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,2,2,power_law_1.2,0.08005887985229491
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,2,2,power_law_1.2,1.2715916442871094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,2,2,power_law_1.2,0.03048959970474243
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,2,2,power_law_1.2,0.14302847862243653
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,2,2,power_law_1.2,0.1951091194152832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,2,2,power_law_1.2,0.06525055885314941
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,2,2,power_law_1.2,0.07027455806732177
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,2,2,power_law_1.2,11.596953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,2,2,power_law_1.2,0.07251840114593507
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,2,2,power_law_1.2,2.5366246032714845
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,2,2,power_law_1.2,0.09454848289489745
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,4,1,balanced,0.07987648010253906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,4,1,balanced,0.10042495727539062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,4,1,balanced,0.16176063537597657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,2,2,power_law_1.2,0.11201151847839355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,4,1,balanced,0.16238271713256835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,4,1,balanced,0.17462335586547853
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,4,1,balanced,0.1756870460510254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,4,1,balanced,0.17690624237060545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,4,1,balanced,0.1780064010620117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,4,1,balanced,0.17824319839477537
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,4,1,balanced,0.17744768142700196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,4,1,balanced,0.1800377655029297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,4,1,balanced,0.18105535507202147
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,4,1,balanced,0.18351680755615235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,4,1,balanced,0.1896454429626465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,4,1,balanced,0.24696384429931642
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,4,1,balanced,0.25344383239746093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,4,1,balanced,0.26080320358276365
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,4,1,balanced,0.3660927963256836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,4,1,balanced,0.4440723037719726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,4,1,balanced,0.6345651245117188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,4,1,balanced,0.8143859100341796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,4,1,balanced,1.2351385498046876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,4,1,balanced,1.6469715881347657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,4,1,balanced,2.474298858642578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,2,2,power_law_1.2,0.48942207336425786
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,4,1,balanced,3.2960818481445315
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,4,1,balanced,0.04947455883026123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,4,1,balanced,0.06582848072052003
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,4,1,balanced,0.09287424087524414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,4,1,balanced,4.836223449707031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,4,1,balanced,0.10255231857299804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,4,1,balanced,0.10403327941894532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,4,1,balanced,0.1046720027923584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,4,1,balanced,0.10243328094482422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,4,1,balanced,0.10402367591857911
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,4,1,balanced,0.10386240005493164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,4,1,balanced,0.10579392433166504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,4,1,balanced,0.10622912406921388
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,4,1,balanced,0.10893376350402832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,4,1,balanced,0.11110464096069336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,4,1,balanced,0.11586943626403809
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,4,1,balanced,0.1188441562652588
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,4,1,balanced,6.654615478515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,4,1,balanced,0.11839103698730469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,4,1,balanced,0.09227583885192872
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,4,1,balanced,0.1667206382751465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,4,1,balanced,0.21505727767944335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,4,1,balanced,0.2731180763244629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,4,1,balanced,0.16516159057617188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,4,1,balanced,0.3842176055908203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,4,1,balanced,0.4761248016357422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,4,1,balanced,0.7038150024414063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,4,1,balanced,0.9516460418701171
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,4,1,balanced,1.4194796752929686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,4,1,balanced,1.8906521606445312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,2,2,power_law_1.2,0.2575116729736328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,2,2,power_law_1.2,0.3174489593505859
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,4,1,balanced,0.022228479385375977
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,4,1,balanced,0.05821568012237549
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,2,2,power_law_1.2,0.7637042999267578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,4,1,balanced,0.035676159858703614
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,4,1,balanced,0.05737664222717285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,4,1,balanced,0.057578239440917965
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,4,1,balanced,0.05751359939575196
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,4,1,balanced,0.0577791976928711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,4,1,balanced,0.05743360042572021
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,4,1,balanced,0.05849855899810791
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,4,1,balanced,0.057670397758483885
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,4,1,balanced,0.06151936054229736
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,4,1,balanced,0.057031679153442386
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,4,1,balanced,0.061331839561462397
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,4,1,balanced,0.06260543823242187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,2,2,power_law_1.2,1.0540428924560548
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,4,1,balanced,0.057343997955322266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,2,2,power_law_1.2,1.3906343078613281
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,4,1,balanced,0.057056641578674315
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,4,1,balanced,0.05731840133666992
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,4,1,balanced,0.08472000122070313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,4,1,balanced,0.10939519882202149
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,4,1,balanced,0.08059264183044433
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,4,1,power_law_1.01,0.07871808052062988
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,4,1,power_law_1.01,0.0966431999206543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,4,1,balanced,0.13147263526916503
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,4,1,power_law_1.01,0.10964544296264649
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,4,1,power_law_1.01,0.1551212787628174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,4,1,power_law_1.01,0.1760972785949707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,4,1,power_law_1.01,0.1792857551574707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,4,1,power_law_1.01,0.17796863555908202
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,4,1,power_law_1.01,0.17824703216552734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,4,1,power_law_1.01,0.1778374481201172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,4,1,power_law_1.01,0.19051647186279297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,4,1,power_law_1.01,0.22439744949340817
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,4,1,power_law_1.01,0.22699008941650392
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,4,1,power_law_1.01,0.23625791549682615
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,4,1,power_law_1.01,0.24785087585449217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,4,1,power_law_1.01,0.25549247741699216
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,4,1,power_law_1.01,0.2958009529113769
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,4,1,power_law_1.01,0.34315521240234376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,4,1,power_law_1.01,0.4219392013549805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,4,1,power_law_1.01,0.5027507019042969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,4,1,balanced,0.1896236801147461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,4,1,power_law_1.01,0.6766368103027344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,4,1,power_law_1.01,0.8664646148681641
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,4,1,balanced,0.24385471343994142
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,4,1,balanced,0.35571521759033203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,4,1,power_law_1.01,1.3006732177734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,4,1,power_law_1.01,1.6916249084472657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,4,1,power_law_1.01,0.05444096088409424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,4,1,power_law_1.01,0.06388800144195557
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,4,1,power_law_1.01,0.06898816108703613
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,4,1,power_law_1.01,0.09242367744445801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,4,1,power_law_1.01,0.09959744453430176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,4,1,power_law_1.01,0.10107968330383302
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,4,1,power_law_1.01,0.10081791877746582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,4,1,power_law_1.01,0.10094207763671874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,4,1,power_law_1.01,0.1012435245513916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,4,1,power_law_1.01,0.10348735809326173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,4,1,power_law_1.01,0.10437055587768555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,4,1,power_law_1.01,0.10534144401550294
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,4,1,power_law_1.01,0.1086508846282959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,4,1,power_law_1.01,0.11852160453796387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,4,1,power_law_1.01,2.5036927795410158
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,4,1,power_law_1.01,0.12535296440124513
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,4,1,power_law_1.01,0.13757887840270996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,4,1,power_law_1.01,0.13663552284240724
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,4,1,power_law_1.01,0.1608051109313965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,4,1,power_law_1.01,0.18802816390991212
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,4,1,power_law_1.01,0.24282432556152345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,4,1,power_law_1.01,3.299032897949219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,4,1,balanced,0.46483711242675785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,4,1,power_law_1.01,0.2890425682067871
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,4,1,balanced,0.7138668823242187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,4,1,power_law_1.01,0.39335872650146486
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,4,1,balanced,0.9985369873046874
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,4,1,power_law_1.01,0.029568641185760497
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,4,1,power_law_1.01,0.037703039646148684
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,4,1,power_law_1.01,0.04223423957824707
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,4,1,power_law_1.01,0.056340479850769044
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,4,1,power_law_1.01,0.05751296043395996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,4,1,power_law_1.01,0.05784832000732422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,4,1,power_law_1.01,0.057960958480834955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,4,1,power_law_1.01,0.5033254241943359
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,4,1,power_law_1.01,0.05669760227203369
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,4,1,power_law_1.01,0.058094720840454105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,4,1,power_law_1.01,0.7304544067382812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,4,1,power_law_1.01,0.05915391921997071
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,4,1,power_law_1.01,4.964876098632812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,4,1,power_law_1.01,0.9754796600341796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,4,1,power_law_1.01,6.656479492187501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,4,1,power_law_1.01,1.4491078186035158
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,4,1,power_law_1.01,1.9291404724121093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,4,1,power_law_1.01,0.05682752132415771
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,4,1,power_law_1.01,0.1051308822631836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,4,1,power_law_1.01,0.05830463886260986
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,4,1,power_law_1.01,0.05899519920349121
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,4,1,power_law_1.01,0.06194623947143555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,4,1,power_law_1.01,0.06421120166778564
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,4,1,power_law_1.01,0.06756671905517578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,4,1,power_law_1.01,0.07322944164276123
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,4,1,power_law_1.01,0.08948351860046386
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,4,1,power_law_1.01,0.13538432121276855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,4,1,power_law_1.2,0.07896768093109131
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,4,1,power_law_1.2,0.09473535537719727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,4,1,power_law_1.2,0.10336447715759278
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,4,1,power_law_1.01,0.16746368408203124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,4,1,power_law_1.2,0.1702751922607422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,4,1,power_law_1.2,0.1749715232849121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,4,1,power_law_1.2,0.17583040237426756
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,4,1,power_law_1.01,0.23201152801513675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,4,1,power_law_1.2,0.1756377601623535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,4,1,power_law_1.2,0.17875776290893555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,4,1,power_law_1.2,0.18749439239501953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,4,1,power_law_1.2,0.22476671218872069
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,4,1,power_law_1.2,0.22890047073364256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,4,1,power_law_1.2,0.1583135986328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,4,1,power_law_1.2,0.23574783325195314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,4,1,power_law_1.2,0.24881023406982422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,4,1,power_law_1.2,0.25767616271972654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,4,1,power_law_1.2,0.3041062355041504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,4,1,power_law_1.2,0.34459457397460935
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,4,1,power_law_1.2,0.42146175384521484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,4,1,power_law_1.01,0.29627071380615233
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,4,1,power_law_1.2,0.5076959991455079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,4,1,power_law_1.2,0.6794445037841796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,4,1,power_law_1.2,0.8954783630371093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,4,1,power_law_1.01,0.42689662933349604
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,4,1,power_law_1.2,1.2441670227050783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,4,1,power_law_1.2,1.6824441528320313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,4,1,power_law_1.01,0.5617407989501954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,4,1,power_law_1.2,0.06738239765167237
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,4,1,power_law_1.2,0.06677760124206543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,4,1,power_law_1.2,0.0929958438873291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,4,1,power_law_1.2,0.096429443359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,4,1,power_law_1.2,0.10076352119445801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,4,1,power_law_1.2,0.10110912322998047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,4,1,power_law_1.2,0.10161664009094237
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,4,1,power_law_1.2,0.10308480262756348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,4,1,power_law_1.2,0.10377087593078613
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,4,1,power_law_1.2,0.10594688415527345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,4,1,power_law_1.2,0.10714176177978516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,4,1,power_law_1.2,0.10727359771728515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,4,1,power_law_1.2,0.11980416297912597
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,4,1,power_law_1.2,0.05195648193359374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,4,1,power_law_1.2,0.12924160003662108
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,4,1,power_law_1.2,0.13898176193237305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,4,1,power_law_1.2,0.13735808372497557
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,4,1,power_law_1.2,2.426144714355469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,4,1,power_law_1.2,0.159104642868042
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,4,1,power_law_1.01,0.8386259460449219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,4,1,power_law_1.2,0.18789375305175782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,4,1,power_law_1.2,0.23822975158691406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,4,1,power_law_1.2,0.2884876823425293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,4,1,power_law_1.2,0.3941196823120117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,4,1,power_law_1.2,3.276402587890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,4,1,power_law_1.2,0.5007059097290039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,4,1,power_law_1.2,0.027075839042663575
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,4,1,power_law_1.2,0.03581248044967651
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,4,1,power_law_1.2,0.03829119920730591
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,4,1,power_law_1.2,0.05547135829925537
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,4,1,power_law_1.2,0.054930558204650884
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,4,1,power_law_1.2,0.05684288024902344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,4,1,power_law_1.2,0.05734720230102539
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,4,1,power_law_1.2,0.05618368148803711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,4,1,power_law_1.2,0.057392001152038574
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,4,1,power_law_1.2,0.05707071781158447
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,4,1,power_law_1.2,0.057956480979919435
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,4,1,power_law_1.01,1.1418905639648438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,4,1,power_law_1.2,0.05969535827636718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,4,1,power_law_1.2,0.7240646362304688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,4,1,power_law_1.2,0.06245632171630859
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,4,1,power_law_1.2,0.0666918420791626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,4,1,power_law_1.2,4.879805297851563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,4,1,power_law_1.2,0.9783987426757813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,4,1,power_law_1.2,0.0586303997039795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,4,1,power_law_1.2,1.4394931030273437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,4,1,power_law_1.2,0.06964352130889892
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,4,1,power_law_1.2,0.07350143909454346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,4,1,power_law_1.2,6.569249267578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,4,1,power_law_1.2,0.09046079635620116
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,4,1,power_law_1.2,0.10676992416381836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,4,1,power_law_1.2,0.3034662437438965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,4,1,power_law_1.2,1.942729034423828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,1,4,balanced,0.25169408798217774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,1,4,balanced,0.24542591094970706
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,1,4,balanced,0.32325759887695316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,1,4,balanced,0.47014526367187504
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,4,1,power_law_1.2,0.1390790367126465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,1,4,balanced,0.471459846496582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,1,4,balanced,0.4718131256103516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,1,4,balanced,0.47581951141357426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,1,4,balanced,0.47645824432373046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,1,4,balanced,0.4805491256713867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,1,4,balanced,0.48086910247802733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,1,4,balanced,0.48368000030517583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,1,4,balanced,0.49128959655761717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,1,4,balanced,0.4946598434448243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,1,4,balanced,0.5040166473388672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,1,4,balanced,0.5544780731201172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,1,4,balanced,0.5537279891967773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,1,4,balanced,0.5675059127807617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,1,4,balanced,0.8740377807617188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,1,4,balanced,0.764843521118164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,1,4,balanced,1.2730508422851563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,1,4,balanced,1.4849203491210938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,4,1,power_law_1.2,0.17097280502319337
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,1,4,balanced,2.2759129333496095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,1,4,balanced,2.978726501464844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,4,1,power_law_1.2,0.2408940887451172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,1,4,balanced,4.488241882324219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,1,4,balanced,0.1142848014831543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,1,4,balanced,0.1132748794555664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,1,4,balanced,0.16319488525390624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,1,4,balanced,0.17139711380004882
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,1,4,balanced,0.17240447998046876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,1,4,balanced,0.1733171272277832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,1,4,balanced,0.1749350357055664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,1,4,balanced,0.17518463134765624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,1,4,balanced,0.17589376449584962
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,1,4,balanced,0.1765068817138672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,1,4,balanced,0.17854591369628908
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,1,4,balanced,0.18084352493286132
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,1,4,balanced,6.573804931640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,1,4,balanced,0.18142976760864257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,1,4,balanced,0.1903718376159668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,1,4,balanced,0.18869632720947266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,1,4,balanced,0.1930598449707031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,1,4,balanced,0.2032307243347168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,1,4,balanced,0.251461124420166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,1,4,balanced,0.253436164855957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,1,4,balanced,0.31504896163940427
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,1,4,balanced,0.4124492645263672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,1,4,balanced,0.542050552368164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,1,4,balanced,0.7026956939697266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,1,4,balanced,1.1339558410644532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,1,4,balanced,1.4552268981933594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,1,4,balanced,10.753420410156249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,1,4,balanced,2.2460147094726564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,1,4,balanced,2.9459405517578126
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,4,1,power_law_1.2,0.44055297851562497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,1,4,balanced,14.6787451171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,4,1,power_law_1.2,0.575175666809082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,4,1,power_law_1.2,0.8475955200195312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,4,1,power_law_1.2,1.1613542175292968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,1,4,balanced,0.05446400165557861
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,1,4,balanced,0.08008831977844239
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,1,4,balanced,0.0796992015838623
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,1,4,balanced,0.05549952030181885
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,1,4,balanced,0.08081151962280272
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,1,4,balanced,0.08163968086242676
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,1,4,balanced,0.08107775688171387
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,1,4,balanced,0.08219136238098144
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,1,4,balanced,0.08222847938537597
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,1,4,balanced,0.08384896278381347
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,1,4,balanced,0.08424192428588867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,1,4,balanced,0.07892352104187011
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,1,4,balanced,0.07952640056610108
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,1,4,balanced,0.07906816005706788
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,1,4,balanced,0.07917952060699462
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,1,4,balanced,0.07896575927734376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,1,4,balanced,0.07871103763580323
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,1,4,balanced,0.13303423881530763
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,1,4,balanced,0.13544960021972657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,1,4,balanced,0.17428224563598632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,1,4,power_law_1.01,0.2559040069580078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,1,4,power_law_1.01,0.2891622352600097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,1,4,power_law_1.01,0.4283660888671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,1,4,power_law_1.01,0.46799873352050786
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,1,4,power_law_1.01,0.46981632232666015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,1,4,balanced,0.23368192672729493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,1,4,power_law_1.01,0.4742451095581055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,1,4,power_law_1.01,0.48294910430908206
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,1,4,balanced,0.3238438415527344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,1,4,power_law_1.01,0.4860940933227539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,1,4,power_law_1.01,0.4893247985839844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,1,4,power_law_1.01,0.5070809555053711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,1,4,power_law_1.01,0.49946624755859376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,1,4,power_law_1.01,0.5367782211303711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,1,4,power_law_1.01,0.5615347290039063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,1,4,power_law_1.01,0.6441305541992188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,1,4,power_law_1.01,0.7023744201660156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,1,4,power_law_1.01,0.8056960296630858
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,1,4,power_law_1.01,0.9366194915771484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,1,4,power_law_1.01,1.4079335021972657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,1,4,power_law_1.01,1.8081459045410155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,1,4,power_law_1.01,2.396506805419922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,1,4,power_law_1.01,3.5068722534179684
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,1,4,balanced,0.4235059356689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,1,4,power_law_1.01,4.662787780761719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,1,4,power_law_1.01,10.709775390625001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,1,4,power_law_1.01,7.5725732421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,1,4,power_law_1.01,0.11699968338012696
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,1,4,balanced,0.6191769790649414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,1,4,power_law_1.01,0.1324492835998535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,1,4,power_law_1.01,0.1692927932739258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,1,4,power_law_1.01,0.17127552032470703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,1,4,power_law_1.01,0.17268224716186525
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,1,4,power_law_1.01,0.17488895416259767
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,1,4,power_law_1.01,0.17577856063842773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,1,4,power_law_1.01,0.17723775863647462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,1,4,power_law_1.01,0.1793062400817871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,1,4,power_law_1.01,0.18167808532714844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,1,4,power_law_1.01,0.19457920074462892
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,1,4,power_law_1.01,0.1976460838317871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,1,4,power_law_1.01,0.21221120834350585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,1,4,power_law_1.01,0.2288012886047363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,1,4,power_law_1.01,0.24247680664062501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,1,4,power_law_1.01,0.27232128143310547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,1,4,power_law_1.01,0.33854209899902343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,1,4,power_law_1.01,0.42965633392333985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,1,4,power_law_1.01,0.14317952156066896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,1,4,power_law_1.01,0.5963571166992188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,1,4,power_law_1.01,0.7265459442138672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,1,4,power_law_1.01,1.090271987915039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,1,4,balanced,0.7923519897460938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,1,4,power_law_1.01,20.2014697265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,1,4,power_law_1.01,2.306712341308594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,1,4,power_law_1.01,1.416661834716797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,1,4,power_law_1.01,0.06254591941833496
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,1,4,power_law_1.01,0.06692480087280274
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,1,4,power_law_1.01,0.07463551998138428
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,1,4,power_law_1.01,0.07899392127990723
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,1,4,power_law_1.01,0.07893887996673585
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,1,4,power_law_1.01,0.07935359954833984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,1,4,power_law_1.01,0.08035712242126465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,1,4,power_law_1.01,26.505417480468747
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,1,4,power_law_1.01,0.08025343894958496
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,1,4,power_law_1.01,0.08188799858093262
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,1,4,power_law_1.01,0.08152064323425293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,1,4,power_law_1.01,0.08264448165893554
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,1,4,balanced,1.2297714996337892
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,1,4,power_law_1.01,0.12434687614440917
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,1,4,balanced,1.6638220214843749
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,1,4,power_law_1.01,4.982789001464844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,1,4,power_law_1.01,3.0063372802734376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,1,4,power_law_1.01,6.0030755615234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,1,4,power_law_1.01,33.69536865234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,1,4,power_law_1.01,0.09843328475952148
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,1,4,power_law_1.01,0.12690815925598145
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,1,4,power_law_1.01,0.13918975830078123
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,1,4,power_law_1.01,0.0948031997680664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,1,4,power_law_1.01,0.6452249908447266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,1,4,power_law_1.01,0.1977459144592285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,1,4,power_law_1.01,0.16119424819946288
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,1,4,power_law_1.01,0.26826879501342776
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,1,4,power_law_1.2,0.2566771125793457
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,1,4,power_law_1.01,0.8189389038085938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,1,4,power_law_1.2,0.44006526947021485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,1,4,power_law_1.2,0.4724582290649414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,1,4,power_law_1.2,0.4720959854125977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,1,4,power_law_1.2,0.47764480590820313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,1,4,power_law_1.2,0.4829286575317383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,1,4,power_law_1.2,0.2702079963684082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,1,4,power_law_1.2,0.4888358306884766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,1,4,power_law_1.2,0.4886783981323243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,1,4,power_law_1.2,0.4920703887939453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,1,4,power_law_1.2,0.49937408447265624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,1,4,power_law_1.2,0.5450022506713867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,1,4,power_law_1.2,0.5575513458251953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,1,4,power_law_1.2,0.674777603149414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,1,4,power_law_1.2,0.7508544158935547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,1,4,power_law_1.2,0.8202572631835938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,1,4,power_law_1.2,0.9613683319091797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,1,4,power_law_1.2,1.3792807006835939
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,1,4,power_law_1.2,1.6392063903808594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,1,4,power_law_1.01,0.3751424026489258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,1,4,power_law_1.2,2.552238006591797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,1,4,power_law_1.2,3.070513916015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,1,4,power_law_1.2,5.926968383789062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,1,4,power_law_1.2,6.929384765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,1,4,power_law_1.2,12.650413818359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,1,4,power_law_1.01,0.4545062255859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,1,4,power_law_1.2,0.11416319847106933
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,1,4,power_law_1.2,0.13467647552490233
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,1,4,power_law_1.2,0.14428288459777833
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,1,4,power_law_1.2,0.1718604850769043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,1,4,power_law_1.2,0.17271167755126954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,1,4,power_law_1.2,0.17537664413452148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,1,4,power_law_1.2,0.17611391067504883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,1,4,power_law_1.2,0.17825023651123045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,1,4,power_law_1.2,0.18018560409545897
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,1,4,power_law_1.2,0.18144767761230468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,1,4,power_law_1.2,0.18542848587036134
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,1,4,power_law_1.2,0.1963238334655762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,1,4,power_law_1.2,0.20032768249511718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,1,4,power_law_1.2,0.21576192855834958
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,1,4,power_law_1.2,15.16339111328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,1,4,power_law_1.2,0.2349990463256836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,1,4,power_law_1.2,0.24686080932617188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,1,4,power_law_1.2,0.27241600036621094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,1,4,power_law_1.2,0.34548862457275387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,1,4,power_law_1.2,0.42926593780517575
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,1,4,power_law_1.01,1.3368870544433595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,1,4,power_law_1.2,0.5620198440551758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,1,4,power_law_1.2,0.865475845336914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,1,4,power_law_1.2,1.033570556640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,1,4,power_law_1.2,1.5890663146972657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,1,4,power_law_1.01,1.7928256225585937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,1,4,power_law_1.2,2.3835289001464846
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,1,4,power_law_1.2,0.05968128204345703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,1,4,power_law_1.2,0.07143040180206299
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,1,4,power_law_1.2,0.06687359809875489
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,1,4,power_law_1.2,0.07973631858825683
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,1,4,power_law_1.2,0.07935999870300293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,1,4,power_law_1.2,0.07990015983581543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,1,4,power_law_1.2,0.08041983604431152
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,1,4,power_law_1.2,0.08088576316833496
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,1,4,power_law_1.2,0.08163071632385253
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,1,4,power_law_1.2,0.08246272087097169
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,1,4,power_law_1.2,0.0832703971862793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,1,4,power_law_1.2,0.09492351531982421
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,1,4,power_law_1.2,0.1025830364227295
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,1,4,power_law_1.2,0.11000831604003906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,1,4,power_law_1.01,2.753313293457031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,1,4,power_law_1.01,3.6556478881835934
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,1,4,power_law_1.2,0.13551360130310058
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,1,4,power_law_1.2,0.1708185577392578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,1,4,power_law_1.2,2.906585693359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,1,4,power_law_1.2,0.26683391571044923
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,1,4,power_law_1.2,0.3577356719970703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,1,4,power_law_1.2,28.137956542968748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,1,4,power_law_1.2,4.547421569824219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,1,4,power_law_1.2,6.908466186523438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,1,4,power_law_1.2,0.9420326232910157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,1,4,power_law_1.2,1.6971827697753905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,1,4,power_law_1.2,33.78611572265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,2,2,balanced,0.12046463966369629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,2,2,balanced,0.1536243152618408
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,2,2,balanced,0.2733555221557617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,2,2,balanced,0.36535423278808593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,2,2,balanced,0.36624256134033206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,2,2,balanced,0.3667123031616211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,2,2,balanced,0.3723225784301758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,2,2,balanced,0.37299968719482424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,2,2,balanced,0.377789421081543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,2,2,balanced,0.378260498046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,2,2,balanced,0.3822963333129883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,2,2,balanced,0.38622718811035156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,2,2,balanced,0.3895987319946289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,2,2,balanced,0.39662334442138675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,2,2,balanced,0.5172185516357422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,2,2,balanced,0.5026867294311523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,2,2,balanced,0.4368652725219727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,2,2,balanced,0.6955506896972656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,2,2,balanced,0.7738162994384765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,2,2,balanced,1.1763008117675782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,2,2,balanced,1.4782579040527344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,2,2,balanced,2.27797119140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,2,2,balanced,2.903073425292969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,2,2,balanced,4.477743225097656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,1,4,power_law_1.2,0.13960831642150878
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,1,4,power_law_1.2,0.205664005279541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,2,2,balanced,6.495774536132812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,2,2,balanced,0.07295231819152832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,2,2,balanced,0.08923775672912597
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,1,4,power_law_1.2,0.515233268737793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,2,2,balanced,0.13666560173034667
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,2,2,balanced,0.1458841609954834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,2,2,balanced,0.14416255950927734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,2,2,balanced,0.15030783653259278
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,2,2,balanced,0.14608256340026854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,2,2,balanced,0.15243007659912108
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,2,2,balanced,0.14876288414001465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,2,2,balanced,0.15179519653320311
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,2,2,balanced,0.15344511985778808
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,2,2,balanced,0.15696895599365235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,2,2,balanced,0.15953663825988768
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,2,2,balanced,0.1680703926086426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,2,2,balanced,0.17173120498657227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,2,2,balanced,0.17753984451293944
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,2,2,balanced,0.17615232467651368
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,2,2,balanced,0.2438656044006348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,2,2,balanced,0.2603724861145019
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,2,2,balanced,0.3324224090576172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,2,2,balanced,0.41266048431396485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,2,2,balanced,0.5908607864379883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,2,2,balanced,0.8195520019531249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,2,2,balanced,1.1987712097167968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,2,2,balanced,10.351456298828126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,2,2,balanced,1.565244140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,1,4,power_law_1.2,0.8047500610351562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,2,2,balanced,2.341955871582031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,2,2,balanced,3.1315405273437498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,2,2,balanced,14.587380371093749
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,1,4,power_law_1.2,1.3519448852539062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,1,4,power_law_1.2,3.836756591796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,2,2,balanced,0.048071680068969724
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,2,2,balanced,0.08385279655456543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,2,2,balanced,0.030140159130096433
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,2,2,balanced,0.08383359909057617
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,2,2,balanced,0.08290559768676757
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,2,2,balanced,0.0838809585571289
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,2,2,balanced,0.08385408401489258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,2,2,balanced,0.08390912055969238
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,2,2,balanced,0.08343935966491699
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,2,2,balanced,0.08615296363830567
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,2,2,balanced,0.08583935737609863
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,2,2,balanced,0.08775679588317871
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,2,2,balanced,0.08153471946716309
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,2,2,balanced,0.08193280220031737
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,2,2,balanced,0.0824345588684082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,1,4,power_law_1.2,3.0665573120117187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,2,2,balanced,0.08286463737487793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,2,2,balanced,0.08291839599609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,2,2,balanced,0.11961728096008302
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,2,2,balanced,0.11390591621398927
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,2,2,balanced,0.2029952049255371
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,2,2,balanced,0.16096895217895507
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,2,2,power_law_1.01,0.12671999931335448
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,2,2,power_law_1.01,0.17297279357910156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,2,2,power_law_1.01,0.19469440460205079
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,2,2,balanced,0.28503295898437503
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,2,2,power_law_1.01,0.3643392181396484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,2,2,power_law_1.01,0.3646886444091797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,2,2,power_law_1.01,0.37112960815429685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,2,2,power_law_1.01,0.37423614501953123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,2,2,power_law_1.01,0.3780595016479492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,2,2,power_law_1.01,0.3798476791381836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,2,2,power_law_1.01,0.378155517578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,2,2,power_law_1.01,0.3806681442260742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,2,2,power_law_1.01,0.38348415374755856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,2,2,power_law_1.01,0.3930931091308594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,2,2,power_law_1.01,0.4856691360473633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,2,2,power_law_1.01,0.5619583892822265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,2,2,power_law_1.01,0.6018572616577148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,2,2,power_law_1.01,0.666519012451172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,2,2,power_law_1.01,0.8328947448730469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,2,2,balanced,0.37568641662597657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,2,2,power_law_1.01,1.1748941040039063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,2,2,power_law_1.01,1.814787902832031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,2,2,balanced,0.5526809692382812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,2,2,power_law_1.01,1.6495692443847656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,2,2,power_law_1.01,5.546307983398437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,2,2,power_law_1.01,3.639742736816406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,2,2,power_law_1.01,4.121196899414063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,2,2,power_law_1.01,0.09200256347656251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,2,2,power_law_1.01,0.1129651165008545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,2,2,power_law_1.01,0.1464742374420166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,2,2,power_law_1.01,0.1459993553161621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,2,2,power_law_1.01,0.14796031951904295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,2,2,power_law_1.01,0.14812543869018555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,2,2,power_law_1.01,0.15029248237609863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,2,2,power_law_1.01,0.15081088066101073
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,2,2,power_law_1.01,0.15186816215515136
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,2,2,power_law_1.01,0.15365632057189943
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,2,2,power_law_1.01,0.16481279373168944
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,2,2,power_law_1.01,0.16794496536254883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,2,2,power_law_1.01,0.17915008544921876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,2,2,power_law_1.01,0.18601984024047852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,2,2,power_law_1.01,0.19235328674316404
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,2,2,power_law_1.01,0.2274687957763672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,2,2,power_law_1.01,0.07707647800445556
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,2,2,power_law_1.01,0.27876352310180663
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,2,2,power_law_1.01,0.33343616485595706
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,2,2,power_law_1.01,0.42208000183105465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,2,2,power_law_1.01,0.530835189819336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,2,2,power_law_1.01,7.218772583007812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,2,2,power_law_1.01,1.0291059112548828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,2,2,power_law_1.01,0.6779174041748047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,2,2,power_law_1.01,1.465482177734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,2,2,balanced,0.7486515045166016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,2,2,power_law_1.01,0.048586239814758306
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,2,2,power_law_1.01,0.05376383781433105
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,2,2,power_law_1.01,0.06548223972320558
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,2,2,power_law_1.01,0.08235648155212402
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,2,2,power_law_1.01,0.08178048133850098
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,2,2,power_law_1.01,0.08193535804748535
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,2,2,power_law_1.01,0.08202495574951171
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,2,2,power_law_1.01,0.08243583679199219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,2,2,power_law_1.01,0.08260224342346192
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,2,2,power_law_1.01,0.08315648078918457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,2,2,power_law_1.01,14.143504638671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,2,2,power_law_1.01,0.0836736011505127
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,2,2,balanced,1.176294403076172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,2,2,power_law_1.01,0.10385663986206055
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,2,2,balanced,1.59957763671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,2,2,power_law_1.01,2.744879455566406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,2,2,power_law_1.01,2.280925445556641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,2,2,power_law_1.01,4.078275756835938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,2,2,power_law_1.01,17.61685546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,2,2,power_law_1.01,0.08887935638427734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,2,2,power_law_1.01,0.08574720382690429
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,2,2,power_law_1.01,0.27607807159423825
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,2,2,power_law_1.01,0.10570752143859863
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,2,2,power_law_1.01,0.11712127685546876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,2,2,power_law_1.01,0.11094271659851074
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,2,2,power_law_1.01,0.1538534355163574
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,2,2,power_law_1.01,0.19188224792480468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,2,2,power_law_1.2,0.12751487731933592
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,2,2,power_law_1.2,0.15582592010498048
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,2,2,power_law_1.2,0.1726464080810547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,2,2,power_law_1.2,0.36378368377685544
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,2,2,power_law_1.2,0.3653171157836914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,2,2,power_law_1.2,0.36768512725830077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,2,2,power_law_1.2,0.3750233459472656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,2,2,power_law_1.2,0.3786163330078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,2,2,power_law_1.01,0.24592512130737304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,2,2,power_law_1.2,0.3784115219116211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,2,2,power_law_1.2,0.3777676773071289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,2,2,power_law_1.2,0.38032257080078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,2,2,power_law_1.2,0.383502082824707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,2,2,power_law_1.2,0.39097599029541014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,2,2,power_law_1.2,0.470250244140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,2,2,power_law_1.2,0.5732608032226563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,2,2,power_law_1.2,0.6008204650878907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,2,2,power_law_1.2,0.7049305725097657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,2,2,power_law_1.2,0.8215039825439453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,2,2,power_law_1.2,1.020975341796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,2,2,power_law_1.2,1.6052172851562498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,2,2,power_law_1.2,2.117972412109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,2,2,power_law_1.2,3.2458010864257814
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,2,2,power_law_1.01,0.42604927062988285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,2,2,power_law_1.2,4.120499267578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,2,2,power_law_1.01,0.6097574234008789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,2,2,power_law_1.2,6.153289184570313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,2,2,power_law_1.2,0.08453120231628418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,2,2,power_law_1.2,0.10183936119079591
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,2,2,power_law_1.2,0.10709376335144043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,2,2,power_law_1.2,0.1458304023742676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,2,2,power_law_1.2,0.14623616218566896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,2,2,power_law_1.2,0.14786304473876952
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,2,2,power_law_1.2,0.14843008041381836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,2,2,power_law_1.2,0.15075072288513183
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,2,2,power_law_1.2,0.15188096046447755
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,2,2,power_law_1.2,0.1513036823272705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,2,2,power_law_1.2,0.1532736015319824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,2,2,power_law_1.2,0.1647475242614746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,2,2,power_law_1.2,0.16774272918701172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,2,2,power_law_1.2,0.17911552429199218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,2,2,power_law_1.2,0.18772991180419923
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,2,2,power_law_1.2,0.19941503524780274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,2,2,power_law_1.2,0.230515193939209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,2,2,power_law_1.2,7.589968872070313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,2,2,power_law_1.2,0.2777292823791504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,2,2,power_law_1.2,0.29779327392578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,2,2,power_law_1.01,0.8330124664306642
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,2,2,power_law_1.2,0.4288819122314453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,2,2,power_law_1.2,0.5418112182617187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,2,2,power_law_1.01,1.2090726470947266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,2,2,power_law_1.2,1.0380723571777344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,2,2,power_law_1.2,0.8264409637451171
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,2,2,power_law_1.01,1.615764465332031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,2,2,power_law_1.2,0.04509183883666992
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,2,2,power_law_1.2,0.05603328227996827
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,2,2,power_law_1.2,0.056775679588317876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,2,2,power_law_1.2,0.08094207763671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,2,2,power_law_1.2,0.08118271827697754
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,2,2,power_law_1.2,0.08104448318481446
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,2,2,power_law_1.2,0.08171520233154297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,2,2,power_law_1.2,0.08209024429321289
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,2,2,power_law_1.2,0.08231040000915527
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,2,2,power_law_1.2,0.08244223594665527
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,2,2,power_law_1.2,0.08331007957458496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,2,2,power_law_1.2,1.4110887145996094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,2,2,power_law_1.2,0.09356927871704102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,2,2,power_law_1.2,2.1816908264160153
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,2,2,power_law_1.2,0.09553919792175293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,2,2,power_law_1.2,0.10160767555236816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,2,2,power_law_1.2,2.870008239746094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,2,2,power_law_1.2,0.15603839874267578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,2,2,power_law_1.01,2.34440185546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,2,2,power_law_1.2,21.0501171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,2,2,power_law_1.2,0.30950912475585934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,2,2,power_law_1.2,16.24483154296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,2,2,power_law_1.2,4.295367736816407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,2,2,power_law_1.2,0.7832832336425781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,2,2,power_law_1.2,0.08602368354797363
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,2,2,power_law_1.2,0.10694527626037598
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,2,2,power_law_1.2,0.12201215744018554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,4,1,balanced,0.10357439994812012
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,4,1,balanced,0.1356825637817383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,4,1,balanced,0.26396223068237307
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,2,2,power_law_1.2,0.16735103607177734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,4,1,balanced,0.25681343078613283
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,4,1,balanced,0.2672480010986328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,4,1,balanced,0.2585817527770996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,4,1,balanced,0.25979904174804686
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,2,2,power_law_1.2,1.6007437133789062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,4,1,balanced,0.25987968444824217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,4,1,balanced,0.26409664154052737
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,4,1,balanced,0.2643788719177246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,4,1,balanced,0.271265926361084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,4,1,balanced,0.27035327911376955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,4,1,balanced,0.29552896499633785
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,2,2,power_law_1.2,0.2561433601379394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,4,1,balanced,0.4232896041870117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,4,1,balanced,0.4169535827636719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,4,1,balanced,0.27084800720214847
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,4,1,balanced,0.4260582351684571
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,4,1,balanced,0.6020684814453124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,4,1,balanced,0.7062540435791016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,4,1,balanced,1.117952651977539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,4,1,balanced,1.4025830078125001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,4,1,balanced,2.098768005371094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,4,1,balanced,2.8181198120117186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,4,1,balanced,4.139496459960937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,4,1,balanced,0.06748223781585694
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,4,1,balanced,0.08535103797912598
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,4,1,balanced,0.1403219223022461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,4,1,balanced,5.555416870117187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,2,2,power_law_1.2,0.5092851257324218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,4,1,balanced,0.1564793586730957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,4,1,balanced,0.15398015975952148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,4,1,balanced,0.15467776298522948
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,4,1,balanced,0.15317952156066894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,4,1,balanced,0.15275263786315918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,4,1,balanced,0.15481663703918458
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,4,1,balanced,0.15538304328918456
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,4,1,balanced,0.15721407890319825
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,4,1,balanced,0.15959487915039064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,4,1,balanced,0.16416704177856445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,4,1,balanced,0.16671936035156248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,4,1,balanced,0.17202367782592773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,4,1,balanced,0.17264383316040038
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,4,1,balanced,0.22579839706420896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,4,1,balanced,0.23806783676147464
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,4,1,balanced,0.31877376556396486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,4,1,balanced,0.3951129531860352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,4,1,balanced,0.5665689468383789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,4,1,balanced,8.263589477539062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,4,1,balanced,0.747864990234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,4,1,balanced,1.153403549194336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,4,1,balanced,0.13853631973266603
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,2,2,power_law_1.2,0.5676774215698243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,4,1,balanced,1.5137600708007812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,4,1,balanced,2.2594784545898436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,4,1,balanced,3.0109292602539064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,4,1,balanced,12.4831103515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,2,2,power_law_1.2,1.2247052764892579
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,4,1,balanced,0.027980799674987795
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,4,1,balanced,0.05083392143249512
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,4,1,balanced,0.08087424278259278
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,4,1,balanced,0.0814688014984131
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,4,1,balanced,0.0814361572265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,4,1,balanced,0.0799174404144287
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,4,1,balanced,0.08159168243408203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,4,1,balanced,0.08122112274169921
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,4,1,balanced,0.08159551620483399
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,4,1,balanced,0.0838425636291504
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,4,1,balanced,0.08514368057250976
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,4,1,balanced,0.08567872047424316
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,4,1,balanced,0.08750271797180176
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,4,1,balanced,0.07932735919952393
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,4,1,balanced,0.0792787218093872
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,4,1,balanced,0.08060864448547364
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,4,1,balanced,0.08105152130126954
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,2,2,power_law_1.2,2.103829803466797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,4,1,balanced,0.12250047683715821
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,4,1,balanced,0.11594752311706542
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,4,1,balanced,0.1671238327026367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,4,1,power_law_1.01,0.1039475154876709
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,4,1,power_law_1.01,0.1327616024017334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,4,1,power_law_1.01,0.1638528060913086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,4,1,balanced,0.20666879653930664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,4,1,power_law_1.01,0.2597350311279297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,4,1,power_law_1.01,0.2584799957275391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,4,1,power_law_1.01,0.26150783538818356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,4,1,power_law_1.01,0.2619001579284668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,4,1,power_law_1.01,0.2585215950012207
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,4,1,balanced,0.29248512268066407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,4,1,power_law_1.01,0.26216320037841795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,4,1,power_law_1.01,0.2827359962463379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,4,1,power_law_1.01,0.3550867080688477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,4,1,power_law_1.01,0.36763648986816405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,4,1,power_law_1.01,0.38035648345947265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,4,1,power_law_1.01,0.401580810546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,4,1,power_law_1.01,0.4186105728149414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,4,1,power_law_1.01,0.45689727783203127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,4,1,power_law_1.01,0.5273491287231445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,4,1,power_law_1.01,0.6794400024414062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,4,1,balanced,0.3786406326293945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,4,1,power_law_1.01,0.8349241638183594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,4,1,power_law_1.01,1.152576675415039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,4,1,power_law_1.01,1.5195628356933593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,4,1,power_law_1.01,2.196216888427734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,4,1,power_law_1.01,2.898828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,4,1,power_law_1.01,4.3228564453125005
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,4,1,power_law_1.01,0.0685478401184082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,4,1,power_law_1.01,0.08340607643127441
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,4,1,power_law_1.01,0.10301631927490233
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,4,1,power_law_1.01,0.13901760101318358
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,4,1,power_law_1.01,0.14845952033996582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,4,1,power_law_1.01,0.1493183994293213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,4,1,power_law_1.01,0.15082240104675293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,4,1,power_law_1.01,0.15127103805541992
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,4,1,balanced,0.5780108642578126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,4,1,power_law_1.01,0.15808064460754395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,4,1,power_law_1.01,0.15465984344482422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,4,1,power_law_1.01,0.15499520301818848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,4,1,power_law_1.01,0.15502464294433593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,4,1,power_law_1.01,0.16063488006591797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,4,1,power_law_1.01,0.1701145553588867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,4,1,power_law_1.01,5.712994384765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,4,1,power_law_1.01,0.17673023223876955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,4,1,power_law_1.01,0.178143367767334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,4,1,power_law_1.01,0.19716735839843752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,4,1,power_law_1.01,0.22749183654785154
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,4,1,balanced,0.7482015991210937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,4,1,power_law_1.01,0.2710329627990723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,4,1,power_law_1.01,0.3490707015991211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,4,1,power_law_1.01,0.42597183227539065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,4,1,power_law_1.01,0.5861625671386719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,4,1,balanced,1.2107603454589844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,4,1,balanced,1.6871014404296876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,4,1,power_law_1.01,0.03880127906799317
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,4,1,power_law_1.01,0.05677120208740234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,4,1,power_law_1.01,0.05866879940032958
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,4,1,power_law_1.01,0.7834374237060546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,4,1,power_law_1.01,0.07939263820648193
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,4,1,power_law_1.01,0.08503104209899902
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,4,1,power_law_1.01,0.08021247863769532
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,4,1,power_law_1.01,0.08134592056274415
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,4,1,power_law_1.01,0.07991487979888916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,4,1,power_law_1.01,0.07997119903564454
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,4,1,power_law_1.01,0.08083776473999024
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,4,1,power_law_1.01,0.0818387222290039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,4,1,power_law_1.01,0.08202431678771974
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,4,1,power_law_1.01,0.08882687568664552
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,4,1,power_law_1.01,1.1570829010009764
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,4,1,power_law_1.01,8.518812255859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,4,1,power_law_1.01,1.5438490295410157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,4,1,power_law_1.01,0.1324998378753662
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,4,1,power_law_1.01,2.257648620605469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,4,1,power_law_1.01,11.474691162109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,4,1,power_law_1.01,2.9657342529296877
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,4,1,power_law_1.01,0.3455302429199219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,4,1,power_law_1.01,0.08332799911499024
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,4,1,power_law_1.01,0.09145088195800781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,4,1,power_law_1.01,0.0945299243927002
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,4,1,power_law_1.01,0.10538816452026367
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,4,1,power_law_1.01,0.15544320106506349
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,4,1,power_law_1.2,0.10385855674743652
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,4,1,power_law_1.2,0.12967103958129883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,4,1,power_law_1.2,0.15102527618408204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,4,1,power_law_1.2,0.2601587104797363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,4,1,power_law_1.2,0.25089664459228517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,4,1,power_law_1.2,0.2608966445922852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,4,1,power_law_1.2,0.26240575790405274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,4,1,power_law_1.2,0.2629171180725097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,4,1,power_law_1.2,0.26695104598999025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,4,1,power_law_1.2,0.27999616622924806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,4,1,power_law_1.2,0.36277889251708983
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,4,1,power_law_1.2,0.37354751586914064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,4,1,power_law_1.2,0.37519809722900394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,4,1,power_law_1.2,0.3989139175415039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,4,1,power_law_1.2,0.42401470184326173
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,4,1,power_law_1.01,0.2017171287536621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,4,1,power_law_1.2,0.46970176696777344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,4,1,power_law_1.2,0.5226496124267579
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,4,1,power_law_1.2,0.6731212615966797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,4,1,power_law_1.01,0.2509689521789551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,4,1,power_law_1.2,0.8370706939697266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,4,1,power_law_1.2,1.1774726104736328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,4,1,power_law_1.2,1.5390176391601562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,4,1,power_law_1.2,2.1942707824707033
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,4,1,power_law_1.01,0.4430329513549805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,4,1,power_law_1.2,2.8908453369140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,4,1,power_law_1.01,0.6639826965332032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,4,1,power_law_1.2,0.06765503883361816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,4,1,power_law_1.2,0.08058431625366211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,4,1,power_law_1.2,0.0879270362854004
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,4,1,power_law_1.01,0.9139647674560546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,4,1,power_law_1.2,0.15183615684509277
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,4,1,power_law_1.2,0.15205504417419433
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,4,1,power_law_1.2,0.15232192039489748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,4,1,power_law_1.2,0.15261759757995605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,4,1,power_law_1.2,0.1536832046508789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,4,1,power_law_1.2,0.15429504394531252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,4,1,power_law_1.2,0.1554905605316162
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,4,1,power_law_1.2,0.1573683166503906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,4,1,power_law_1.2,0.1570918369293213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,4,1,power_law_1.2,0.1717728042602539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,4,1,power_law_1.2,0.13887552261352537
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,4,1,power_law_1.2,0.18147647857666016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,4,1,power_law_1.2,0.1832467269897461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,4,1,power_law_1.2,4.292489013671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,4,1,power_law_1.2,0.19616575241088868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,4,1,power_law_1.2,0.22763008117675781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,4,1,power_law_1.2,0.271146240234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,4,1,power_law_1.2,0.3449951934814453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,4,1,power_law_1.01,1.3697573852539062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,4,1,power_law_1.2,0.42401920318603514
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,4,1,power_law_1.2,0.6062387084960937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,4,1,power_law_1.2,5.669096679687501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,4,1,power_law_1.2,0.7903366088867188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,4,1,power_law_1.01,1.8154765319824218
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,4,1,power_law_1.2,0.0355679988861084
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,4,1,power_law_1.2,0.049723520278930664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,4,1,power_law_1.2,0.05246016025543213
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,4,1,power_law_1.2,0.07942399978637696
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,4,1,power_law_1.2,0.07859519958496095
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,4,1,power_law_1.2,0.08025088310241699
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,4,1,power_law_1.2,0.08048704147338867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,4,1,power_law_1.2,0.07978752136230469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,4,1,power_law_1.2,0.08044672012329102
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,4,1,power_law_1.2,0.08095295906066895
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,4,1,power_law_1.2,0.08171008110046388
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,4,1,power_law_1.2,0.08253376007080078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,4,1,power_law_1.2,0.0888640022277832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,4,1,power_law_1.2,0.09363136291503907
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,4,1,power_law_1.2,0.09796159744262696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,4,1,power_law_1.2,1.1711775970458984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,4,1,power_law_1.2,8.4295556640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,4,1,power_law_1.2,1.5332485961914064
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,4,1,power_law_1.2,0.20453887939453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,4,1,power_law_1.2,0.08187775611877442
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,4,1,power_law_1.2,2.282604217529297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,4,1,power_law_1.2,0.1086137580871582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,4,1,power_law_1.2,11.711304931640624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,4,1,power_law_1.2,0.13205632209777832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,4,1,power_law_1.2,0.1553926372528076
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,4,1,power_law_1.2,2.976237487792969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,4,balanced,0.0833523178100586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,4,balanced,0.09797632217407226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,4,balanced,0.15003520011901855
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,4,1,power_law_1.2,0.25364736557006834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,4,balanced,0.29716991424560546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,4,balanced,0.5191206359863282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,4,balanced,0.9539686584472656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,4,balanced,0.9489459228515624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,4,balanced,0.947925796508789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,4,balanced,1.0456524658203126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,4,balanced,1.045379867553711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,4,balanced,1.053416976928711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,4,balanced,1.068826904296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,4,balanced,1.0672537231445314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,4,balanced,1.091440658569336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,4,balanced,1.1108633422851564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,4,balanced,1.1165503692626952
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,4,balanced,1.1443814086914064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,4,balanced,1.1873689270019532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,4,balanced,1.2467123413085939
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,4,balanced,1.3315545654296874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,4,balanced,1.4452761840820312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,4,balanced,2.1749081420898437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,4,balanced,2.4176704406738283
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,4,balanced,3.382243957519531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,4,balanced,4.422763671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,4,1,power_law_1.2,0.3557875061035156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,4,balanced,0.04597248077392578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,4,balanced,0.06130303859710693
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,4,balanced,0.08844799995422363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,4,balanced,0.1571020793914795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,4,balanced,0.27221759796142575
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,4,balanced,0.4852057647705078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,4,balanced,7.011421508789063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,4,balanced,0.4863475036621094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,4,balanced,0.4914547348022461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,4,balanced,0.5252467346191406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,4,1,power_law_1.2,1.3898579406738283
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,4,balanced,0.5234636688232422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,4,1,power_law_1.2,0.4543993759155273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,4,balanced,0.5257254409790039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,4,balanced,0.5292633438110352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,4,balanced,0.5314841461181641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,4,balanced,0.536572151184082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,4,balanced,0.5390156936645508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,4,balanced,0.5430412673950196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,4,balanced,0.5577881622314453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,4,balanced,0.612993278503418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,4,balanced,0.6209600067138672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,4,balanced,0.6859264373779297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,4,balanced,0.7592076873779297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,4,balanced,0.8205888366699219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,4,balanced,0.9851827239990234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,4,balanced,1.422040252685547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,4,balanced,10.582760009765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,4,balanced,1.6149427795410154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,4,balanced,2.335758056640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,4,balanced,3.0128256225585934
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,4,1,power_law_1.2,0.6805305480957031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,4,1,power_law_1.2,0.9107513427734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,1,4,balanced,0.032170240879058835
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,1,4,balanced,0.04081535816192627
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,1,4,balanced,0.2906572723388672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,1,4,balanced,0.2922419166564941
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,1,4,balanced,0.05911935806274414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,1,4,balanced,0.2905843162536621
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,1,4,balanced,0.2930572891235351
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,1,4,balanced,0.2925081634521484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,1,4,balanced,0.29220991134643554
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,1,4,balanced,0.09230463981628419
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,1,4,balanced,0.29559680938720706
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,1,4,balanced,0.2960767936706543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,1,4,balanced,0.29655296325683594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,1,4,balanced,0.29893375396728517
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,1,4,balanced,0.3051084709167481
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,1,4,balanced,0.30943487167358397
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,1,4,balanced,0.31947519302368166
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,1,4,balanced,0.3287884902954102
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,1,4,balanced,0.3466495895385742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,1,4,balanced,0.1562060832977295
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,1,4,balanced,0.36701953887939454
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,1,4,balanced,0.289881591796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,4,power_law_1.01,0.29740543365478517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,4,power_law_1.01,0.14438912391662598
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,4,power_law_1.01,0.36198272705078127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,4,power_law_1.01,0.584527359008789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,4,power_law_1.01,0.4781951904296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,4,power_law_1.01,0.6495053100585937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,4,power_law_1.01,0.8429682922363281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,4,power_law_1.01,0.8780095672607422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,4,power_law_1.01,0.9832473754882812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,4,power_law_1.01,0.9885453033447267
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,1,4,balanced,0.2911129570007324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,4,power_law_1.01,0.9941490936279298
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,4,power_law_1.01,1.0072972869873047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,4,power_law_1.01,0.9388492584228516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,4,power_law_1.01,1.0103948974609376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,4,power_law_1.01,1.0324044799804688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,4,power_law_1.01,1.0912806701660156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,4,power_law_1.01,1.1434815979003905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,4,power_law_1.01,1.2665433502197265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,4,power_law_1.01,1.3712806701660156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,4,power_law_1.01,1.7802931213378905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,4,power_law_1.01,1.9787084960937498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,4,power_law_1.01,2.547950134277344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,4,1,power_law_1.2,1.8426617431640626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,4,power_law_1.01,3.0902810668945313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,4,power_law_1.01,4.113734436035156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,4,power_law_1.01,0.08845696449279786
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,4,power_law_1.01,0.14851584434509277
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,4,power_law_1.01,0.17121280670166014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,4,power_law_1.01,0.2519692802429199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,4,power_law_1.01,5.497276000976562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,4,power_law_1.01,0.28976255416870117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,4,power_law_1.01,0.4050649642944336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,4,power_law_1.01,0.3903744125366211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,4,power_law_1.01,0.43871231079101564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,4,power_law_1.01,7.701015014648438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,4,power_law_1.01,0.4415270233154297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,4,power_law_1.01,0.49396480560302736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,4,power_law_1.01,0.4813695907592773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,4,power_law_1.01,0.49128448486328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,4,power_law_1.01,0.48251777648925775
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,4,power_law_1.01,0.49911041259765626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,4,power_law_1.01,0.5113382339477539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,4,power_law_1.01,10.33607666015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,4,power_law_1.01,0.5315596771240234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,4,power_law_1.01,0.5899711990356445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,4,power_law_1.01,0.7042265319824219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,4,power_law_1.01,0.6171980667114257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,4,power_law_1.01,0.767504653930664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,4,power_law_1.01,0.9878221130371093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,4,power_law_1.01,1.1487257385253906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,4,power_law_1.01,1.547052764892578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,1,4,balanced,0.5817011260986329
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,1,4,power_law_1.01,0.06602240085601807
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,1,4,balanced,0.5472435379028321
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,1,4,balanced,1.1230528259277344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,1,4,balanced,0.8121651458740236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,4,power_law_1.01,1.89947265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,1,4,power_law_1.01,0.09878144264221192
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,1,4,power_law_1.01,0.10744192123413085
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,1,4,power_law_1.01,0.1470246410369873
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,1,4,power_law_1.01,0.18044544219970704
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,1,4,power_law_1.01,0.2406399917602539
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,1,4,power_law_1.01,0.22665216445922853
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,1,4,power_law_1.01,0.2371084785461426
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,1,4,power_law_1.01,0.2434163284301758
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,1,4,power_law_1.01,0.25264768600463866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,4,power_law_1.01,2.7105892944335936
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,1,4,power_law_1.01,0.2632947158813477
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,1,4,power_law_1.01,0.25253631591796877
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,1,4,power_law_1.01,0.2642969512939453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,1,4,power_law_1.01,0.26664831161499025
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,1,4,power_law_1.01,0.28354303359985356
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,1,4,power_law_1.01,0.28812288284301757
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,4,power_law_1.01,3.587580261230469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,4,power_law_1.01,0.3463193511962891
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,1,4,power_law_1.01,0.17822208404541015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,1,4,power_law_1.01,0.32156417846679686
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,1,4,power_law_1.01,0.3851123046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,1,4,power_law_1.01,0.3360268783569336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,1,4,power_law_1.01,0.42594047546386715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,4,power_law_1.2,0.14401920318603517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,4,power_law_1.2,0.29700992584228514
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,4,power_law_1.2,0.33043712615966797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,1,4,power_law_1.01,0.6455462646484376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,4,power_law_1.2,0.43756671905517575
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,4,power_law_1.2,0.5702220916748046
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,1,4,power_law_1.01,0.537660140991211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,4,power_law_1.2,0.6621990203857422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,4,power_law_1.2,0.7955673980712891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,4,power_law_1.2,0.8759232330322266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,4,power_law_1.2,0.8390707397460938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,4,power_law_1.2,0.9870873260498048
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,4,power_law_1.2,0.986709747314453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,4,power_law_1.2,0.9879116821289063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,4,power_law_1.2,0.97723388671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,4,power_law_1.2,1.0257062530517578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,4,power_law_1.2,1.0910873413085938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,4,power_law_1.2,1.1211097717285157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,4,power_law_1.2,1.1486975860595703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,1,4,power_law_1.01,0.851805419921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,4,power_law_1.2,1.3033395385742188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,4,power_law_1.2,1.4444837951660157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,4,power_law_1.2,1.7719590759277346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,4,power_law_1.2,2.111109161376953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,4,power_law_1.2,2.7524273681640627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,4,power_law_1.2,3.2240280151367187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,4,power_law_1.2,0.08647551536560058
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,4,power_law_1.2,0.14470272064208983
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,4,power_law_1.2,4.3949581909179685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,4,power_law_1.2,0.15018367767333984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,4,power_law_1.2,0.24564735412597657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,4,power_law_1.2,0.27125503540039064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,4,power_law_1.2,0.29312255859374997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,4,power_law_1.2,0.40557697296142575
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,4,power_law_1.2,5.508999633789062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,4,power_law_1.2,0.41496959686279294
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,4,power_law_1.2,0.4343731307983399
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,4,power_law_1.2,0.489167366027832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,4,power_law_1.2,0.4640499114990234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,4,power_law_1.2,0.48590335845947263
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,4,power_law_1.2,0.4812633514404297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,4,power_law_1.2,0.4978995132446289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,4,power_law_1.2,0.5083993530273437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,4,power_law_1.2,0.5208166503906251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,4,power_law_1.2,0.528007698059082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,4,power_law_1.2,0.6073574447631837
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,4,power_law_1.2,8.481632080078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,4,power_law_1.2,0.6343577575683594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,4,power_law_1.2,0.7297830200195312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,4,power_law_1.2,0.8093965148925781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,4,power_law_1.2,1.0100621032714845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,4,power_law_1.2,1.158608627319336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,4,power_law_1.2,1.6092620849609376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,1,4,power_law_1.2,0.06421376228332519
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,1,4,power_law_1.2,0.09794048309326173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,4,power_law_1.2,2.0652825927734377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,4,power_law_1.2,11.037050781249999
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,1,4,power_law_1.2,0.10110719680786133
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,1,4,power_law_1.2,0.13435263633728028
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,1,4,power_law_1.2,0.1828531265258789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,1,4,power_law_1.2,0.16473344802856443
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,1,4,power_law_1.2,0.21830783843994142
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,1,4,power_law_1.2,0.22306304931640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,1,4,power_law_1.2,0.2393190383911133
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,1,4,power_law_1.2,0.23893632888793945
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,1,4,power_law_1.2,0.25625984191894535
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,1,4,power_law_1.2,0.25993600845336917
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,1,4,power_law_1.2,0.25618175506591795
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,1,4,power_law_1.2,0.2665983963012695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,4,power_law_1.2,3.7351245117187504
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,1,4,power_law_1.2,0.2702681541442871
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,1,4,power_law_1.2,0.2870681571960449
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,4,power_law_1.2,2.96967041015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,1,4,power_law_1.01,1.0658841705322266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,1,4,power_law_1.01,1.7411929321289061
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,1,4,power_law_1.01,2.088553009033203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,1,4,power_law_1.2,0.30256128311157227
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,1,4,power_law_1.2,0.33808895111083986
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,1,4,power_law_1.2,0.3658790588378906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,1,4,power_law_1.2,0.4249190521240235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,2,balanced,0.07493631839752198
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,2,balanced,0.09745152473449707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,2,balanced,0.1418995189666748
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,1,4,power_law_1.2,0.4864076614379883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,2,balanced,0.29379072189331057
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,2,balanced,0.5223334503173829
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,2,balanced,0.953658905029297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,2,balanced,0.9519602966308593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,2,balanced,0.9564211273193359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,2,balanced,0.958182373046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,2,balanced,0.9660326385498047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,2,balanced,0.9660224151611327
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,1,4,power_law_1.2,0.6663654327392579
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,2,balanced,1.0471564483642577
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,2,balanced,1.0583052825927735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,2,balanced,1.0505292510986328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,2,balanced,1.0785523223876954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,2,balanced,1.0857497406005858
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,2,balanced,1.0909017944335937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,1,4,power_law_1.2,0.8661325073242188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,2,balanced,1.1642880249023437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,2,balanced,1.2243660736083986
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,1,4,power_law_1.2,1.2552499389648437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,2,balanced,1.2843994140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,2,balanced,1.393377227783203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,2,balanced,2.1161958312988283
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,2,balanced,2.2890713500976565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,2,balanced,0.04594560146331787
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,2,balanced,0.06222208023071289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,2,balanced,0.09112575531005859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,2,balanced,3.1633062744140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,2,balanced,0.16305536270141602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,2,balanced,4.074919738769531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,2,balanced,0.29314687728881833
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,2,balanced,0.5096345520019531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,2,balanced,0.5089548873901367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,2,balanced,6.07114990234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,2,balanced,0.5130483245849609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,2,balanced,0.5160268783569336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,2,balanced,0.5240793609619141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,2,balanced,0.5627699279785157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,2,balanced,0.5649856185913086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,2,balanced,0.5735769653320313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,2,balanced,0.5737523269653321
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,2,balanced,0.580948486328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,2,balanced,0.5895808029174805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,2,balanced,0.6123775863647462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,2,balanced,0.6474674987792969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,2,balanced,0.6790118408203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,2,balanced,8.892413330078124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,2,balanced,0.7399718475341797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,2,balanced,0.8330303955078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,2,balanced,0.9423948669433594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,2,balanced,1.4203955078125001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,2,balanced,1.6399743652343752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,2,balanced,0.5068544006347656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,2,balanced,2.3258035278320315
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,2,balanced,3.018266906738281
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,1,4,power_law_1.2,1.4872921752929689
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,1,4,power_law_1.2,2.4210418701171874
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,1,4,power_law_1.2,3.288355712890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,2,2,balanced,0.04996736049652099
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,2,2,balanced,0.0414412784576416
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,2,2,balanced,0.3120691108703613
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,2,2,balanced,0.3121561622619629
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,2,2,balanced,0.31371648788452144
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,2,2,balanced,0.1000716781616211
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,2,2,balanced,0.3133452796936035
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,2,2,balanced,0.06707071781158447
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,2,2,balanced,0.31531648635864257
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,2,2,balanced,0.3168255996704102
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,2,2,balanced,0.31863807678222655
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,2,2,balanced,0.3214591979980469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,2,2,balanced,0.32465534210205077
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,2,2,balanced,0.33295745849609376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,2,2,balanced,0.3399385452270508
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,2,2,balanced,0.3558195114135742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,2,2,balanced,0.16863616943359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,2,2,balanced,0.37255039215087893
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,2,2,balanced,0.40145919799804686
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,2,2,balanced,0.4358208084106446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,2,power_law_1.01,0.09743359565734863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,2,power_law_1.01,0.138470401763916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,2,power_law_1.01,0.20209280014038086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,2,power_law_1.01,0.3660774230957031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,2,power_law_1.01,0.4598591995239258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,2,power_law_1.01,0.5583961486816407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,2,power_law_1.01,0.7533964538574219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,2,power_law_1.01,0.7694207763671874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,2,power_law_1.01,0.8164147186279298
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,2,2,balanced,0.31084543228149414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,2,power_law_1.01,0.8207052612304688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,2,power_law_1.01,0.8442483520507812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,2,2,balanced,0.310567684173584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,2,power_law_1.01,0.9607782745361328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,2,power_law_1.01,0.9683033752441407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,2,power_law_1.01,0.9665433502197265
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,2,2,balanced,0.31106048583984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,2,power_law_1.01,1.0087065887451172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,2,power_law_1.01,1.0297369384765624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,2,power_law_1.01,1.0574886322021484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,2,power_law_1.01,1.1723417663574218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,2,power_law_1.01,1.3152204895019532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,2,power_law_1.01,1.5250918579101562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,2,power_law_1.01,1.7781619262695312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,2,power_law_1.01,2.2178111267089844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,2,power_law_1.01,2.796423645019531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,2,power_law_1.01,3.868104858398438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,2,power_law_1.01,0.061669120788574214
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,2,power_law_1.01,0.08896639823913574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,2,power_law_1.01,4.722603454589843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,2,power_law_1.01,0.1278425598144531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,2,power_law_1.01,0.20680704116821289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,2,power_law_1.01,0.26864383697509764
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,2,power_law_1.01,0.3205683135986328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,2,power_law_1.01,0.41554817199707034
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,2,power_law_1.01,0.4224678421020508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,2,power_law_1.01,0.44066432952880863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,2,power_law_1.01,6.77493896484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,2,power_law_1.01,0.4409075164794922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,2,power_law_1.01,0.4585292816162109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,2,power_law_1.01,0.5097548675537109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,2,power_law_1.01,0.5090227127075195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,2,power_law_1.01,0.5094464111328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,2,power_law_1.01,0.5264588928222655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,2,power_law_1.01,0.5233177566528321
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,2,power_law_1.01,0.5579571151733399
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,2,power_law_1.01,8.559522705078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,2,power_law_1.01,0.5866700744628905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,2,power_law_1.01,0.6258265686035156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,2,power_law_1.01,0.7183001708984376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,2,power_law_1.01,0.8721087646484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,2,power_law_1.01,0.9794636535644532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,2,power_law_1.01,1.1612582397460938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,2,power_law_1.01,1.4774259948730468
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,2,2,power_law_1.01,0.05801983833312988
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,2,power_law_1.01,1.8809510803222658
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,2,2,power_law_1.01,0.0740505599975586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,2,2,power_law_1.01,0.08799103736877442
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,2,2,power_law_1.01,0.11974399566650391
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,2,2,power_law_1.01,0.14956671714782716
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,2,2,power_law_1.01,0.18763647079467774
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,2,2,balanced,0.634260482788086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,2,2,power_law_1.01,0.24371456146240233
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,2,power_law_1.01,2.6365502929687503
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,2,2,balanced,0.7310374450683594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,2,2,power_law_1.01,0.25486848831176756
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,2,2,balanced,1.0568256378173828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,2,2,power_law_1.01,0.2678016090393066
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,2,2,power_law_1.01,0.26961536407470704
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,2,2,power_law_1.01,0.2800896072387695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,2,2,power_law_1.01,0.27621503829956057
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,2,2,balanced,1.403884735107422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,2,2,power_law_1.01,0.28482303619384763
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,2,2,power_law_1.01,0.2988198471069336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,2,2,power_law_1.01,0.2845363235473633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,2,power_law_1.01,3.326077575683594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,2,2,power_law_1.01,0.3072524833679199
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,2,2,power_law_1.01,0.3230502319335938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,2,2,power_law_1.01,0.3518092727661133
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,2,2,power_law_1.01,0.4548646545410156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,2,2,power_law_1.01,0.3853273773193359
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,2,2,power_law_1.01,0.646135025024414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,2,power_law_1.2,0.0967027187347412
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,2,2,power_law_1.01,0.5349939346313477
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,2,2,power_law_1.01,0.8778777313232421
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,2,power_law_1.2,0.18905088424682617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,2,power_law_1.2,0.13811840057373045
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,2,2,power_law_1.01,1.1726118469238282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,2,power_law_1.2,0.3467340850830078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,2,power_law_1.2,0.4283443069458007
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,2,power_law_1.2,0.5416447830200195
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,2,2,power_law_1.01,1.449548797607422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,2,power_law_1.2,0.7221056365966797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,2,power_law_1.2,0.7323455810546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,2,power_law_1.2,0.7772697448730469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,2,2,power_law_1.01,2.173322296142578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,2,power_law_1.2,0.8076595306396485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,2,power_law_1.2,0.8494220733642578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,2,power_law_1.2,0.9577458953857422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,2,power_law_1.2,0.9651277160644531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,2,power_law_1.2,0.986240005493164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,2,power_law_1.2,1.0174387359619141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,2,power_law_1.2,1.035242233276367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,2,power_law_1.2,1.1038156890869142
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,2,power_law_1.2,1.2056153869628905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,2,power_law_1.2,1.3141786193847655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,2,power_law_1.2,1.5659507751464843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,2,power_law_1.2,1.8959065246582032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,2,power_law_1.2,2.3593177795410156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,2,power_law_1.2,0.08918527603149415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,2,power_law_1.2,0.06155648231506348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,2,power_law_1.2,2.765910949707031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,2,power_law_1.2,0.12361087799072265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,2,power_law_1.2,0.2959103965759277
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,2,power_law_1.2,4.075438232421876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,2,power_law_1.2,0.18381439208984376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,2,power_law_1.2,0.41218814849853513
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,2,power_law_1.2,0.4277171325683594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,2,power_law_1.2,0.437468147277832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,2,power_law_1.2,4.891818237304688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,2,power_law_1.2,0.44653568267822263
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,2,power_law_1.2,0.4593983840942383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,2,power_law_1.2,0.5086284637451172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,2,power_law_1.2,0.5089664077758789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,2,power_law_1.2,0.5203891372680663
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,2,power_law_1.2,0.5353523254394531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,2,power_law_1.2,0.5241664123535157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,2,power_law_1.2,0.5685145568847656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,2,power_law_1.2,0.5960409545898437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,2,power_law_1.2,0.6583500671386718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,2,power_law_1.2,0.7408525085449218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,2,power_law_1.2,7.206200561523437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,2,power_law_1.2,0.8946585845947265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,2,power_law_1.2,1.0071731567382813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,2,power_law_1.2,1.1747122955322264
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,2,2,power_law_1.2,0.059520001411437984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,2,power_law_1.2,1.5284927368164063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,2,power_law_1.2,8.609971313476562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,2,2,power_law_1.2,0.0741644811630249
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,2,2,power_law_1.2,0.0838425636291504
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,2,2,power_law_1.2,0.11493120193481446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,2,power_law_1.2,1.9850738525390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,2,2,power_law_1.2,0.1387660789489746
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,2,2,power_law_1.2,0.17469696044921873
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,2,2,power_law_1.2,0.24497024536132814
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,2,2,power_law_1.2,0.2632844734191894
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,2,2,power_law_1.2,0.25921407699584964
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,2,2,power_law_1.2,0.26551424026489256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,2,power_law_1.2,0.2637311935424805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,2,2,power_law_1.2,0.27248512268066405
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,2,2,power_law_1.2,0.2844044876098633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,2,2,power_law_1.2,0.3042406463623047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,2,2,power_law_1.2,0.3138534355163574
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,2,2,power_law_1.2,0.32651649475097655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,2,power_law_1.2,3.390943908691406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,2,2,power_law_1.2,0.3532684707641601
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,2,2,power_law_1.2,0.28695295333862303
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,2,2,power_law_1.2,0.5457894515991211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,2,power_law_1.2,2.653541259765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,2,2,power_law_1.01,2.884998474121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,2,2,power_law_1.2,0.44689662933349605
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,2,2,power_law_1.2,0.6191628646850587
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,2,2,power_law_1.2,0.8379174041748048
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,2,2,power_law_1.2,1.1319923400878906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,2,2,power_law_1.2,1.5052415466308593
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,2,2,power_law_1.2,1.9884672546386717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,1,balanced,0.08534720420837402
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,1,balanced,0.06577087879180908
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,1,balanced,0.1342815971374512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,1,balanced,0.23280319213867187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,1,balanced,0.4536153411865234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,1,balanced,0.8275405120849608
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,1,balanced,0.8304838562011719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,1,balanced,0.842200927734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,1,balanced,0.8459629058837891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,1,balanced,0.8511980438232423
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,1,balanced,0.870794906616211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,1,balanced,0.8787065887451172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,1,balanced,0.8435366058349609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,1,balanced,0.8446867370605469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,1,balanced,0.8531782531738281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,1,balanced,0.8649062347412109
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,2,2,power_law_1.2,3.264698791503906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,1,balanced,1.0177638244628906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,1,balanced,1.052738571166992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,1,balanced,0.9824915313720703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,1,balanced,1.102845458984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,1,balanced,1.2539628601074218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,1,balanced,1.3245497131347657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,1,balanced,0.05846784114837646
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,1,balanced,0.07603392124176026
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,1,balanced,0.1055065631866455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,1,balanced,0.17685951232910155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,1,balanced,1.4441658020019532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,1,balanced,0.3209836959838867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,1,balanced,0.5946131134033203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,2,2,power_law_1.2,4.454110717773437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,1,balanced,0.5969702529907226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,1,balanced,0.599793930053711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,1,balanced,4.285565490722656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,1,balanced,2.1733631896972656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,1,balanced,0.6028435134887695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,1,balanced,0.6098035049438477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,1,balanced,0.622177276611328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,1,balanced,2.313885498046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,1,balanced,0.6240512084960937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,1,balanced,0.6273414230346679
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,1,balanced,0.687742691040039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,1,balanced,0.6923852539062499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,1,balanced,0.7124524688720703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,1,balanced,0.7216896057128906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,1,balanced,0.7440153503417969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,1,balanced,3.310506896972656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,1,balanced,0.7889170837402344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,1,balanced,0.8831577301025391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,1,balanced,0.905185317993164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,1,balanced,0.9985343933105468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,1,balanced,1.5608921813964844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,1,balanced,0.593884162902832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,1,balanced,1.739976348876953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,1,balanced,2.501921844482422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,1,balanced,3.160641784667969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,4,1,balanced,0.07004032135009766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,4,1,balanced,0.06388991832733154
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,4,1,balanced,0.08883135795593262
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,4,1,balanced,0.12747136116027832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,4,1,balanced,0.3827916717529297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,4,1,balanced,0.20145536422729493
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,4,1,balanced,0.3817830276489258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,4,1,balanced,0.3837830352783203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,4,1,balanced,0.3809132766723633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,4,1,balanced,0.38290241241455075
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,4,1,balanced,0.3868294525146484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,4,1,balanced,0.3875929641723633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,4,1,balanced,0.3900627136230469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,4,1,balanced,0.39299774169921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,4,1,balanced,0.3819219207763672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,4,1,balanced,0.3971488189697266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,4,1,balanced,0.40345985412597657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,4,1,balanced,0.41786880493164064
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,4,1,balanced,0.43053760528564455
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,4,1,balanced,0.45599105834960935
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,4,1,balanced,0.485522575378418
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,4,1,balanced,0.5413254547119141
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,4,1,balanced,0.5923065567016602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,1,power_law_1.01,0.08522560119628905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,1,power_law_1.01,0.23210815429687498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,1,power_law_1.01,0.13320768356323243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,1,power_law_1.01,0.06591231822967529
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,1,power_law_1.01,0.3400115203857422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,1,power_law_1.01,0.45854911804199217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,1,power_law_1.01,0.6603314971923828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,1,power_law_1.01,0.6235212707519532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,1,power_law_1.01,0.6994739532470703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,1,power_law_1.01,0.7561119842529297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,1,power_law_1.01,0.673667221069336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,1,power_law_1.01,0.7666118621826172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,1,power_law_1.01,0.6693011474609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,1,power_law_1.01,0.8581811523437499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,1,power_law_1.01,0.907921905517578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,1,power_law_1.01,0.7907609558105468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,1,power_law_1.01,1.0277254486083984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,1,power_law_1.01,1.255824661254883
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,4,1,balanced,0.8828383636474608
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,1,power_law_1.01,1.348059539794922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,4,1,balanced,1.024659194946289
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,4,1,balanced,1.4823840332031248
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,4,1,balanced,1.9414572143554687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,1,power_law_1.01,1.074640655517578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,1,power_law_1.01,1.7932249450683593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,1,power_law_1.01,0.054897918701171874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,1,power_law_1.01,0.07597760200500489
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,1,power_law_1.01,0.10612095832824706
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,1,power_law_1.01,2.304149169921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,1,power_law_1.01,0.17795200347900392
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,1,power_law_1.01,0.26226432800292965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,1,power_law_1.01,0.3656883239746094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,1,power_law_1.01,2.7328326416015623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,1,power_law_1.01,1.2198982238769531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,1,power_law_1.01,0.46358398437500004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,1,power_law_1.01,0.48997760772705073
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,1,power_law_1.01,0.4956460952758789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,1,power_law_1.01,0.5053580856323242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,1,power_law_1.01,0.5269292831420899
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,1,power_law_1.01,3.7201715087890626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,1,power_law_1.01,0.5359411239624023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,1,power_law_1.01,0.543067512512207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,1,power_law_1.01,0.5524876785278321
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,1,power_law_1.01,0.6197945785522461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,1,power_law_1.01,0.6238227081298828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,1,power_law_1.01,0.6583757019042968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,1,power_law_1.01,0.6819071960449219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,1,power_law_1.01,0.7254201507568359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,1,power_law_1.01,0.8015187072753907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,1,power_law_1.01,4.643413696289063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,1,power_law_1.01,0.8876927947998047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,1,power_law_1.01,1.5348588562011718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,1,power_law_1.01,1.0395584106445312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,1,power_law_1.01,1.222974090576172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,4,1,power_law_1.01,0.0638259220123291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,1,power_law_1.01,1.6038681030273438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,4,1,power_law_1.01,0.09337727546691894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,1,power_law_1.01,1.9545663452148438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,4,1,power_law_1.01,0.12651328086853025
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,4,1,power_law_1.01,0.16711103439331054
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,4,1,power_law_1.01,0.20342975616455078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,1,power_law_1.01,2.667222900390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,4,1,power_law_1.01,0.2827027130126953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,4,1,power_law_1.01,0.29941951751708984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,4,1,power_law_1.01,0.3042873573303223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,1,power_law_1.01,3.3428121948242184
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,4,1,power_law_1.01,0.3111225509643555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,4,1,power_law_1.01,0.32724929809570313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,4,1,power_law_1.01,0.34086593627929684
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,4,1,power_law_1.01,0.3449548721313477
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,4,1,power_law_1.01,0.35485118865966797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,4,1,power_law_1.01,0.3720908737182617
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,4,1,power_law_1.01,0.38159488677978515
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,4,1,power_law_1.01,0.41092353820800775
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,4,1,power_law_1.01,0.45820030212402346
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,4,1,power_law_1.01,0.07676159858703613
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,4,1,power_law_1.01,0.6072051239013672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,4,1,power_law_1.01,0.7194976043701172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,1,power_law_1.2,0.06649343967437744
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,4,1,power_law_1.01,0.9165625762939452
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,4,1,power_law_1.01,0.5124230575561524
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,4,1,power_law_1.01,1.1425542449951172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,1,power_law_1.2,0.13266048431396485
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,4,1,power_law_1.01,1.6533721923828124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,1,power_law_1.2,0.21740800857543946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,1,power_law_1.2,0.6166432189941407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,4,1,power_law_1.01,2.0696479797363283
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,1,power_law_1.2,0.6489286041259765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,1,power_law_1.2,0.0854201602935791
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,1,power_law_1.2,0.672380142211914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,1,power_law_1.2,0.6998438262939454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,1,power_law_1.2,0.7497792053222656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,1,power_law_1.2,0.3182784080505371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,1,power_law_1.2,0.4419532775878906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,1,power_law_1.2,0.8387500762939453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,1,power_law_1.2,0.8712531280517577
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,1,power_law_1.2,0.9355955505371094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,1,power_law_1.2,0.6631673431396484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,1,power_law_1.2,1.3105670166015624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,1,power_law_1.2,1.117061767578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,1,power_law_1.2,0.76700927734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,1,power_law_1.2,1.407024688720703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,1,power_law_1.2,1.6118476867675782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,1,power_law_1.2,1.9249658203125002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,1,power_law_1.2,1.0702816009521485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,1,power_law_1.2,2.324753875732422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,1,power_law_1.2,2.877445068359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,1,power_law_1.2,0.05118080139160156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,1,power_law_1.2,0.07676415920257569
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,1,power_law_1.2,0.10525247573852539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,1,power_law_1.2,1.2715020751953126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,1,power_law_1.2,0.16794815063476562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,1,power_law_1.2,0.2495315170288086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,1,power_law_1.2,0.3466604614257812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,1,power_law_1.2,0.4577036666870117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,1,power_law_1.2,0.48119487762451174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,1,power_law_1.2,3.8161547851562503
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,1,power_law_1.2,0.4917440032958984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,1,power_law_1.2,0.5037721633911133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,1,power_law_1.2,0.5184211349487304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,1,power_law_1.2,0.5310950469970703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,1,power_law_1.2,0.5468838500976563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,1,power_law_1.2,0.5632102584838867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,1,power_law_1.2,0.6324563217163086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,1,power_law_1.2,0.6517311859130859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,1,power_law_1.2,0.6633548736572266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,1,power_law_1.2,0.7110623931884765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,1,power_law_1.2,4.723496398925781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,1,power_law_1.2,0.7523212432861328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,1,power_law_1.2,0.8430579376220703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,1,power_law_1.2,0.9058969879150391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,1,power_law_1.2,1.100806427001953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,1,power_law_1.2,1.2845362854003906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,1,power_law_1.2,1.6501196289062499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,1,power_law_1.2,2.0056723022460936
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,4,1,power_law_1.2,0.06357376098632812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,4,1,power_law_1.2,0.0778278398513794
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,4,1,power_law_1.2,0.09109888076782227
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,4,1,power_law_1.2,0.121015043258667
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,4,1,power_law_1.2,0.1557196807861328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,1,power_law_1.2,2.7034585571289065
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,4,1,power_law_1.2,0.19585792541503905
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,4,1,power_law_1.2,0.27638271331787106
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,1,power_law_1.2,3.3776846313476563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,4,1,power_law_1.2,0.29570240020751953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,4,1,power_law_1.2,0.3007040023803711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,4,1,power_law_1.2,0.3129696083068848
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,4,1,power_law_1.2,0.32596096038818356
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,4,1,power_law_1.2,0.3508927917480469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,4,1,power_law_1.2,0.3390707015991211
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,4,1,power_law_1.2,0.3692377471923828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,4,1,power_law_1.2,0.38940479278564455
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,4,1,power_law_1.01,3.008594665527344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,4,1,power_law_1.2,0.40642753601074216
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,4,1,power_law_1.01,4.076962585449219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,4,1,power_law_1.2,0.443040657043457
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,4,1,power_law_1.2,0.5161331176757813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,4,1,power_law_1.2,0.7463577270507813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,4,1,power_law_1.2,0.5840396881103516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,4,balanced,0.07246719837188721
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,4,balanced,0.08615167617797852
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,4,1,power_law_1.2,1.250491485595703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,4,balanced,0.12431872367858887
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,4,1,power_law_1.2,0.8918681335449218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,4,balanced,0.4599078369140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,4,balanced,0.8200153350830078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,4,1,power_law_1.2,1.6165536499023438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,4,balanced,0.8222246551513672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,4,balanced,0.8328025817871094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,4,balanced,0.8867161560058594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,4,balanced,0.9018598175048828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,4,balanced,0.9044134521484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,4,balanced,0.9058354949951172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,4,balanced,0.9225574493408203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,4,balanced,0.9300390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,4,balanced,0.9364582061767578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,4,balanced,0.9565401458740235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,4,balanced,0.9761663818359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,4,balanced,1.0069171142578124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,4,balanced,1.0682355499267577
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,4,balanced,1.1582579040527343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,4,balanced,1.2668402862548827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,4,balanced,1.8651878356933593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,4,balanced,2.065154571533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,4,1,power_law_1.2,2.4171328735351563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,4,balanced,2.9569241333007814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,4,balanced,0.03986176013946533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,4,balanced,0.05633408069610596
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,4,balanced,0.07868288040161134
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,4,balanced,0.26076799392700195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,4,balanced,0.13652607917785645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,4,balanced,3.805050964355469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,4,balanced,0.4201152038574219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,4,balanced,0.42721664428710937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,4,balanced,0.4229888153076172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,4,balanced,0.24049280166625975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,4,balanced,0.45529857635498044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,4,balanced,0.4553433609008789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,4,balanced,0.4570111846923828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,4,balanced,0.4593280029296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,4,balanced,0.4607551956176758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,4,balanced,0.4664934539794922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,4,balanced,0.46797054290771484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,4,balanced,0.4722137451171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,4,balanced,0.48579841613769525
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,4,balanced,0.5333990478515626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,4,balanced,0.5382924652099609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,4,balanced,6.104707641601562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,4,balanced,0.6012531280517578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,4,balanced,0.6625215911865234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,4,balanced,0.7231423950195313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,4,balanced,0.8757926177978514
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,4,balanced,8.259990844726563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,4,balanced,1.232642593383789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,4,balanced,1.4296946716308594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,4,balanced,2.0875852966308592
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,4,balanced,2.6438516235351566
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,4,1,power_law_1.2,3.260115051269531
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,1,4,balanced,0.037552640438079835
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,1,4,balanced,0.03085311889648438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,1,4,balanced,0.05346047878265381
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,1,4,balanced,0.08155648231506347
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,1,4,balanced,0.13632512092590332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,1,4,balanced,0.24935935974121093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,1,4,balanced,0.2501555252075195
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,1,4,balanced,0.2515135955810547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,1,4,balanced,0.25108736038208007
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,1,4,balanced,0.25143680572509763
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,1,4,balanced,0.2513088035583496
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,1,4,balanced,0.2523379135131836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,1,4,balanced,0.25409023284912113
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,1,4,balanced,0.2547251129150391
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,1,4,balanced,0.25492736816406253
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,1,4,balanced,0.2580428886413574
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,1,4,balanced,0.26291967391967774
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,1,4,balanced,0.2675686454772949
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,1,4,balanced,0.27602176666259765
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,1,4,balanced,0.28402816772460937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,1,4,balanced,0.30034303665161133
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,1,4,balanced,0.3177587127685547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,4,power_law_1.01,0.12003840446472167
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,4,power_law_1.01,0.25957632064819336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,4,power_law_1.01,0.32024192810058594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,4,power_law_1.01,0.34328575134277345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,4,power_law_1.01,0.47395198822021484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,4,power_law_1.01,0.5156556701660155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,4,power_law_1.01,0.6273036956787109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,4,power_law_1.01,0.6642867279052734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,4,power_law_1.01,0.7692671966552734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,4,power_law_1.01,0.7814527893066406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,4,power_law_1.01,0.8192639923095703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,4,power_law_1.01,0.8326156616210938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,1,4,balanced,0.2500057601928711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,4,power_law_1.01,0.8454450988769532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,4,power_law_1.01,0.8607897949218749
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,4,power_law_1.01,0.8667046356201172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,4,power_law_1.01,0.9203072357177735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,4,power_law_1.01,0.9507968139648437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,4,power_law_1.01,1.0498802947998047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,4,power_law_1.01,1.1630233764648437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,4,power_law_1.01,1.41085693359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,4,power_law_1.01,1.6754637145996092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,4,power_law_1.01,2.167741394042969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,4,power_law_1.01,2.6698202514648437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,4,power_law_1.01,3.6176882934570314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,4,power_law_1.01,4.655469970703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,4,power_law_1.01,6.639828491210937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,1,4,balanced,0.46816638946533207
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,1,4,balanced,0.5120230484008789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,1,4,balanced,0.7174451446533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,4,power_law_1.01,0.0800153636932373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,4,power_law_1.01,0.13017600059509277
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,4,power_law_1.01,0.17069952011108397
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,4,power_law_1.01,0.22870271682739257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,4,power_law_1.01,0.2578508758544922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,4,power_law_1.01,0.2944895935058594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,4,power_law_1.01,0.3216896057128906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,4,power_law_1.01,0.3555020904541016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,1,4,balanced,0.9158822631835937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,4,power_law_1.01,0.35975040435791017
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,4,power_law_1.01,0.3761100769042969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,4,power_law_1.01,0.384192008972168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,4,power_law_1.01,0.40203903198242186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,4,power_law_1.01,0.3880998229980469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,4,power_law_1.01,0.41634048461914064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,4,power_law_1.01,0.43593982696533207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,4,power_law_1.01,0.44678398132324215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,4,power_law_1.01,0.5008806228637696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,4,power_law_1.01,0.5348287963867188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,4,power_law_1.01,0.6010188674926757
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,4,power_law_1.01,0.6714214324951172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,4,power_law_1.01,0.8596800231933595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,4,power_law_1.01,1.0202662658691406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,4,power_law_1.01,8.743414916992188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,1,4,power_law_1.01,0.0584665584564209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,4,power_law_1.01,1.4165849304199218
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,1,4,power_law_1.01,0.08698111534118652
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,1,4,power_law_1.01,0.10976384162902833
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,4,power_law_1.01,1.738727722167969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,1,4,power_law_1.01,0.12490367889404297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,1,4,power_law_1.01,0.15445247650146485
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,1,4,power_law_1.01,0.17432064056396485
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,1,4,power_law_1.01,0.21015935897827148
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,1,4,power_law_1.01,0.21113727569580076
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,1,4,power_law_1.01,0.22394752502441406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,4,power_law_1.01,2.4615065002441403
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,1,4,power_law_1.01,0.20366079330444337
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,1,4,power_law_1.01,0.217205753326416
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,1,4,power_law_1.01,0.2112588882446289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,4,power_law_1.01,3.010969543457031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,1,4,power_law_1.01,0.22368000030517576
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,1,4,power_law_1.01,0.23409408569335938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,1,4,power_law_1.01,0.23932544708251954
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,1,4,power_law_1.01,0.24545919418334963
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,4,power_law_1.01,0.3865395355224609
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,1,4,power_law_1.01,0.24784255981445313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,1,4,power_law_1.01,0.2902976036071777
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,1,4,power_law_1.01,0.3310156631469726
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,1,4,power_law_1.01,0.4948070526123047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,1,4,power_law_1.01,0.27358591079711914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,1,4,power_law_1.01,0.3801599884033203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,4,power_law_1.2,0.12084863662719728
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,1,4,power_law_1.01,0.5988224029541016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,4,power_law_1.2,0.23556863784790041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,4,power_law_1.2,0.24749055862426758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,4,power_law_1.2,0.33973888397216795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,4,power_law_1.2,0.4559436798095703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,4,power_law_1.2,0.4578931045532227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,4,power_law_1.2,0.6498252868652343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,4,power_law_1.2,0.6625663757324218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,4,power_law_1.2,0.7434982299804688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,4,power_law_1.2,0.7782975769042968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,4,power_law_1.2,0.8268748474121093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,4,power_law_1.2,0.8436006164550781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,4,power_law_1.2,0.8592537689208986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,4,power_law_1.2,0.8370508575439454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,4,power_law_1.2,0.884136962890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,4,power_law_1.2,0.9052249908447265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,4,power_law_1.2,1.0116710662841797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,4,power_law_1.2,1.115445785522461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,4,power_law_1.2,1.2116748809814453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,4,power_law_1.2,1.4709017944335936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,4,power_law_1.2,1.7353126525878906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,4,power_law_1.2,2.2031948852539065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,4,power_law_1.2,5.256687622070312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,4,power_law_1.2,7.139465942382813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,4,power_law_1.2,0.08039551734924316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,4,power_law_1.2,0.12975104331970216
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,4,power_law_1.2,0.1410214424133301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,4,power_law_1.2,0.18767871856689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,4,power_law_1.2,0.229881591796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,4,power_law_1.2,0.2462361526489258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,4,power_law_1.2,2.826087646484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,4,power_law_1.2,0.31862783432006836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,4,power_law_1.2,0.351907844543457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,4,power_law_1.2,3.9181338500976564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,4,power_law_1.2,0.34521343231201174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,4,power_law_1.2,0.36019966125488284
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,4,power_law_1.2,0.38270847320556645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,4,power_law_1.2,0.40985984802246095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,4,power_law_1.2,0.38927871704101563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,4,power_law_1.2,0.41474048614501957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,4,power_law_1.2,0.42750209808349615
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,4,power_law_1.2,0.4325337600708008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,4,power_law_1.2,0.4639871978759766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,4,power_law_1.2,0.5294438552856445
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,1,4,power_law_1.01,0.8110784149169922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,4,power_law_1.2,10.228537597656251
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,1,4,power_law_1.01,1.0818982696533204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,4,power_law_1.2,0.6246720123291015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,1,4,power_law_1.01,1.4117567443847656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,4,power_law_1.2,0.7194982147216796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,4,power_law_1.2,0.8916134643554688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,4,power_law_1.2,0.5615744018554688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,1,4,power_law_1.01,1.7441317749023437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,1,4,power_law_1.2,0.05859839916229248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,4,power_law_1.2,1.124000015258789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,4,power_law_1.2,1.4357452392578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,1,4,power_law_1.2,0.11972479820251465
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,1,4,power_law_1.2,0.09329024314880371
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,1,4,power_law_1.2,0.13651583671569825
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,1,4,power_law_1.2,0.14317055702209475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,4,power_law_1.2,1.791535339355469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,1,4,power_law_1.2,0.20033920288085935
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,1,4,power_law_1.2,0.2081804847717285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,1,4,power_law_1.2,0.20555135726928714
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,1,4,power_law_1.2,0.22098175048828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,1,4,power_law_1.2,0.22769664764404296
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,1,4,power_law_1.2,0.24170623779296876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,1,4,power_law_1.2,0.24641408920288085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,4,power_law_1.2,2.5009689331054688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,1,4,power_law_1.2,0.214520320892334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,1,4,power_law_1.2,0.2575347137451172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,1,4,power_law_1.2,0.2623219108581543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,1,4,power_law_1.2,0.32935039520263676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,4,power_law_1.2,3.4373016357421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,1,4,power_law_1.2,0.1922393608093262
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,1,4,power_law_1.2,0.08727168083190918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,1,4,power_law_1.2,0.38015872955322266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,1,4,power_law_1.2,0.3002943992614746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,2,2,balanced,0.06592127799987793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,1,4,power_law_1.2,0.47865985870361333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,2,2,balanced,0.08252799987792969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,1,4,power_law_1.2,0.7881881713867187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,1,4,power_law_1.2,0.6297868728637696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,2,2,balanced,0.12460032463073731
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,2,2,balanced,0.4505702209472656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,2,2,balanced,0.8137177276611329
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,2,2,balanced,0.8143270111083984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,1,4,power_law_1.2,1.4474688720703124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,2,2,balanced,0.8176204681396484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,2,2,balanced,0.8187123107910157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,1,4,power_law_1.2,1.1042918395996093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,2,2,balanced,0.8210176086425781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,2,2,balanced,0.8299839782714844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,2,2,balanced,0.8958016204833985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,2,2,balanced,0.8982579040527344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,2,2,balanced,0.25571327209472655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,2,2,balanced,0.9175872039794921
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,2,2,balanced,0.9243686676025391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,2,2,balanced,0.9045439910888671
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,2,2,balanced,0.9408499145507813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,2,2,balanced,0.98136962890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,2,2,balanced,1.0240998077392578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,2,2,balanced,1.1197644805908202
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,2,2,balanced,1.2001267242431641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,2,2,balanced,1.7905754089355468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,2,2,balanced,1.9999859619140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,2,2,balanced,0.05944704055786133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,2,2,balanced,0.08057472229003906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,2,2,balanced,0.14001279830932617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,2,2,balanced,2.806612548828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,2,2,balanced,0.25336063385009766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,2,2,balanced,0.4397708892822266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,2,2,balanced,0.0397158408164978
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,2,2,balanced,3.4713113403320315
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,2,2,balanced,0.44094081878662106
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,2,2,balanced,0.4427328109741211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,2,2,balanced,0.44696704864501957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,2,2,balanced,0.45558273315429687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,2,2,balanced,0.45144832611083985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,2,2,balanced,0.4852467346191406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,2,2,balanced,0.48858879089355467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,2,2,balanced,0.4951475143432617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,2,2,balanced,0.4958899307250976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,2,2,balanced,5.142998046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,2,2,balanced,0.5003430557250976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,2,2,balanced,0.509062385559082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,2,2,balanced,0.5282035064697266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,2,2,balanced,0.5633472061157226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,2,2,balanced,0.589958381652832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,2,2,balanced,0.6504691314697266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,2,2,balanced,0.7340147399902344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,2,2,balanced,0.849560317993164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,2,2,balanced,6.785592041015624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,2,2,balanced,1.2268326568603516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,2,2,balanced,1.4140582275390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,2,2,balanced,2.0311424255371096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,2,2,balanced,2.6021694946289062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,1,4,power_law_1.2,2.04055419921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,1,4,power_law_1.2,2.7771084594726565
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,2,2,balanced,0.04041600227355957
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,2,2,balanced,0.062080001831054686
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,2,2,balanced,0.2680844879150391
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,2,2,balanced,0.2674675178527832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,2,2,balanced,0.2702783966064453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,2,2,balanced,0.27129472732543947
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,2,2,balanced,0.2716364860534668
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,2,2,balanced,0.27329919815063475
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,2,2,balanced,0.04665215969085693
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,2,2,balanced,0.2740902328491211
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,2,2,balanced,0.14654848098754883
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,2,2,balanced,0.2679334449768066
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,2,2,balanced,0.2802739143371582
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,2,2,balanced,0.08983936309814453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,2,2,balanced,0.29361280441284177
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,2,2,balanced,0.3068300819396973
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,2,2,balanced,0.31991167068481446
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,2,2,balanced,0.3462579345703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,2,2,balanced,0.3727347183227539
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,2,2,balanced,0.2675660705566406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,2,2,balanced,0.26877695083618164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,2,2,power_law_1.01,0.12307328224182128
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,2,2,power_law_1.01,0.19164159774780273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,2,2,power_law_1.01,0.2788889694213867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,2,2,power_law_1.01,0.3655769729614258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,2,2,power_law_1.01,0.4772825622558594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,2,2,power_law_1.01,0.6188313674926758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,2,2,power_law_1.01,0.647842559814453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,2,2,power_law_1.01,0.6720333099365234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,2,2,power_law_1.01,0.08262784004211425
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,2,2,power_law_1.01,0.6847385406494141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,2,2,power_law_1.01,0.7003814697265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,2,2,power_law_1.01,0.81338623046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,2,2,balanced,0.5559193420410156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,2,2,balanced,0.6204275131225586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,2,2,power_law_1.01,0.8193933105468749
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,2,2,power_law_1.01,0.8190374755859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,2,2,power_law_1.01,0.8304217529296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,2,2,balanced,0.8777894592285156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,2,2,balanced,1.164232940673828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,2,2,power_law_1.01,0.9035609436035156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,2,2,power_law_1.01,0.9987852478027344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,2,2,power_law_1.01,0.8606963348388671
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,2,2,power_law_1.01,1.06980224609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,2,2,power_law_1.01,1.3155314636230468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,2,2,power_law_1.01,1.5108966064453124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,2,2,power_law_1.01,1.9072665405273437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,2,2,power_law_1.01,2.325856018066406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,2,2,power_law_1.01,0.07982207775115967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,2,2,power_law_1.01,3.101226196289063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,2,2,power_law_1.01,0.12009216308593751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,2,2,power_law_1.01,0.16288511276245116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,2,2,power_law_1.01,0.2365785598754883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,2,2,power_law_1.01,0.2851968002319336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,2,2,power_law_1.01,4.069888000488281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,2,2,power_law_1.01,0.3620275115966797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,2,2,power_law_1.01,5.653228759765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,2,2,power_law_1.01,0.36990718841552733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,2,2,power_law_1.01,0.3768153762817383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,2,2,power_law_1.01,0.39003902435302734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,2,2,power_law_1.01,0.056605439186096194
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,2,2,power_law_1.01,0.43111934661865237
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,2,2,power_law_1.01,0.43680896759033205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,2,2,power_law_1.01,0.4525683212280273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,2,2,power_law_1.01,0.37626369476318355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,2,2,power_law_1.01,0.46858753204345704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,2,2,power_law_1.01,7.734945068359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,2,2,power_law_1.01,0.470274543762207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,2,2,power_law_1.01,0.43789825439453123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,2,2,power_law_1.01,0.54953857421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,2,2,power_law_1.01,0.6049075317382813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,2,2,power_law_1.01,0.6982438659667969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,2,2,power_law_1.01,0.5041471862792969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,2,2,power_law_1.01,0.8137843322753906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,2,2,power_law_1.01,0.9563865661621094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,2,2,power_law_1.01,0.04845568180084229
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,2,2,power_law_1.01,0.06861184120178224
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,2,2,power_law_1.01,0.07942399978637696
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,2,2,power_law_1.01,0.10971136093139648
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,2,2,power_law_1.01,0.12936448097229003
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,2,2,power_law_1.01,0.15780480384826662
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,2,2,power_law_1.01,1.6672460937499998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,2,2,power_law_1.01,1.2808985900878906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,2,2,power_law_1.01,0.20972415924072269
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,2,2,power_law_1.01,0.21033088684082032
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,2,2,power_law_1.01,0.23417600631713867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,2,2,power_law_1.01,0.22634496688842773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,2,2,power_law_1.01,2.243189697265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,2,2,power_law_1.01,0.22877824783325193
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,2,2,power_law_1.01,0.23302656173706054
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,2,2,power_law_1.01,0.24067199707031248
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,2,2,power_law_1.01,0.24925695419311525
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,2,2,power_law_1.01,0.26274560928344726
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,2,2,power_law_1.01,0.2640627288818359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,2,2,power_law_1.01,2.8043878173828127
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,2,2,power_law_1.01,0.2824320030212403
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,2,2,power_law_1.01,0.39991680145263675
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,2,2,power_law_1.01,0.311148796081543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,2,2,power_law_1.01,0.45641983032226563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,2,2,power_law_1.01,0.3478656005859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,2,2,power_law_1.01,0.5836710357666016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,2,2,power_law_1.01,0.7537113952636718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,2,2,power_law_1.2,0.08317184448242188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,2,2,power_law_1.2,0.12051199913024901
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,2,2,power_law_1.2,0.17095039367675782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,2,2,power_law_1.2,0.27168384552001956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,2,2,power_law_1.2,0.3457459259033203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,2,2,power_law_1.2,0.4311539077758789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,2,2,power_law_1.2,0.6347929763793945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,2,2,power_law_1.2,0.6436300659179688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,2,2,power_law_1.2,0.6534310150146484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,2,2,power_law_1.2,0.6739494323730468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,2,2,power_law_1.2,0.7287270355224609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,2,2,power_law_1.2,0.8164851379394531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,2,2,power_law_1.2,0.8188992309570313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,2,2,power_law_1.2,0.8468723297119141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,2,2,power_law_1.2,0.8698957061767578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,2,2,power_law_1.2,0.8948812866210938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,2,2,power_law_1.2,0.9616588592529297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,2,2,power_law_1.2,1.0396864318847656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,2,2,power_law_1.2,1.1382054138183595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,2,2,power_law_1.2,1.3856486511230468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,2,2,power_law_1.2,1.582434539794922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,2,2,power_law_1.2,2.048810272216797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,2,2,power_law_1.2,2.446210632324219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,2,2,power_law_1.2,3.3175399780273436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,2,2,power_law_1.2,4.1256384277343745
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,2,2,power_law_1.2,6.007830810546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,2,2,power_law_1.2,0.05753215789794922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,2,2,power_law_1.2,0.07991168022155762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,2,2,power_law_1.2,0.11062399864196777
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,2,2,power_law_1.2,0.1468851184844971
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,2,2,power_law_1.2,0.21146368026733398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,2,2,power_law_1.2,0.2629209518432617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,2,2,power_law_1.2,0.35041919708251956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,2,2,power_law_1.2,8.051436767578124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,2,2,power_law_1.2,0.36254974365234377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,2,2,power_law_1.2,0.3681190490722656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,2,2,power_law_1.2,0.38728191375732424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,2,2,power_law_1.2,0.38780670166015624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,2,2,power_law_1.01,1.0057266998291017
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,2,2,power_law_1.2,0.4272563171386719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,2,2,power_law_1.01,1.335188446044922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,2,2,power_law_1.2,0.4377856063842773
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,2,2,power_law_1.01,1.8382374572753908
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,2,2,power_law_1.2,0.4515865707397461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,2,2,power_law_1.2,0.4544371032714844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,2,2,power_law_1.2,0.4705791854858399
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,2,2,power_law_1.2,0.4958220672607422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,2,2,power_law_1.2,0.5199462509155273
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,2,2,power_law_1.01,2.61623046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,2,2,power_law_1.2,0.5686630249023438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,2,2,power_law_1.2,0.6278054428100586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,2,2,power_law_1.2,0.7165030670166016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,2,2,power_law_1.2,0.8667148590087891
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,2,2,power_law_1.2,0.05007487773895264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,2,2,power_law_1.2,1.0041407775878906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,2,2,power_law_1.2,1.3477658081054689
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,2,2,power_law_1.2,1.6783436584472657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,2,2,power_law_1.2,0.07689472198486327
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,2,2,power_law_1.2,0.06761600017547607
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,2,2,power_law_1.2,0.0976576042175293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,2,2,power_law_1.2,0.1251020812988281
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,2,2,power_law_1.2,0.1511347198486328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,2,2,power_law_1.2,0.2036441612243652
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,2,2,power_law_1.2,0.20845951080322264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,2,2,power_law_1.2,2.3123315429687503
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,2,2,power_law_1.2,0.215600643157959
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,2,2,power_law_1.2,0.2222796821594238
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,2,2,power_law_1.2,0.23051647186279295
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,2,2,power_law_1.2,0.24747648239135742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,2,2,power_law_1.2,2.917918701171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,2,2,power_law_1.2,0.25860479354858396
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,2,2,power_law_1.2,0.26878976821899414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,2,2,power_law_1.2,0.27715072631835935
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,2,2,power_law_1.2,0.23749248504638673
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,2,2,power_law_1.2,0.47767807006835933
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,2,2,power_law_1.2,0.37574783325195316
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,2,2,power_law_1.2,0.570967025756836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,2,2,power_law_1.2,0.7918771362304687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,4,1,balanced,0.0599948787689209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,4,1,balanced,0.07672895908355712
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,2,2,power_law_1.2,1.0366950225830078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,4,1,balanced,0.20392576217651368
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,4,1,balanced,0.11950464248657226
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,2,2,power_law_1.2,0.2963494491577149
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,4,1,balanced,0.7122854614257813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,4,1,balanced,0.7251763153076172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,4,1,balanced,0.7289389038085938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,4,1,balanced,0.39232959747314455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,4,1,balanced,0.7239315032958984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,4,1,balanced,0.7105792236328126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,4,1,balanced,0.7326290893554688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,4,1,balanced,0.7387468719482422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,4,1,balanced,0.7450182342529297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,4,1,balanced,0.7513862609863281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,4,1,balanced,0.7549906921386719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,4,1,balanced,0.8541580963134766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,4,1,balanced,0.727306900024414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,4,1,balanced,0.9056678771972656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,2,2,power_law_1.2,1.4581376647949218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,4,1,balanced,0.8988140869140626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,4,1,balanced,0.9404025268554687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,4,1,balanced,1.1445458984374999
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,4,1,balanced,1.2461817932128905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,4,1,balanced,1.0724556732177735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,4,1,balanced,1.8746847534179687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,4,1,balanced,0.05603839874267578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,4,1,balanced,2.0358770751953124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,4,1,balanced,0.07025919914245606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,4,1,balanced,0.15946432113647463
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,2,2,power_law_1.2,2.8475250244140624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,4,1,balanced,0.28213504791259764
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,2,2,power_law_1.2,1.9598629760742188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,4,1,balanced,0.5185331344604492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,4,1,balanced,2.9045022583007816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,4,1,balanced,0.5185676956176758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,4,1,balanced,0.09777919769287109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,4,1,balanced,0.5223859024047852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,4,1,balanced,0.5227167892456055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,4,1,balanced,0.5290892791748047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,4,1,balanced,0.5274431991577149
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,4,1,balanced,3.709443969726563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,4,1,balanced,0.5329081726074218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,4,1,balanced,0.5403545761108399
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,4,1,balanced,0.5455539321899414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,4,1,balanced,0.598047981262207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,4,1,balanced,0.6014822387695313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,4,1,balanced,0.6206867218017578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,4,1,balanced,0.6476716613769531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,4,1,balanced,0.6828749084472656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,4,1,balanced,0.7174150085449218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,4,1,balanced,0.8075296020507812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,2,2,power_law_1.2,4.138067321777344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,4,1,balanced,0.6286118316650391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,4,1,balanced,1.33271484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,4,1,balanced,1.4936965942382812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,4,1,balanced,0.8820275115966798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,4,1,balanced,2.186321258544922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,4,1,balanced,2.758844909667969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,4,1,balanced,0.06212287902832031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,4,1,balanced,0.06770944118499755
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,4,1,balanced,0.08268095970153808
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,4,1,balanced,0.11446144104003905
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,4,1,balanced,0.3301433563232422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,4,1,balanced,0.326033935546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,4,1,balanced,0.3300556945800781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,4,1,balanced,0.17734079360961913
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,4,1,balanced,0.32943870544433596
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,4,1,balanced,0.3323238372802734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,4,1,balanced,0.33567615509033205
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,4,1,balanced,0.3355379104614258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,4,1,balanced,0.3279296112060547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,4,1,balanced,0.32826881408691405
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,4,1,balanced,0.3395801544189453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,4,1,balanced,0.3407417678833008
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,4,1,balanced,0.34687999725341795
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,4,1,balanced,0.3587360000610352
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,4,1,balanced,0.3707040023803711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,4,1,balanced,0.3933158493041992
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,4,1,balanced,0.41620544433593754
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,4,1,balanced,0.4614361572265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,4,1,balanced,0.5077996826171874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,4,1,power_law_1.01,0.0793452787399292
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,4,1,power_law_1.01,0.11927743911743165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,4,1,power_law_1.01,0.29696063995361327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,4,1,power_law_1.01,0.39848449707031247
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,4,1,power_law_1.01,0.5360345458984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,4,1,power_law_1.01,0.2045395278930664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,4,1,power_law_1.01,0.06312191963195801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,4,1,power_law_1.01,0.5823007965087891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,4,1,power_law_1.01,0.6065670394897461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,4,1,power_law_1.01,0.6454975891113282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,4,1,power_law_1.01,0.5694521713256836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,4,1,power_law_1.01,0.6601248168945313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,4,1,power_law_1.01,0.6738893127441405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,4,1,power_law_1.01,0.7398348999023437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,4,1,power_law_1.01,0.7768038177490235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,4,1,power_law_1.01,0.5773747253417969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,4,1,power_law_1.01,1.058383331298828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,4,1,power_law_1.01,1.0570880126953124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,4,1,power_law_1.01,1.1566976165771483
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,4,1,power_law_1.01,0.8736972808837891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,4,1,power_law_1.01,1.3608729553222656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,4,1,power_law_1.01,1.5445202636718751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,4,1,power_law_1.01,0.9125997161865234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,4,1,power_law_1.01,1.9721932983398438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,4,1,power_law_1.01,2.4096173095703124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,4,1,power_law_1.01,0.05084159851074219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,4,1,power_law_1.01,0.06923583984375001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,4,1,power_law_1.01,3.1643743896484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,4,1,power_law_1.01,0.09686079978942871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,4,1,power_law_1.01,0.16295103073120115
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,4,1,power_law_1.01,0.23077503204345703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,4,1,balanced,0.7634310150146485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,4,1,power_law_1.01,0.31724607467651367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,4,1,power_law_1.01,4.021007995605468
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,4,1,balanced,0.8886438751220703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,4,1,power_law_1.01,0.43357761383056637
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,4,1,balanced,1.271173095703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,4,1,power_law_1.01,0.4325747299194336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,4,1,balanced,1.7182995605468752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,4,1,power_law_1.01,0.4400275039672851
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,4,1,power_law_1.01,0.45706241607666015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,4,1,power_law_1.01,0.4034342575073242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,4,1,power_law_1.01,0.4668467330932617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,4,1,power_law_1.01,0.4738163375854493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,4,1,power_law_1.01,0.4805081558227539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,4,1,power_law_1.01,0.5415596771240234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,4,1,power_law_1.01,0.5436294555664063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,4,1,power_law_1.01,0.5701132965087891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,4,1,power_law_1.01,0.5928672027587891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,4,1,power_law_1.01,0.6286988830566406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,4,1,power_law_1.01,0.6960371398925782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,4,1,power_law_1.01,0.7660575866699219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,4,1,power_law_1.01,0.9077235412597655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,4,1,power_law_1.01,1.0665721893310547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,4,1,power_law_1.01,1.3624153137207031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,4,1,power_law_1.01,0.060752000808715824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,4,1,power_law_1.01,1.7083602905273438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,4,1,power_law_1.01,0.07325376033782958
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,4,1,power_law_1.01,0.08805631637573243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,4,1,power_law_1.01,2.2904920959472657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,4,1,power_law_1.01,0.11496383666992187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,4,1,power_law_1.01,0.14942591667175292
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,4,1,power_law_1.01,0.17893375396728514
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,4,1,power_law_1.01,0.24284927368164064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,4,1,power_law_1.01,2.912389831542969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,4,1,power_law_1.01,0.25610431671142575
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,4,1,power_law_1.01,0.2617184066772461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,4,1,power_law_1.01,0.265914249420166
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,4,1,power_law_1.01,0.2815078353881836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,4,1,power_law_1.01,0.29095680236816407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,4,1,power_law_1.01,0.2978291130065918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,4,1,power_law_1.01,0.3050611114501953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,4,1,power_law_1.01,0.31986623764038086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,4,1,power_law_1.01,0.3298419189453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,4,1,power_law_1.01,0.35301952362060546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,4,1,power_law_1.01,0.39096000671386716
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,4,1,power_law_1.01,0.44020225524902346
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,4,1,power_law_1.01,0.624136962890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,4,1,power_law_1.01,0.80176513671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,4,1,power_law_1.01,0.5267340850830078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,4,1,power_law_1.2,0.11896384239196778
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,4,1,power_law_1.01,0.9791744232177735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,4,1,power_law_1.2,0.19095232009887694
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,4,1,power_law_1.2,0.27209791183471677
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,4,1,power_law_1.2,0.3842969512939453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,4,1,power_law_1.2,0.5326009750366211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,4,1,power_law_1.2,0.5560787200927735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,4,1,power_law_1.2,0.5714412689208984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,4,1,power_law_1.01,1.4009164428710936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,4,1,power_law_1.2,0.6024364852905274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,4,1,power_law_1.2,0.6411891174316406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,4,1,power_law_1.2,0.6613568115234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,4,1,power_law_1.2,0.7137773132324219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,4,1,power_law_1.2,0.7409446716308594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,4,1,power_law_1.2,0.8031737518310547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,4,1,power_law_1.2,0.9309439849853515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,4,1,power_law_1.2,1.1000729370117188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,4,1,power_law_1.2,0.5811884689331055
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,4,1,power_law_1.01,1.7957081604003906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,4,1,power_law_1.2,1.0943014526367187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,4,1,power_law_1.2,1.1984070587158204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,4,1,power_law_1.2,1.6232012939453124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,4,1,power_law_1.2,1.9921554565429687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,4,1,power_law_1.2,0.0780505609512329
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,4,1,power_law_1.2,1.3807084655761719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,4,1,power_law_1.2,0.059888000488281246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,4,1,power_law_1.2,2.451588439941406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,4,1,power_law_1.2,0.9267584228515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,4,1,power_law_1.2,0.15185536384582518
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,4,1,power_law_1.2,0.2185696029663086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,4,1,power_law_1.2,0.06967999935150146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,4,1,power_law_1.2,3.248194580078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,4,1,power_law_1.2,0.30320575714111325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,4,1,power_law_1.2,0.394730224609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,4,1,power_law_1.2,0.04477375984191895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,4,1,power_law_1.2,0.4191091156005859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,4,1,power_law_1.2,0.4267827224731445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,4,1,power_law_1.2,0.438328971862793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,4,1,power_law_1.2,0.44946624755859377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,4,1,power_law_1.2,4.0982470703125005
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,4,1,power_law_1.2,0.4603078460693359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,4,1,power_law_1.2,0.47282302856445313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,4,1,power_law_1.2,0.48898944854736326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,4,1,power_law_1.2,0.5527961730957032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,4,1,power_law_1.2,0.5655750274658203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,4,1,power_law_1.2,0.5771833419799804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,4,1,power_law_1.2,0.6174745559692383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,4,1,power_law_1.2,0.6551091003417968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,4,1,power_law_1.2,0.7306092834472656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,4,1,power_law_1.2,0.7907654571533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,4,1,power_law_1.2,0.9484416198730468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,4,1,power_law_1.2,1.108954849243164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,4,1,power_law_1.2,0.0971564769744873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,4,1,power_law_1.2,1.3902239990234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,4,1,power_law_1.2,0.06157760143280029
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,4,1,power_law_1.2,1.7172019958496094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,4,1,power_law_1.2,0.07321919918060302
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,4,1,power_law_1.2,2.3535308837890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,4,1,power_law_1.2,0.08561920166015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,4,1,power_law_1.2,2.951441345214844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,4,1,power_law_1.01,2.6597869873046873
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,4,1,power_law_1.2,0.11028991699218751
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,4,1,power_law_1.2,0.1718560028076172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,4,1,power_law_1.2,0.23782720565795898
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,4,1,power_law_1.2,0.25946367263793946
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,4,1,power_law_1.2,0.2690092849731445
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,4,1,power_law_1.2,0.28126848220825196
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,4,1,power_law_1.2,0.25411903381347656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,4,1,power_law_1.2,0.30211904525756833
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,4,1,power_law_1.2,0.3162431907653809
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,4,1,power_law_1.2,0.29174335479736324
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,4,1,power_law_1.2,0.3350124740600586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,4,1,power_law_1.2,0.349249267578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,4,1,power_law_1.01,3.495729370117188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,4,1,power_law_1.2,0.5019295883178712
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,4,1,power_law_1.2,0.13862784385681154
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,4,1,power_law_1.2,0.37987457275390624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,4,1,power_law_1.2,0.44408702850341797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,4,1,power_law_1.2,0.6420646667480469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,4,1,power_law_1.2,0.7830143737792968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,4,balanced,0.023150079250335694
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,4,1,power_law_1.2,1.095768356323242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,4,balanced,0.023586559295654296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,4,balanced,0.045954561233520506
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,4,balanced,0.07264512062072755
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,4,1,power_law_1.2,1.4033132934570314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,4,balanced,0.033518080711364744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,4,balanced,0.07542016029357909
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,4,balanced,0.09331839561462403
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,4,balanced,0.10059904098510741
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,4,balanced,0.0894041633605957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,4,balanced,0.09333120346069336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,4,balanced,0.09375616073608398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,4,balanced,0.09322879791259767
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,4,balanced,0.09957375526428222
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,4,balanced,0.10396800041198731
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,4,balanced,0.10716927528381348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,4,balanced,0.08895744323730469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,4,balanced,0.12155776023864746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,4,balanced,0.13856255531311035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,4,balanced,0.08884608268737793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,4,balanced,0.19994239807128905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,4,balanced,0.2315635108947754
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,4,1,power_law_1.2,2.0987756347656252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,4,balanced,0.314778881072998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,4,balanced,0.3972825622558594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,4,balanced,0.5685120010375977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,4,balanced,0.7411417388916016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,4,balanced,1.0833229064941405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,4,balanced,0.02634880065917969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,4,balanced,0.026416640281677246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,4,balanced,0.027289600372314454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,4,balanced,0.03386368036270142
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,4,balanced,0.06822656154632568
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,4,balanced,0.059362559318542486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,4,balanced,0.06203392028808594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,4,balanced,1.427060546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,4,balanced,0.07025792121887206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,4,balanced,0.07256959915161133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,4,balanced,0.07602303981781006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,4,balanced,0.06573440074920654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,4,balanced,0.06769279956817627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,4,balanced,0.06735616207122803
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,4,balanced,0.06854527950286865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,4,balanced,0.0693452787399292
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,4,balanced,0.07218431949615478
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,4,balanced,0.0774015998840332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,4,balanced,0.08805760383605957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,4,balanced,0.09490303993225098
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,4,balanced,0.1192191982269287
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,4,balanced,0.13950719833374023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,4,balanced,0.1993075180053711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,4,balanced,0.23937023162841795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,4,balanced,0.337757453918457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,4,balanced,0.4410982513427735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,4,balanced,0.6319705581665038
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,4,balanced,0.8289791870117188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,4,1,power_law_1.2,2.8637127685546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,1,4,balanced,0.026228480339050293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,1,4,balanced,0.02052608013153076
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,1,4,balanced,0.021553919315338135
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,1,4,balanced,0.022702078819274902
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,1,4,balanced,0.03666304111480713
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,1,4,balanced,0.036327679157257084
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,1,4,balanced,0.03701632022857666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,1,4,balanced,0.03665791988372803
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,1,4,balanced,0.037550079822540286
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,1,4,balanced,0.037067520618438723
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,1,4,balanced,0.037962241172790526
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,1,4,balanced,0.03786112070083618
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,1,4,balanced,0.03827584028244019
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,1,4,balanced,0.03848063945770264
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,1,4,balanced,0.04026368141174316
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,1,4,balanced,0.04100607872009278
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,1,4,balanced,0.04347904205322266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,1,4,balanced,0.04705664157867432
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,1,4,balanced,0.03588736057281494
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,1,4,balanced,0.03846271991729736
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,1,4,balanced,0.04387199878692627
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,1,4,balanced,0.06626304149627685
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,1,4,balanced,0.06258560180664062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,1,4,balanced,0.08696191787719727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,4,power_law_1.01,0.027261440753936765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,4,power_law_1.01,0.04082047939300537
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,1,4,balanced,0.10566911697387696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,4,power_law_1.01,0.049122557640075684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,4,power_law_1.01,0.05772031784057617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,4,power_law_1.01,0.06966400146484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,4,power_law_1.01,0.08869376182556152
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,1,4,balanced,0.1465574359893799
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,4,power_law_1.01,0.09087360382080079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,4,power_law_1.01,0.0935923194885254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,4,power_law_1.01,0.08763775825500489
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,4,power_law_1.01,0.09830016136169434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,4,power_law_1.01,0.08808447837829589
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,4,power_law_1.01,0.09452799797058106
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,4,power_law_1.01,0.09956607818603516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,4,power_law_1.01,0.10758015632629395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,4,power_law_1.01,0.043554558753967285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,4,power_law_1.01,0.11627264022827148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,4,power_law_1.01,0.14967040061950682
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,4,power_law_1.01,0.17072895050048828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,4,power_law_1.01,0.22486911773681642
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,4,power_law_1.01,0.2647923278808594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,4,power_law_1.01,0.3593305587768555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,4,power_law_1.01,0.4646067047119141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,4,power_law_1.01,0.6344371032714844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,4,power_law_1.01,0.026771199703216553
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,4,power_law_1.01,0.03175807952880859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,4,power_law_1.01,1.1962303924560547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,4,power_law_1.01,0.03371520042419433
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,4,power_law_1.01,0.03928960084915161
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,4,power_law_1.01,0.04337535858154297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,4,power_law_1.01,1.580563201904297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,4,power_law_1.01,0.814725112915039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,4,power_law_1.01,0.058787841796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,4,power_law_1.01,0.06892032146453857
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,4,power_law_1.01,0.0716096019744873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,4,power_law_1.01,0.07247360229492188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,4,power_law_1.01,0.07482367992401123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,4,power_law_1.01,0.0664512014389038
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,4,power_law_1.01,0.06685440063476562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,4,power_law_1.01,0.06886015892028809
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,4,power_law_1.01,0.07020415782928466
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,4,power_law_1.01,0.09203071594238281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,4,power_law_1.01,0.0717632007598877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,4,power_law_1.01,0.07768959999084472
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,4,power_law_1.01,0.09117823600769043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,4,power_law_1.01,0.10064255714416503
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,4,power_law_1.01,0.12839167594909667
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,4,power_law_1.01,0.15588735580444335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,4,power_law_1.01,0.2138047981262207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,4,power_law_1.01,0.07338240146636962
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,4,power_law_1.01,0.3652608108520508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,4,power_law_1.01,0.47704830169677737
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,1,4,power_law_1.01,0.021839361190795898
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,1,4,power_law_1.01,0.02424448013305664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,1,4,power_law_1.01,0.02642816066741943
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,4,power_law_1.01,0.2580723190307617
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,1,4,power_law_1.01,0.027927041053771973
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,4,power_law_1.01,0.8585881805419922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,1,4,power_law_1.01,0.028908801078796387
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,1,4,power_law_1.01,0.03373440027236939
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,1,4,power_law_1.01,0.03421695947647095
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,1,4,power_law_1.01,0.03534463882446289
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,1,4,power_law_1.01,0.03610879898071289
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,1,4,power_law_1.01,0.036229119300842286
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,1,4,power_law_1.01,0.03691391944885254
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,1,4,power_law_1.01,0.037761280536651606
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,1,4,power_law_1.01,0.038731520175933835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,4,power_law_1.01,0.6921459197998047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,1,4,power_law_1.01,0.039212799072265624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,1,4,power_law_1.01,0.04086143970489502
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,1,4,power_law_1.01,0.035270400047302246
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,1,4,balanced,0.18903680801391604
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,1,4,power_law_1.01,0.054096641540527346
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,1,4,power_law_1.01,0.06298880100250244
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,1,4,power_law_1.01,0.04322432041168213
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,1,4,power_law_1.01,0.04780159950256348
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,1,4,power_law_1.01,0.07352960109710693
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,1,4,power_law_1.01,0.09161855697631835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,4,power_law_1.2,0.02747391939163208
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,4,power_law_1.2,0.03579904079437256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,4,power_law_1.2,0.04456575870513916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,4,power_law_1.2,0.047712001800537106
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,4,power_law_1.2,0.05573247909545899
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,4,power_law_1.2,0.0718284797668457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,4,power_law_1.2,0.0858675193786621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,4,power_law_1.2,0.09135359764099121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,4,power_law_1.2,0.09395071983337402
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,4,power_law_1.2,0.09482111930847167
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,4,power_law_1.2,0.09469183921813965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,4,power_law_1.2,0.0886195182800293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,4,power_law_1.2,0.09056896209716797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,4,power_law_1.2,0.09478272438049316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,4,power_law_1.2,0.098787841796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,4,power_law_1.2,0.10910719871520995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,4,power_law_1.2,0.11878911972045898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,4,power_law_1.2,0.14974720001220704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,4,power_law_1.2,0.17131263732910157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,4,power_law_1.2,0.23036159515380858
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,4,power_law_1.2,0.274836483001709
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,1,4,power_law_1.01,0.12200448036193848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,4,power_law_1.2,0.38356609344482423
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,4,power_law_1.2,0.4808115386962891
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,1,4,power_law_1.01,0.1653606414794922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,1,4,power_law_1.01,0.2998899269104004
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,1,4,power_law_1.01,0.21564159393310547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,4,power_law_1.2,0.026828799247741696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,4,power_law_1.2,0.028115200996398925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,4,power_law_1.2,0.03336704015731812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,1,4,power_law_1.01,0.40629631042480463
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,4,power_law_1.2,0.03684864044189453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,4,power_law_1.2,0.04179327964782715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,4,power_law_1.2,0.06034175872802734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,4,power_law_1.2,0.06720511913299561
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,4,power_law_1.2,0.0690009593963623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,4,power_law_1.2,0.07074944019317628
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,4,power_law_1.2,0.073056001663208
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,4,power_law_1.2,0.073537278175354
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,4,power_law_1.2,0.06607872009277344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,4,power_law_1.2,0.06814464092254638
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,4,power_law_1.2,0.6830707550048828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,4,power_law_1.2,0.06956416130065918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,4,power_law_1.2,0.07074175834655762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,4,power_law_1.2,0.07137152194976806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,4,power_law_1.2,0.0915071964263916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,4,power_law_1.2,0.10335871696472168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,4,power_law_1.2,0.12911359786987303
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,4,power_law_1.2,0.15790080070495605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,4,power_law_1.2,0.2091929626464844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,4,power_law_1.2,0.07891200065612794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,4,power_law_1.2,0.2651417541503906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,4,power_law_1.2,0.5009791946411133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,4,power_law_1.2,0.36806526184082033
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,1,4,power_law_1.2,0.02143872022628784
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,1,4,power_law_1.2,0.02242432117462158
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,1,4,power_law_1.2,0.025676798820495606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,4,power_law_1.2,0.8687474822998047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,1,4,power_law_1.2,0.028039679527282712
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,1,4,power_law_1.2,0.03042304039001465
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,1,4,power_law_1.2,0.03298432111740112
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,1,4,power_law_1.2,0.03434623956680298
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,1,4,power_law_1.2,0.03587455987930298
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,1,4,power_law_1.2,0.035959041118621825
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,1,4,power_law_1.2,0.03684096097946167
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,1,4,power_law_1.2,0.03676160097122193
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,1,4,power_law_1.2,0.03749759912490845
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,1,4,power_law_1.2,0.037880320549011234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,1,4,power_law_1.2,0.03926912069320679
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,1,4,power_law_1.2,0.04091519832611084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,4,power_law_1.2,0.889977569580078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,4,power_law_1.2,0.7005503845214844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,4,power_law_1.2,1.325201873779297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,1,4,power_law_1.2,0.0417574405670166
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,1,4,power_law_1.2,0.044893441200256345
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,1,4,power_law_1.2,0.05128704071044922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,1,4,power_law_1.2,0.057894401550292976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,4,power_law_1.2,1.6821043395996091
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,1,4,power_law_1.2,0.09520511627197266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,2,balanced,0.019805439710617066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,2,balanced,0.02188416004180908
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,2,balanced,0.03174015998840332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,2,balanced,0.0460646390914917
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,2,balanced,0.06897280216217042
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,2,balanced,0.06919936180114747
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,2,balanced,0.07123712062835694
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,2,balanced,0.07160704135894776
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,2,balanced,0.08223615646362305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,2,balanced,0.0829862403869629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,2,balanced,0.08338944435119629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,2,balanced,0.08643839836120606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,2,balanced,0.08703488349914551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,2,balanced,0.08851967811584473
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,2,balanced,0.0926579189300537
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,2,balanced,0.09376640319824218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,2,balanced,0.09989248275756836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,2,balanced,0.10980735778808595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,2,balanced,0.1232038402557373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,2,balanced,0.17784063339233397
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,2,balanced,0.2048588752746582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,2,balanced,0.2766771125793457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,2,balanced,0.35017345428466795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,2,balanced,0.506371841430664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,1,4,power_law_1.2,0.07440383911132813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,2,balanced,0.6582553863525391
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,1,4,power_law_1.2,0.2349465560913086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,2,balanced,0.02521984100341797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,2,balanced,0.025134079456329346
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,1,4,power_law_1.2,0.17240192413330077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,2,balanced,0.02807552099227905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,2,balanced,0.03510528087615967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,2,balanced,1.2700672149658203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,2,balanced,0.06372608184814453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,2,balanced,0.06386559963226318
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,2,balanced,0.0647641611099243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,2,balanced,0.06550399780273437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,2,balanced,0.07271168231964112
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,2,balanced,0.07279488086700439
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,2,balanced,0.07409023761749267
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,2,balanced,0.07471360206604004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,2,balanced,0.0760972785949707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,2,balanced,0.07768703937530518
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,2,balanced,0.07878528118133546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,2,balanced,0.07926527976989746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,2,balanced,0.08351103782653808
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,2,balanced,0.08828288078308105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,2,balanced,0.09652352333068848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,2,balanced,0.11332863807678221
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,2,balanced,0.13011712074279785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,2,balanced,0.18905599594116212
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,2,balanced,0.22073087692260743
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,2,balanced,0.31280639648437497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,2,balanced,0.4135782241821289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,2,balanced,0.593936653137207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,2,balanced,0.7841343688964844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,2,balanced,0.963009262084961
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,1,4,power_law_1.2,0.3015551948547363
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,1,4,power_law_1.2,0.12374272346496581
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,1,4,power_law_1.2,0.5000806427001953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,2,2,balanced,0.02766592025756836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,2,2,balanced,0.026785280704498295
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,2,2,balanced,0.04576511859893799
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,2,2,balanced,0.046403841972351076
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,2,2,balanced,0.04565375804901123
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,2,2,balanced,0.046302719116210936
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,2,2,balanced,0.046489601135253904
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,2,2,balanced,0.04727168083190918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,2,2,balanced,0.0474944019317627
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,2,2,balanced,0.04825088024139405
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,2,2,balanced,0.048951039314270015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,2,2,balanced,0.05014527797698974
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,2,2,balanced,0.05220863819122314
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,2,2,balanced,0.05393152236938477
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,2,2,balanced,0.057866239547729494
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,2,2,balanced,0.03275648117065429
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,2,2,balanced,0.06325759887695312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,2,2,balanced,0.028492801189422608
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,2,2,balanced,0.04534656047821045
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,2,2,balanced,0.04417151927947998
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,2,2,balanced,0.04538879871368408
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,2,power_law_1.01,0.021556479930877687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,1,4,power_law_1.2,0.6417561340332031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,2,power_law_1.01,0.03146752119064331
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,2,power_law_1.01,0.03692032098770141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,2,power_law_1.01,0.042536959648132325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,2,power_law_1.01,0.04968319892883301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,2,power_law_1.01,0.06317440032958985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,2,power_law_1.01,0.06442495822906494
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,2,power_law_1.01,0.06680191993713379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,2,power_law_1.01,0.07843200206756593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,2,power_law_1.01,0.08066047668457031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,2,power_law_1.01,0.08083071708679199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,2,power_law_1.01,0.08340864181518555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,2,power_law_1.01,0.08359807968139649
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,2,power_law_1.01,0.08940032005310058
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,2,power_law_1.01,0.09233152389526367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,2,power_law_1.01,0.09653632164001465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,2,power_law_1.01,0.10554112434387206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,2,power_law_1.01,0.13708288192749024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,2,power_law_1.01,0.1521126365661621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,2,power_law_1.01,0.1857062339782715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,2,power_law_1.01,0.22829952239990234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,2,power_law_1.01,0.29992448806762695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,2,power_law_1.01,0.3930815887451172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,2,power_law_1.01,0.5483443069458007
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,2,power_law_1.01,0.7035968017578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,2,power_law_1.01,1.0472217559814454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,2,power_law_1.01,0.024802560806274417
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,2,power_law_1.01,1.3273202514648437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,2,power_law_1.01,0.027229440212249757
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,2,power_law_1.01,0.03280384063720703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,2,power_law_1.01,0.040652799606323245
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,2,power_law_1.01,0.05682176113128662
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,2,power_law_1.01,0.06215551853179931
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,2,power_law_1.01,0.06204160213470459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,2,power_law_1.01,0.03625600099563599
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,2,power_law_1.01,0.06978047847747802
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,2,power_law_1.01,0.07092095851898193
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,2,power_law_1.01,0.07404416084289552
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,2,power_law_1.01,0.07538943767547608
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,2,power_law_1.01,0.07798912048339843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,2,power_law_1.01,0.07934080123901367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,2,power_law_1.01,0.07206783771514894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,2,power_law_1.01,0.08033791542053223
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,2,2,balanced,0.08611712455749512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,2,power_law_1.01,0.08432512283325196
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,2,2,balanced,0.0950592041015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,2,power_law_1.01,0.09281791687011719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,2,power_law_1.01,0.12650752067565918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,2,power_law_1.01,0.10374655723571777
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,2,power_law_1.01,0.15297280311584474
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,2,power_law_1.01,0.2018880081176758
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,2,2,balanced,0.12465023994445801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,2,power_law_1.01,0.24805631637573242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,2,2,balanced,0.22228864669799803
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,2,power_law_1.01,0.34460929870605467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,2,power_law_1.01,0.4402790451049805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,2,2,power_law_1.01,0.028474879264831544
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,2,2,balanced,0.16023807525634765
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,2,2,power_law_1.01,0.03049344062805176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,2,power_law_1.01,0.6371007919311523
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,2,2,power_law_1.01,0.032983040809631346
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,2,2,power_law_1.01,0.03557375907897949
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,2,2,power_law_1.01,0.04273024082183838
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,2,2,power_law_1.01,0.04461056232452393
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,2,2,power_law_1.01,0.04434815883636475
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,2,2,power_law_1.01,0.04482560157775879
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,2,2,power_law_1.01,0.045226240158081056
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,2,2,power_law_1.01,0.04670720100402832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,2,2,power_law_1.01,0.046813440322875974
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,2,2,power_law_1.01,0.04731391906738282
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,2,2,power_law_1.01,0.04901375770568848
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,2,2,power_law_1.01,0.05593215942382812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,2,power_law_1.01,0.8569292449951172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,2,2,balanced,0.28993791580200196
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,2,2,power_law_1.01,0.02760832071304321
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,2,2,power_law_1.01,0.050913281440734856
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,2,2,power_law_1.01,0.08669183731079103
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,2,2,power_law_1.01,0.0721779203414917
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,2,2,power_law_1.01,0.05244671821594238
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,2,2,power_law_1.01,0.06422783851623534
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,2,power_law_1.2,0.021477119922637938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,2,power_law_1.2,0.031307520866394045
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,2,2,power_law_1.01,0.10283519744873046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,2,power_law_1.2,0.03610624074935913
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,2,power_law_1.2,0.040524802207946777
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,2,power_law_1.2,0.04749567985534668
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,2,2,power_law_1.01,0.14033023834228514
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,2,power_law_1.2,0.06292863845825195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,2,power_law_1.2,0.06302207946777344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,2,power_law_1.2,0.08024064064025879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,2,power_law_1.2,0.07992576122283936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,2,power_law_1.2,0.0809881591796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,2,power_law_1.2,0.08331135749816895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,2,power_law_1.2,0.0861849594116211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,2,power_law_1.2,0.09040639877319337
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,2,power_law_1.2,0.07009024143218993
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,2,power_law_1.2,0.09384575843811035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,2,power_law_1.2,0.09808896064758302
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,2,power_law_1.2,0.10930303573608398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,2,power_law_1.2,0.13806976318359374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,2,power_law_1.2,0.15472384452819826
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,2,power_law_1.2,0.18962047576904298
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,2,power_law_1.2,0.2364441680908203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,2,power_law_1.2,0.31843456268310544
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,2,power_law_1.2,0.4048883056640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,2,power_law_1.2,0.5592563247680664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,2,2,power_law_1.01,0.1788915252685547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,2,power_law_1.2,0.718419189453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,2,power_law_1.2,0.025521919727325437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,2,power_law_1.2,0.02929663896560669
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,2,power_law_1.2,0.03223680019378662
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,2,power_law_1.2,0.03671168088912964
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,2,2,power_law_1.01,0.24851072311401365
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,2,power_law_1.2,0.03970943927764893
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,2,power_law_1.2,0.05658751964569092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,2,power_law_1.2,0.06246912002563476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,2,power_law_1.2,0.07149439811706543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,2,power_law_1.2,0.07302783966064454
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,2,2,power_law_1.01,0.3381836700439453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,2,2,power_law_1.01,0.4386982345581055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,2,power_law_1.2,0.07282559871673584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,2,power_law_1.2,0.07522560119628906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,2,power_law_1.2,0.07605247974395753
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,2,power_law_1.2,0.07885824203491211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,2,power_law_1.2,0.07928319931030273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,2,power_law_1.2,0.08106752395629882
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,2,power_law_1.2,0.08502911567687989
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,2,power_law_1.2,0.09460479736328124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,2,power_law_1.2,0.10433152198791504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,2,power_law_1.2,0.12928128242492676
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,2,2,power_law_1.01,0.6390924835205078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,2,power_law_1.2,0.15391743659973145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,2,power_law_1.2,0.2025676727294922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,2,power_law_1.2,0.25349376678466795
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,2,2,power_law_1.2,0.027612159252166747
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,2,2,power_law_1.2,0.028574719429016116
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,2,2,power_law_1.2,0.0309555196762085
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,2,2,power_law_1.2,0.033070080280303955
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,2,2,power_law_1.2,0.03472768068313599
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,2,2,power_law_1.2,0.04180863857269287
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,2,2,power_law_1.2,0.04386303901672363
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,2,2,power_law_1.2,0.04469120025634766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,2,power_law_1.2,0.05968512058258056
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,2,2,power_law_1.2,0.045263361930847165
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,2,2,power_law_1.2,0.046069760322570805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,2,2,power_law_1.2,0.04680319786071777
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,2,2,power_law_1.2,0.047365121841430664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,2,2,power_law_1.2,0.04852352142333984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,2,2,power_law_1.2,0.04999040126800537
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,2,power_law_1.2,0.44426113128662104
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,2,2,power_law_1.2,0.05474944114685058
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,2,power_law_1.2,1.0979110717773437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,2,power_law_1.2,0.34689537048339847
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,2,power_law_1.2,0.8314444732666015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,2,power_law_1.2,0.6344627380371094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,2,power_law_1.2,1.3336370849609376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,2,2,power_law_1.2,0.0707583999633789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,2,2,power_law_1.2,0.05262720108032226
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,2,2,power_law_1.2,0.08409472465515136
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,2,2,power_law_1.2,0.05952383995056152
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,2,2,power_law_1.2,0.09700480461120606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,4,1,balanced,0.0291596794128418
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,2,2,power_law_1.2,0.1344883155822754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,4,1,balanced,0.029983360767364502
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,1,balanced,0.05695040225982666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,1,balanced,0.04099584102630615
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,1,balanced,0.08164735794067382
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,1,balanced,0.08475584030151367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,1,balanced,0.0846656036376953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,1,balanced,0.08589823722839356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,1,balanced,0.08423999786376954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,1,balanced,0.0850175952911377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,1,balanced,0.08552000045776367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,1,balanced,0.10363583564758301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,1,balanced,0.10382847785949707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,1,balanced,0.10698559761047363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,1,balanced,0.10884672164916993
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,1,balanced,0.1118943977355957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,1,balanced,0.11529664039611816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,1,balanced,0.12678079605102538
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,1,balanced,0.13575551986694337
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,1,balanced,0.18847551345825195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,1,balanced,0.21001920700073243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,1,balanced,0.27487615585327146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,1,balanced,0.3400089645385742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,1,balanced,0.4813222503662109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,1,balanced,0.6193465423583985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,1,balanced,0.904299545288086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,2,2,power_law_1.2,0.18529535293579102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,1,balanced,1.1829766082763673
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,2,2,power_law_1.2,0.24164735794067385
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,1,balanced,0.032876799106597895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,1,balanced,0.03980992078781128
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,1,balanced,0.05084671974182129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,1,balanced,0.08732992172241211
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,2,2,power_law_1.2,0.34330623626708984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,1,balanced,0.08784704208374024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,1,balanced,0.08958592414855956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,1,balanced,0.0903065586090088
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,1,balanced,0.08846912384033204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,1,balanced,0.08723008155822753
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,2,2,power_law_1.2,0.5170969772338867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,1,balanced,0.03367295980453491
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,1,balanced,0.08825471878051758
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,2,2,power_law_1.2,0.702182388305664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,1,balanced,0.10173184394836425
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,1,balanced,0.12020095825195312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,1,balanced,0.10182016372680665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,1,balanced,0.13408512115478516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,1,balanced,0.10198911666870117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,1,balanced,0.14782655715942383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,1,balanced,0.21984895706176757
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,1,balanced,0.10277888298034668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,1,balanced,0.24733631134033204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,1,balanced,0.34668224334716796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,1,balanced,0.44244033813476563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,1,balanced,0.6451007843017578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,1,balanced,0.8440128326416015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,1,balanced,0.10529472351074218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,1,balanced,0.10880895614624024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,1,balanced,0.11403455734252929
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,2,2,power_law_1.2,0.9214835357666015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,4,1,balanced,0.04005119800567627
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,4,1,balanced,0.041069440841674805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,4,1,balanced,0.06559616088867187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,4,1,balanced,0.06523392200469971
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,4,1,balanced,0.04265664100646972
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,4,1,balanced,0.049628162384033205
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,4,1,balanced,0.06475327968597412
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,4,1,balanced,0.06494336128234864
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,4,1,balanced,0.0660691213607788
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,4,1,balanced,0.06619328022003174
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,4,1,balanced,0.06727359771728517
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,4,1,balanced,0.06821184158325196
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,4,1,balanced,0.0683135986328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,4,1,balanced,0.0650707197189331
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,4,1,balanced,0.06982656002044678
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,4,1,balanced,0.07134655952453614
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,4,1,balanced,0.0744652795791626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,4,1,balanced,0.0784006404876709
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,4,1,balanced,0.08612992286682128
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,4,1,balanced,0.09296128273010254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,4,1,power_law_1.01,0.028959360122680668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,4,1,power_law_1.01,0.035301120281219484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,4,1,balanced,0.06561791896820068
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,1,power_law_1.01,0.050240640640258794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,1,power_law_1.01,0.057573761940002445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,1,power_law_1.01,0.07562687873840332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,1,power_law_1.01,0.07685823917388916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,1,power_law_1.01,0.0786233615875244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,1,power_law_1.01,0.08232831954956055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,1,power_law_1.01,0.08468671798706054
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,1,power_law_1.01,0.0897152042388916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,1,power_law_1.01,0.1004601573944092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,1,power_law_1.01,0.10345791816711425
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,1,power_law_1.01,0.10726719856262208
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,1,power_law_1.01,0.10997759819030761
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,1,power_law_1.01,0.11296192169189453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,1,power_law_1.01,0.1206220817565918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,1,power_law_1.01,0.04388607978820801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,1,power_law_1.01,0.14081472396850586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,1,power_law_1.01,0.1630668830871582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,1,power_law_1.01,0.19366399765014647
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,1,power_law_1.01,0.22715776443481445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,1,power_law_1.01,0.2965913581848144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,1,power_law_1.01,0.3675686264038086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,1,power_law_1.01,0.5090086364746094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,1,power_law_1.01,0.6516799926757812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,1,power_law_1.01,0.9380966186523437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,1,power_law_1.01,1.2241248321533202
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,4,1,balanced,0.14700927734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,1,power_law_1.01,0.033693439960479736
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,4,1,balanced,0.13049344062805174
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,4,1,balanced,0.1992403221130371
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,4,1,balanced,0.25374847412109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,1,power_law_1.01,0.04883584022521973
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,1,power_law_1.01,0.05921919822692871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,1,power_law_1.01,0.08297408103942872
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,1,power_law_1.01,0.0797817611694336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,1,power_law_1.01,0.08699456214904785
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,4,1,balanced,0.35994945526123046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,1,power_law_1.01,0.08698623657226563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,1,power_law_1.01,0.08762751579284668
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,4,1,balanced,0.47130561828613277
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,1,power_law_1.01,0.10008831977844239
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,1,power_law_1.01,0.1006009578704834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,1,power_law_1.01,0.10300095558166504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,1,power_law_1.01,0.10543871879577638
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,1,power_law_1.01,0.10766976356506348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,1,power_law_1.01,0.11462016105651855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,1,power_law_1.01,0.12151679992675782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,1,power_law_1.01,0.13775551795959473
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,1,power_law_1.01,0.1577625560760498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,1,power_law_1.01,0.18604352951049805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,1,power_law_1.01,0.2317574310302734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,1,power_law_1.01,0.2806796836853027
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,1,power_law_1.01,0.38338432312011717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,1,power_law_1.01,0.030412158966064452
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,1,power_law_1.01,0.4863059234619141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,1,power_law_1.01,0.6892473602294922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,4,1,power_law_1.01,0.03984512090682983
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,4,1,power_law_1.01,0.04188096046447754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,1,power_law_1.01,0.8908985900878907
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,4,1,power_law_1.01,0.0449894380569458
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,4,1,power_law_1.01,0.04805503845214844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,4,1,power_law_1.01,0.05124032020568847
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,4,1,power_law_1.01,0.05963583946228027
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,4,1,power_law_1.01,0.062231678962707516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,4,1,power_law_1.01,0.0637279987335205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,1,power_law_1.01,0.08438015937805175
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,4,1,power_law_1.01,0.06465792179107666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,4,1,power_law_1.01,0.06480447769165039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,4,1,power_law_1.01,0.0664031982421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,4,1,power_law_1.01,0.06932223796844482
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,4,1,power_law_1.01,0.07234623908996582
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,4,1,power_law_1.01,0.06853439807891845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,1,power_law_1.01,0.0411411190032959
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,4,1,power_law_1.01,0.07780608177185058
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,4,1,power_law_1.01,0.09715392112731933
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,4,1,power_law_1.01,0.07477375984191895
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,4,1,power_law_1.01,0.08444031715393066
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,4,1,power_law_1.01,0.11082624435424804
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,4,1,power_law_1.01,0.14086336135864258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,4,1,power_law_1.01,0.17169279098510742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,4,1,power_law_1.2,0.02915839910507202
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,4,1,power_law_1.2,0.03700223922729492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,1,power_law_1.2,0.04310848236083985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,1,power_law_1.2,0.04940415859222412
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,1,power_law_1.2,0.07673600196838379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,1,power_law_1.2,0.07967296123504639
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,1,power_law_1.2,0.08196415901184081
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,1,power_law_1.2,0.08517760276794434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,1,power_law_1.2,0.05757503986358643
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,1,power_law_1.2,0.08648832321166992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,1,power_law_1.2,0.09229503631591797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,1,power_law_1.2,0.102739839553833
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,4,1,power_law_1.01,0.23483007431030273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,1,power_law_1.2,0.10408127784729004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,1,power_law_1.2,0.10906240463256836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,1,power_law_1.2,0.11098239898681642
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,1,power_law_1.2,0.1137235164642334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,1,power_law_1.2,0.12125184059143065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,1,power_law_1.2,0.14866175651550292
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,1,power_law_1.2,0.1641644859313965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,1,power_law_1.2,0.1950182342529297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,1,power_law_1.2,0.2275020790100098
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,1,power_law_1.2,0.29682815551757813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,1,power_law_1.2,0.3678662490844727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,1,power_law_1.2,0.5104844665527344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,4,1,power_law_1.01,0.29830720901489255
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,1,power_law_1.2,0.03288640022277832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,1,power_law_1.2,0.03410880088806152
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,4,1,power_law_1.01,0.4341113662719726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,1,power_law_1.2,0.653288345336914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,1,power_law_1.2,0.040700159072875976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,1,power_law_1.2,0.04797120094299316
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,4,1,power_law_1.01,0.5714764785766602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,1,power_law_1.2,0.05896895885467529
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,1,power_law_1.2,0.0800063991546631
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,1,power_law_1.2,0.08440256118774414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,1,power_law_1.2,0.0860377597808838
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,1,power_law_1.2,0.0867795181274414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,4,1,power_law_1.01,0.8576179504394531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,1,power_law_1.2,0.08746944427490234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,1,power_law_1.2,0.10083392143249512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,1,power_law_1.2,0.1007852840423584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,4,1,power_law_1.01,1.1457734680175782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,1,power_law_1.2,0.10784255981445312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,1,power_law_1.2,0.1101203155517578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,1,power_law_1.2,0.11458815574645995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,1,power_law_1.2,0.13980799674987793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,1,power_law_1.2,0.12891967773437502
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,1,power_law_1.2,0.1634592056274414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,1,power_law_1.2,0.19037439346313476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,1,power_law_1.2,0.2334502410888672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,1,power_law_1.2,0.28510719299316406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,1,power_law_1.2,0.08429311752319336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,4,1,power_law_1.2,0.040167040824890136
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,1,power_law_1.2,0.10468480110168457
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,4,1,power_law_1.2,0.04188799858093262
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,4,1,power_law_1.2,0.044614400863647465
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,4,1,power_law_1.2,0.04680768013000488
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,4,1,power_law_1.2,0.050645117759704594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,4,1,power_law_1.2,0.060375041961669915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,1,power_law_1.2,0.3855526351928711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,4,1,power_law_1.2,0.062300801277160645
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,4,1,power_law_1.2,0.06320320129394531
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,4,1,power_law_1.2,0.06515647888183593
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,4,1,power_law_1.2,0.06570752143859863
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,4,1,power_law_1.2,0.06678656101226807
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,4,1,power_law_1.2,0.06976384162902832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,4,1,power_law_1.2,0.07110784053802491
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,4,1,power_law_1.2,0.07897791862487794
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,4,1,power_law_1.2,0.0825376033782959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,1,power_law_1.2,0.4874310302734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,1,power_law_1.2,0.9408678436279297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,4,1,power_law_1.2,0.07463039875030517
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,4,1,power_law_1.2,0.10981311798095703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,4,1,power_law_1.2,0.09102911949157715
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,4,1,power_law_1.2,0.13074624061584472
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,1,power_law_1.2,0.6926726531982422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,4,1,power_law_1.2,0.1762540817260742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,4,1,power_law_1.2,0.22242752075195313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,4,1,power_law_1.2,0.3180287933349609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,4,balanced,0.04804096221923828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,4,balanced,0.05920000076293945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,4,balanced,0.07425792217254638
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,4,balanced,0.1312166404724121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,4,balanced,0.22944000244140622
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,4,balanced,0.23387008666992187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,4,balanced,0.2705484771728516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,4,balanced,0.2729830360412598
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,4,1,power_law_1.2,0.41682689666748046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,4,balanced,0.27208703994750977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,4,balanced,0.2773004722595215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,4,balanced,0.27487232208251955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,1,power_law_1.2,1.2270867156982423
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,4,balanced,0.27818239212036133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,4,balanced,0.28723840713500975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,4,balanced,0.283636474609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,4,balanced,0.30876672744750977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,4,balanced,0.3125862312316895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,4,balanced,0.34289920806884766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,4,balanced,0.3779993438720703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,4,balanced,0.5114931106567383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,4,balanced,0.29571712493896485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,4,balanced,0.5785753631591797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,4,balanced,0.8123506927490235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,4,balanced,1.0125145721435547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,4,balanced,1.464837188720703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,4,balanced,1.9195353698730468
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,4,1,power_law_1.2,0.6086457443237305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,4,balanced,0.032104959487915044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,4,balanced,0.03749504089355469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,4,balanced,0.05346303939819336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,4,balanced,0.08303487777709961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,4,balanced,0.13371007919311523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,4,balanced,0.1363929557800293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,4,balanced,0.1522163200378418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,4,balanced,0.15509632110595703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,4,balanced,0.158472957611084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,4,balanced,0.16047359466552735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,4,balanced,0.14943360328674316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,4,balanced,0.15146623611450197
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,4,balanced,0.15215999603271485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,4,balanced,0.15523839950561524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,4,balanced,0.1563596820831299
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,4,balanced,0.1587763214111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,4,balanced,0.16923648834228516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,4,balanced,0.1989695930480957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,4,balanced,0.20577535629272461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,4,balanced,3.012720642089844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,4,balanced,0.25024000167846683
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,4,balanced,0.28848896026611326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,4,balanced,0.4012249755859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,4,balanced,4.257583312988281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,4,balanced,0.4710105514526367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,4,balanced,0.6612940979003906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,4,balanced,0.8559117126464845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,1,power_law_1.2,0.8936121368408203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,4,balanced,1.24023681640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,4,balanced,1.6590757751464842
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,4,1,power_law_1.2,0.8009689331054688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,1,4,balanced,0.022771201133728027
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,1,4,balanced,0.08370304107666016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,1,4,balanced,0.08376447677612305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,1,4,balanced,0.024286720752716064
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,1,4,balanced,0.08424703598022461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,1,4,balanced,0.03461888074874878
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,1,4,balanced,0.08454655647277833
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,1,4,balanced,0.05121920108795166
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,1,4,balanced,0.08463616371154785
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,1,4,balanced,0.08381695747375488
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,1,4,balanced,0.0846463966369629
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,1,4,balanced,0.08492287635803222
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,1,4,balanced,0.08580351829528808
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,1,4,balanced,0.08657792091369629
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,1,4,balanced,0.08813695907592774
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,1,4,balanced,0.09175040245056151
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,1,4,balanced,0.09373567581176759
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,1,4,balanced,0.09902976036071778
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,4,1,power_law_1.2,1.1886495971679687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,1,4,balanced,0.10392959594726563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,1,4,balanced,0.0828428840637207
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,1,4,balanced,0.0831283187866211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,4,power_law_1.01,0.07351679801940918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,4,power_law_1.01,0.09832832336425781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,4,power_law_1.01,0.10833791732788085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,4,power_law_1.01,0.15664511680603027
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,4,power_law_1.01,0.1957695960998535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,4,power_law_1.01,0.22488704681396485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,4,power_law_1.01,0.26061439514160156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,4,power_law_1.01,0.26494464874267576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,4,power_law_1.01,0.2677375984191895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,4,power_law_1.01,0.27143039703369143
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,4,power_law_1.01,0.2750937652587891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,4,power_law_1.01,0.27215999603271485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,4,power_law_1.01,0.2722112083435059
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,4,power_law_1.01,0.2919731140136719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,4,power_law_1.01,0.30035583496093754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,4,power_law_1.01,0.3268889617919922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,4,power_law_1.01,0.3629759979248047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,4,power_law_1.01,0.41467647552490233
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,4,power_law_1.01,0.454015998840332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,4,power_law_1.01,0.5842649459838867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,4,power_law_1.01,0.7088230133056641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,4,power_law_1.01,0.9399807739257813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,4,power_law_1.01,1.2164672088623045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,4,power_law_1.01,1.9081715393066407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,4,power_law_1.01,2.4009202575683597
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,1,4,balanced,0.14278143882751465
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,1,4,balanced,0.2618137550354004
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,1,4,balanced,0.1532313632965088
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,1,4,balanced,0.20798336029052736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,4,power_law_1.01,0.051663360595703124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,4,power_law_1.01,0.058871040344238276
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,1,4,balanced,0.38148223876953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,4,power_law_1.01,0.07412992000579835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,4,power_law_1.01,0.08157312393188476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,4,power_law_1.01,0.09474431991577148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,4,power_law_1.01,0.13054847717285156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,4,power_law_1.01,0.14225919723510744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,4,power_law_1.01,0.14733440399169923
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,4,power_law_1.01,0.1503667163848877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,4,power_law_1.01,0.15410688400268555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,4,power_law_1.01,0.15338239669799805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,4,power_law_1.01,0.14802687644958495
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,4,power_law_1.01,0.1508249568939209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,4,power_law_1.01,0.15353216171264647
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,1,4,balanced,0.49764865875244146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,4,power_law_1.01,0.15492863655090333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,4,power_law_1.01,3.157715148925781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,4,power_law_1.01,0.1600614356994629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,4,power_law_1.01,0.17061376571655273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,4,power_law_1.01,0.20525695800781252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,4,power_law_1.01,0.21943296432495116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,4,power_law_1.01,0.28568960189819337
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,4,power_law_1.01,0.3309542465209961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,4,power_law_1.01,0.43624961853027344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,4,power_law_1.01,0.5133094406127929
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,4,power_law_1.01,0.7274278259277344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,1,4,power_law_1.01,0.044441599845886234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,1,4,power_law_1.01,0.04402431964874268
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,4,power_law_1.01,0.9561138916015626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,1,4,power_law_1.01,0.05905920028686523
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,1,4,power_law_1.01,0.0569817590713501
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,1,4,power_law_1.01,0.07631103992462158
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,1,4,power_law_1.01,0.07730559825897217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,4,power_law_1.01,1.4137715148925782
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,1,4,power_law_1.01,0.08137215614318848
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,1,4,power_law_1.01,0.07974656105041504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,4,power_law_1.01,1.7547354125976562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,1,4,power_law_1.01,0.08800383567810058
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,1,4,power_law_1.01,0.08669440269470215
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,1,4,power_law_1.01,0.07106688022613525
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,1,4,power_law_1.01,0.08805760383605957
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,1,4,power_law_1.01,0.0809932804107666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,1,4,power_law_1.01,0.10814080238342286
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,1,4,power_law_1.01,0.08504832267761231
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,4,power_law_1.01,4.45565185546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,4,1,power_law_1.2,1.5820364379882812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,1,4,power_law_1.01,0.0355404806137085
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,1,4,power_law_1.01,0.09524991989135742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,1,4,power_law_1.01,0.08981375694274903
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,1,4,power_law_1.01,0.12393728256225586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,1,4,power_law_1.01,0.15123583793640136
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,4,power_law_1.2,0.07368703842163085
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,1,4,power_law_1.01,0.18336383819580077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,4,power_law_1.2,0.08031359672546387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,4,power_law_1.2,0.10016127586364747
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,4,power_law_1.2,0.15568384170532226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,4,power_law_1.2,0.16854143142700195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,4,power_law_1.2,0.22296703338623045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,4,power_law_1.2,0.25569536209106447
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,4,power_law_1.2,0.26565631866455075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,4,power_law_1.2,0.2664422416687012
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,4,power_law_1.2,0.26869504928588867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,4,power_law_1.2,0.2802457618713379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,4,power_law_1.2,0.27339008331298825
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,4,power_law_1.2,0.2835251235961914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,4,power_law_1.2,0.2817766380310059
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,1,4,power_law_1.01,0.22524543762207033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,4,power_law_1.2,0.3034457588195801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,4,power_law_1.2,0.325401611328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,4,power_law_1.2,0.35715328216552733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,4,power_law_1.2,0.41964542388916015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,4,power_law_1.2,0.474444808959961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,4,power_law_1.2,0.6091775894165039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,4,power_law_1.2,0.7466713714599609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,4,power_law_1.2,1.116015396118164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,4,power_law_1.2,1.2695219421386719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,4,power_law_1.2,2.045366973876953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,1,4,power_law_1.01,0.2970035171508789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,4,power_law_1.2,0.04944511890411377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,4,power_law_1.2,2.578490905761719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,4,power_law_1.2,0.05523712158203124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,4,power_law_1.2,0.0879257583618164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,4,power_law_1.2,0.08914943695068359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,4,power_law_1.2,0.12829055786132812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,4,power_law_1.2,0.14215423583984374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,4,power_law_1.2,0.14769280433654786
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,4,power_law_1.2,0.1527680015563965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,4,power_law_1.2,0.1572159957885742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,4,power_law_1.2,0.15704319953918458
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,4,power_law_1.2,0.15042688369750976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,4,power_law_1.2,0.152359676361084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,4,power_law_1.2,0.07555456161499023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,4,power_law_1.2,0.15557248115539551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,4,power_law_1.2,0.15804160118103028
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,4,power_law_1.2,0.16190208435058592
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,4,power_law_1.2,0.2091993522644043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,4,power_law_1.2,0.22965631484985352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,4,power_law_1.2,0.29103103637695316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,4,power_law_1.2,0.3397568130493164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,4,power_law_1.2,0.45727489471435545
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,1,4,power_law_1.01,0.43229312896728517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,4,power_law_1.2,5.005332336425782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,4,power_law_1.2,0.7729318237304688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,4,power_law_1.2,0.6008102416992187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,1,4,power_law_1.01,0.5931814575195313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,1,4,power_law_1.2,0.03460223913192749
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,1,4,power_law_1.01,0.8716287994384766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,1,4,power_law_1.2,0.037489919662475585
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,1,4,power_law_1.2,0.04950272083282471
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,1,4,power_law_1.2,0.05185152053833007
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,1,4,power_law_1.2,0.05158016204833984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,1,4,power_law_1.2,0.07599232196807862
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,4,power_law_1.2,1.0379942321777345
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,1,4,power_law_1.2,0.0783193588256836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,1,4,power_law_1.2,0.08270336151123046
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,1,4,power_law_1.2,0.07878143787384033
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,1,4,power_law_1.2,0.0848102378845215
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,1,4,power_law_1.01,1.1391500854492187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,1,4,power_law_1.2,0.08579071998596191
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,1,4,power_law_1.2,0.08755968093872071
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,1,4,power_law_1.2,0.08797311782836914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,1,4,power_law_1.2,0.09093376159667968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,1,4,power_law_1.2,0.08826239585876465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,4,power_law_1.2,0.17477119445800782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,4,power_law_1.2,3.607924499511719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,4,power_law_1.2,1.475572509765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,1,4,power_law_1.2,0.10178303718566895
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,1,4,power_law_1.2,0.12155391693115233
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,1,4,power_law_1.2,0.09608575820922852
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,1,4,power_law_1.2,0.14078592300415038
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,1,4,power_law_1.2,0.18199167251586915
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,1,4,power_law_1.2,0.20694400787353517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,2,balanced,0.043610877990722656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,2,balanced,0.05607552051544189
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,2,balanced,0.07771647930145263
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,2,balanced,0.1187980842590332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,2,balanced,0.22653823852539062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,4,power_law_1.2,1.9649382019042967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,2,balanced,0.2268275260925293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,2,balanced,0.22590335845947265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,2,balanced,0.22845695495605467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,2,balanced,0.25892351150512694
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,2,balanced,0.2611136054992676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,2,balanced,0.2623999977111816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,2,balanced,0.26813568115234376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,2,balanced,0.2690444755554199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,2,balanced,0.27371648788452146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,2,balanced,0.2793574333190918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,2,balanced,0.2824038314819336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,2,balanced,0.2931174468994141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,2,balanced,0.3183705520629883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,2,balanced,0.34732799530029296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,2,balanced,0.4614156723022461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,2,balanced,0.5299456024169922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,2,balanced,0.7311078643798827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,2,balanced,0.9319526672363281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,2,balanced,1.3519500732421874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,2,balanced,1.7884889221191407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,1,4,power_law_1.2,0.2959526443481445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,2,balanced,0.03303807973861694
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,2,balanced,0.03612799882888794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,2,balanced,0.053614082336425783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,2,balanced,0.08974592208862305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,2,balanced,2.650360412597656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,2,balanced,0.13920127868652343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,2,balanced,0.14185088157653808
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,2,balanced,0.14239871978759766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,2,balanced,0.14293631553649902
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,2,balanced,0.15723008155822754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,2,balanced,0.15861503601074217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,2,balanced,0.1590028762817383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,2,balanced,0.16007936477661133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,2,balanced,0.16220031738281251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,2,balanced,0.16443904876708984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,2,balanced,0.16616575241088866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,2,balanced,3.5020928955078126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,2,balanced,0.17717632293701172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,2,balanced,0.18910463333129884
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,2,balanced,0.20890495300292972
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,2,balanced,0.2304921531677246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,2,balanced,0.16848512649536135
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,2,balanced,0.27091968536376954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,2,balanced,0.394257926940918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,2,balanced,0.4525196838378906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,2,balanced,0.6350579071044922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,2,balanced,0.826266860961914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,2,balanced,1.1891225433349608
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,2,balanced,1.5488858032226562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,1,4,power_law_1.2,0.3789145660400391
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,1,4,power_law_1.2,1.0816960144042969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,1,4,power_law_1.2,0.7888294219970703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,1,4,power_law_1.2,0.601957130432129
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,2,2,balanced,0.028023040294647215
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,2,2,balanced,0.09165184020996094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,2,2,balanced,0.09217151641845703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,2,2,balanced,0.09220735549926758
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,2,2,balanced,0.09244159698486329
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,2,2,balanced,0.030410239696502684
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,2,2,balanced,0.09263615608215332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,2,2,balanced,0.09342592239379884
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,2,2,balanced,0.09479424476623535
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,2,2,balanced,0.0951961612701416
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,2,2,balanced,0.09705344200134278
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,2,2,balanced,0.09897727966308593
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,2,2,balanced,0.10313856124877929
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,2,2,balanced,0.10780672073364257
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,2,2,balanced,0.11506815910339356
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,2,2,balanced,0.12222463607788085
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,2,2,balanced,0.04035840034484863
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,2,2,balanced,0.057148160934448244
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,2,2,balanced,0.09046784400939942
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,2,2,balanced,0.08982144355773926
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,2,2,balanced,0.0903116798400879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,2,power_law_1.01,0.05582719802856445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,2,power_law_1.01,0.06710271835327149
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,2,power_law_1.01,0.08782464027404785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,2,power_law_1.01,0.1131340789794922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,2,power_law_1.01,0.1526848030090332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,2,power_law_1.01,0.21607040405273437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,2,power_law_1.01,0.21918975830078127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,2,power_law_1.01,0.22180992126464844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,2,power_law_1.01,0.25416959762573244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,2,power_law_1.01,0.2570047950744629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,2,power_law_1.01,0.25857408523559566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,2,power_law_1.01,0.26222719192504884
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,2,power_law_1.01,0.2650470352172852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,2,power_law_1.01,0.27522304534912106
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,2,power_law_1.01,0.2797529602050781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,2,power_law_1.01,0.2859929656982422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,2,power_law_1.01,0.32718719482421876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,2,power_law_1.01,0.3666700744628906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,1,4,power_law_1.2,1.5452685546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,2,power_law_1.01,0.40708351135253906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,2,power_law_1.01,0.513836784362793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,2,power_law_1.01,0.6228403091430664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,2,power_law_1.01,0.8749209594726561
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,2,power_law_1.01,1.0747698974609374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,2,power_law_1.01,1.5927807617187502
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,2,power_law_1.01,2.0145765686035153
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,2,2,balanced,0.1723686408996582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,2,power_law_1.01,2.9237631225585936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,2,power_law_1.01,0.03215744018554688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,2,power_law_1.01,0.06032896041870117
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,2,2,balanced,0.2627814483642578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,2,2,balanced,0.18913408279418947
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,2,power_law_1.01,0.08646400451660156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,2,power_law_1.01,0.0934220790863037
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,2,power_law_1.01,0.12920703887939453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,2,power_law_1.01,0.13086591720581056
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,2,power_law_1.01,0.13443584442138673
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,2,power_law_1.01,0.14794624328613282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,2,power_law_1.01,0.14939776420593262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,2,power_law_1.01,0.04940159797668457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,2,power_law_1.01,0.15270015716552734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,2,power_law_1.01,0.1555238437652588
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,2,power_law_1.01,4.008523559570312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,2,power_law_1.01,0.16049280166625976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,2,power_law_1.01,0.16312704086303711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,2,power_law_1.01,0.16396160125732423
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,2,power_law_1.01,0.16780031204223633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,2,power_law_1.01,0.17886079788208006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,2,power_law_1.01,0.20482303619384767
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,2,power_law_1.01,0.22710399627685546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,2,power_law_1.01,0.2700390434265137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,2,power_law_1.01,0.32328319549560547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,2,power_law_1.01,0.415283203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,2,power_law_1.01,0.5273600006103516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,2,2,balanced,0.3383871841430664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,2,power_law_1.01,0.7274944305419921
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,2,2,power_law_1.01,0.030159358978271482
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,2,2,power_law_1.01,0.03873023986816406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,2,2,power_law_1.01,0.05443840026855469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,2,2,power_law_1.01,0.059024639129638665
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,2,2,power_law_1.01,0.07948800086975097
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,2,2,power_law_1.01,0.08214655876159668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,2,power_law_1.01,1.2900621032714843
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,2,2,power_law_1.01,0.08509056091308594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,2,2,power_law_1.01,0.08412672042846679
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,2,2,power_law_1.01,0.08778495788574218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,2,power_law_1.01,1.676691131591797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,2,2,power_law_1.01,0.0896332836151123
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,2,2,power_law_1.01,0.09423744201660156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,2,2,power_law_1.01,0.09469951629638672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,2,power_law_1.01,0.8766515350341797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,2,2,power_law_1.01,0.09880064010620118
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,2,2,power_law_1.01,0.09867775917053223
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,2,2,balanced,0.49427711486816406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,2,2,power_law_1.01,0.19865087509155271
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,2,2,balanced,0.6565964508056641
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,2,2,power_law_1.01,0.04503039836883545
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,2,2,power_law_1.01,0.11347968101501464
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,2,2,power_law_1.01,0.14518143653869628
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,2,2,power_law_1.01,0.10238335609436036
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,2,2,power_law_1.01,0.12987008094787597
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,2,power_law_1.2,0.054851841926574704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,2,power_law_1.2,0.06240896224975586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,2,power_law_1.2,0.08618368148803711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,2,power_law_1.2,0.11029120445251465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,2,power_law_1.2,0.1242956829071045
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,2,2,power_law_1.01,0.21607295989990236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,2,power_law_1.2,0.2168230438232422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,2,power_law_1.2,0.22042240142822264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,2,power_law_1.2,0.22093183517456055
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,2,2,power_law_1.01,0.28758399963378906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,2,power_law_1.2,0.2535833549499512
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,2,2,power_law_1.01,0.3837939071655273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,2,power_law_1.2,0.2539596748352051
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,2,power_law_1.2,0.25464576721191406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,2,power_law_1.2,0.26669183731079105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,2,power_law_1.2,0.2664857673645019
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,2,power_law_1.2,0.26739072799682617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,2,power_law_1.2,0.2766182327270508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,2,power_law_1.2,0.29111167907714847
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,2,power_law_1.2,0.3282406234741211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,2,power_law_1.2,0.37515777587890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,2,power_law_1.2,0.4135001754760742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,2,power_law_1.2,0.5326988983154297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,2,power_law_1.2,0.6423270416259765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,2,power_law_1.2,0.9198758697509766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,2,power_law_1.2,1.1144729614257813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,2,power_law_1.2,1.6194367980957032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,2,power_law_1.2,0.03341567993164062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,2,power_law_1.2,0.046058239936828616
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,2,power_law_1.2,2.1362022399902343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,2,power_law_1.2,0.0590284776687622
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,2,power_law_1.2,0.0871615982055664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,2,power_law_1.2,0.08590463638305665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,2,power_law_1.2,0.13167360305786133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,2,power_law_1.2,0.1321779155731201
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,2,power_law_1.2,0.13483263969421386
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,2,power_law_1.2,0.14831999778747557
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,2,power_law_1.2,0.1523251247406006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,2,power_law_1.2,0.15610495567321778
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,2,power_law_1.2,0.15752063751220705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,2,power_law_1.2,0.16172927856445313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,2,power_law_1.2,0.1647500801086426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,2,power_law_1.2,0.16657791137695313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,2,power_law_1.2,0.17180543899536133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,2,power_law_1.2,0.17879039764404298
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,2,power_law_1.2,0.21061248779296876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,2,power_law_1.2,0.23191680908203124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,2,power_law_1.2,0.2809228706359863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,2,power_law_1.2,0.3293376159667969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,2,power_law_1.2,0.4293183898925781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,2,power_law_1.2,0.5386252975463867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,2,power_law_1.2,0.7578495788574219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,2,power_law_1.2,4.035663452148437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,2,2,power_law_1.01,0.5217177581787109
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,2,2,power_law_1.2,0.029802238941192626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,2,2,power_law_1.2,0.03662463903427124
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,2,2,power_law_1.2,0.04530432224273682
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,2,2,power_law_1.2,0.050344958305358886
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,2,2,power_law_1.2,0.05833343982696533
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,2,2,power_law_1.2,0.07865344047546387
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,2,2,power_law_1.01,0.8034355163574218
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,2,2,power_law_1.2,0.08277888298034668
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,2,2,power_law_1.01,1.038040313720703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,2,2,power_law_1.2,0.08874496459960937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,2,2,power_law_1.2,0.08699904441833496
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,2,2,power_law_1.2,0.09120384216308594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,2,2,power_law_1.2,0.0928217601776123
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,2,2,power_law_1.2,0.09774975776672364
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,2,2,power_law_1.01,1.4072434997558594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,2,2,power_law_1.2,0.09946240425109863
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,2,2,power_law_1.2,0.10570112228393555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,2,2,power_law_1.2,0.10594304084777832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,2,2,power_law_1.2,0.08798336029052735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,2,power_law_1.2,0.8961907196044923
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,2,power_law_1.2,3.0095205688476563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,2,power_law_1.2,1.7665472412109373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,2,power_law_1.2,1.2929817199707032
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,2,2,power_law_1.2,0.1167897605895996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,2,2,power_law_1.2,0.14546560287475585
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,2,2,power_law_1.2,0.22976383209228515
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,2,2,power_law_1.2,0.16088703155517578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,2,2,power_law_1.2,0.281582088470459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,1,balanced,0.04717504024505616
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,1,balanced,0.06459328174591064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,1,balanced,0.08457280158996582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,1,balanced,0.13190272331237793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,1,balanced,0.23959104537963868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,1,balanced,0.2402195167541504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,1,balanced,0.24002944946289065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,1,balanced,0.24070079803466796
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,2,2,power_law_1.2,0.4432755279541015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,1,balanced,0.24085056304931643
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,1,balanced,0.2407462310791016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,1,balanced,0.24123775482177737
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,1,balanced,0.2711123275756836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,1,balanced,0.2757036781311035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,1,balanced,0.2798553657531738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,1,balanced,0.2832287979125977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,1,balanced,0.2894073677062988
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,1,balanced,0.2990496063232422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,1,balanced,0.32150527954101565
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,2,2,power_law_1.2,0.5681804656982422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,1,balanced,0.3475219345092774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,1,balanced,0.4542956924438476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,1,balanced,0.5051980972290039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,1,balanced,0.7012358093261719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,1,balanced,0.880679702758789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,1,balanced,1.3170451354980468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,1,balanced,0.03955071926116943
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,1,balanced,0.047205119132995604
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,1,balanced,1.7446400451660158
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,1,balanced,0.06684927940368653
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,1,balanced,0.1078758430480957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,1,balanced,0.17201215744018555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,1,balanced,0.17386688232421874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,1,balanced,0.1732441520690918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,1,balanced,0.1756742477416992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,1,balanced,0.17662208557128906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,1,balanced,0.17726463317871094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,1,balanced,2.50486083984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,1,balanced,0.17772287368774414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,1,balanced,0.20025087356567384
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,1,balanced,0.20492927551269532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,1,balanced,0.20379711151123048
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,1,balanced,0.20786815643310547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,1,balanced,0.21264896392822266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,1,balanced,0.2252115249633789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,1,balanced,0.23714303970336914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,1,balanced,3.391845703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,1,balanced,0.2623308753967285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,1,balanced,0.2851532745361328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,1,balanced,0.20402559280395507
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,1,balanced,0.43724224090576175
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,1,balanced,0.47699775695800783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,1,balanced,0.6631327819824219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,1,balanced,0.8613958740234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,1,balanced,1.2449830627441405
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,2,2,power_law_1.2,0.8902937316894531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,1,balanced,1.6443090820312498
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,2,2,power_law_1.2,1.014993896484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,2,2,power_law_1.2,1.5874278259277343
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,4,1,balanced,0.03943743944168091
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,4,1,balanced,0.042662401199340824
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,4,1,balanced,0.11321279525756836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,4,1,balanced,0.053754878044128415
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,4,1,balanced,0.11441216468811036
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,4,1,balanced,0.11524224281311035
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,4,1,balanced,0.0735097599029541
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,4,1,balanced,0.11576959609985352
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,4,1,balanced,0.11710144042968751
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,4,1,balanced,0.11823871612548828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,4,1,balanced,0.11984512329101563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,4,1,balanced,0.12128128051757811
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,4,1,balanced,0.12439104080200196
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,4,1,balanced,0.12783231735229492
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,4,1,balanced,0.1127455997467041
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,4,1,balanced,0.13513855934143065
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,4,1,balanced,0.14307647705078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,4,1,balanced,0.15864959716796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,4,1,balanced,0.1734227180480957
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,4,1,balanced,0.11344256401062011
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,4,1,balanced,0.11376895904541015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,1,power_law_1.01,0.03898175954818726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,1,power_law_1.01,0.06549952030181885
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,1,power_law_1.01,0.08452351570129395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,1,power_law_1.01,0.1129478359222412
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,1,power_law_1.01,0.1364179229736328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,1,power_law_1.01,0.2174188804626465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,1,power_law_1.01,0.22863168716430665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,1,power_law_1.01,0.23335872650146486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,1,power_law_1.01,0.23476032257080076
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,1,power_law_1.01,0.23491136550903322
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,1,power_law_1.01,0.2488863945007324
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,2,2,power_law_1.2,2.098635559082031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,1,power_law_1.01,0.26963455200195313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,1,power_law_1.01,0.2793868827819824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,1,power_law_1.01,0.27129791259765623
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,4,1,balanced,0.25315519332885744
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,4,1,balanced,0.405689582824707
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,4,1,balanced,0.28656511306762694
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,1,power_law_1.01,0.28495168685913086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,1,power_law_1.01,0.2885875129699707
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,4,1,balanced,0.5298579025268555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,1,power_law_1.01,0.30357631683349606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,1,power_law_1.01,0.3676486587524414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,1,power_law_1.01,0.39506622314453127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,1,power_law_1.01,0.4877900695800781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,1,power_law_1.01,0.579486083984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,4,1,balanced,0.7891808319091796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,1,power_law_1.01,0.7752812957763672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,1,power_law_1.01,0.03734911918640137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,1,power_law_1.01,0.9583551788330078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,1,power_law_1.01,0.049745922088623044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,1,power_law_1.01,0.07037888050079347
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,1,power_law_1.01,1.3606098937988282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,1,power_law_1.01,0.08790143966674804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,1,power_law_1.01,0.1179532814025879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,1,power_law_1.01,0.1619276809692383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,1,power_law_1.01,0.15942272186279297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,1,power_law_1.01,0.16615488052368163
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,1,power_law_1.01,1.7767205810546876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,1,power_law_1.01,0.16958784103393554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,1,power_law_1.01,0.16971071243286134
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,1,power_law_1.01,0.17379199981689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,1,power_law_1.01,0.19689151763916016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,1,power_law_1.01,0.20196096420288087
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,1,power_law_1.01,0.2028223991394043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,1,power_law_1.01,0.2056300735473633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,1,power_law_1.01,0.20777471542358397
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,1,power_law_1.01,0.22059392929077148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,1,power_law_1.01,0.23547584533691407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,1,power_law_1.01,0.2635539245605469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,1,power_law_1.01,2.5991809082031248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,1,power_law_1.01,0.2982169532775879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,1,power_law_1.01,0.3538399887084961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,1,power_law_1.01,0.4436217498779297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,1,power_law_1.01,0.5348006439208984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,1,power_law_1.01,0.7267231750488281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,1,power_law_1.01,3.412032775878906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,1,power_law_1.01,0.9230207824707032
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,4,1,power_law_1.01,0.03946304082870483
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,1,power_law_1.01,1.3040357971191407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,4,1,power_law_1.01,0.054444160461425785
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,4,1,power_law_1.01,0.06647168159484863
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,4,1,power_law_1.01,0.09881471633911133
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,4,1,power_law_1.01,0.07472576141357422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,1,power_law_1.01,1.6890521240234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,4,1,power_law_1.01,0.10423551559448244
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,4,1,power_law_1.01,0.10646400451660157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,4,1,power_law_1.01,0.10973631858825683
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,4,1,power_law_1.01,0.10934271812438964
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,4,1,power_law_1.01,0.11549440383911133
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,4,1,power_law_1.01,0.1174124813079834
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,4,1,power_law_1.01,0.12094016075134277
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,4,1,power_law_1.01,0.12815232276916505
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,4,1,balanced,1.053533477783203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,4,1,power_law_1.01,0.04504896163940429
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,4,1,power_law_1.01,0.15636608123779297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,4,1,power_law_1.01,0.14157695770263673
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,4,1,power_law_1.01,0.13746432304382322
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,4,1,power_law_1.01,0.18516416549682618
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,4,1,power_law_1.01,0.2139046478271484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,4,1,power_law_1.01,0.2804908752441406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,1,power_law_1.2,0.04347583770751953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,1,power_law_1.2,0.06643712043762207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,1,power_law_1.2,0.08255999565124512
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,4,1,power_law_1.01,0.35272830963134766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,1,power_law_1.2,0.10571392059326172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,1,power_law_1.2,0.13045632362365722
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,1,power_law_1.2,0.2242259216308594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,1,power_law_1.2,0.23122304916381836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,1,power_law_1.2,0.23540800094604492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,1,power_law_1.2,0.2353139114379883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,1,power_law_1.2,0.23621376037597655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,1,power_law_1.2,0.25067136764526365
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,1,power_law_1.2,0.26870336532592776
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,1,power_law_1.2,0.2710291290283203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,1,power_law_1.2,0.27866239547729493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,1,power_law_1.2,0.2828940773010254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,1,power_law_1.2,0.28912576675415036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,1,power_law_1.2,0.30773248672485354
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,1,power_law_1.2,0.368579216003418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,1,power_law_1.2,0.3991596984863281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,1,power_law_1.2,0.5001779174804687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,1,power_law_1.2,0.5800326538085938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,1,power_law_1.2,0.7726566314697265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,1,power_law_1.2,0.9758316802978516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,1,power_law_1.2,1.3689222717285157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,4,1,power_law_1.01,0.5014060974121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,4,1,power_law_1.01,0.6514054107666015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,1,power_law_1.2,0.037391359806060794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,1,power_law_1.2,0.05007743835449219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,4,1,power_law_1.01,0.9674463653564453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,1,power_law_1.2,0.06601408004760742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,1,power_law_1.2,0.08099455833435058
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,4,1,power_law_1.01,1.2674591827392578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,1,power_law_1.2,0.11291135787963866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,1,power_law_1.2,0.15684543609619142
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,1,power_law_1.2,0.16401920318603516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,1,power_law_1.2,0.16536447525024414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,1,power_law_1.2,0.17042623519897462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,1,power_law_1.2,0.17104768753051758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,1,power_law_1.2,0.1737830352783203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,4,1,power_law_1.01,1.9083110046386718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,1,power_law_1.2,0.19799039840698243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,1,power_law_1.2,1.7961888122558594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,1,power_law_1.2,0.20811967849731444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,1,power_law_1.2,0.21614336013793944
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,4,1,power_law_1.01,2.5856781005859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,1,power_law_1.2,0.22131391525268557
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,1,power_law_1.2,0.2476268768310547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,1,power_law_1.2,0.1972991943359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,1,power_law_1.2,0.26654399871826173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,1,power_law_1.2,0.3133497619628906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,1,power_law_1.2,0.20396223068237304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,1,power_law_1.2,0.3559987258911133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,1,power_law_1.2,0.44625728607177734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,4,1,power_law_1.2,0.0398035192489624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,1,power_law_1.2,0.5478604888916015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,4,1,power_law_1.2,0.044882559776306154
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,4,1,power_law_1.2,0.053626880645751954
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,4,1,power_law_1.2,0.06214144229888916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,4,1,power_law_1.2,0.07109119892120361
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,4,1,power_law_1.2,0.0997862434387207
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,4,1,power_law_1.2,0.10461695671081543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,4,1,power_law_1.2,0.10836416244506837
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,4,1,power_law_1.2,0.11024255752563476
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,4,1,power_law_1.2,0.1121132755279541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,1,power_law_1.2,0.7326700592041016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,4,1,power_law_1.2,0.11652671813964843
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,4,1,power_law_1.2,0.12274239540100099
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,4,1,power_law_1.2,0.12539584159851075
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,4,1,power_law_1.2,0.13398207664489745
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,4,1,power_law_1.2,0.14492032051086426
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,4,1,power_law_1.2,0.15253952026367187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,1,power_law_1.2,0.9255699157714844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,1,power_law_1.2,2.5302476501464843
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,4,1,power_law_1.2,0.17172927856445314
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,4,1,power_law_1.2,0.2151103973388672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,4,1,power_law_1.2,0.2586489677429199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,1,power_law_1.2,1.3068006896972657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,4,1,power_law_1.2,0.355425910949707
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,4,1,power_law_1.2,0.4566835021972656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,4,1,power_law_1.2,0.6627410888671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,1,4,balanced,0.08219264030456544
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,1,4,balanced,0.09585663795471192
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,1,4,balanced,0.14631808280944825
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,1,4,balanced,0.3043801689147949
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,1,power_law_1.2,3.380783386230469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,1,4,balanced,0.5189440155029297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,4,1,power_law_1.2,0.8762944030761719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,1,4,balanced,0.668911361694336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,1,4,balanced,0.7659865570068359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,1,4,balanced,0.7731110382080078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,1,4,balanced,0.7541043090820312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,1,4,balanced,0.7685388946533204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,1,4,balanced,0.7690739440917969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,1,4,balanced,0.7768473815917969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,1,4,balanced,0.7878745269775391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,1,4,balanced,0.7824960327148437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,1,4,balanced,0.7989593505859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,1,4,balanced,0.8125145721435547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,1,4,balanced,0.8111090850830077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,1,4,balanced,0.8552601623535157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,1,4,balanced,0.9100915527343749
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,4,1,power_law_1.2,1.3110226440429689
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,1,4,balanced,1.3661016845703124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,1,4,balanced,1.4410572814941407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,1,power_law_1.2,1.6952204895019531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,1,4,balanced,2.101405487060547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,1,4,balanced,2.58790283203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,1,4,balanced,0.05009920120239257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,1,4,balanced,0.0630463981628418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,1,4,balanced,0.10621312141418457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,1,4,balanced,0.16402687072753905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,1,4,balanced,3.6458419799804687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,1,4,balanced,0.2856038475036621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,1,4,balanced,0.3496883010864258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,1,4,balanced,0.37512447357177736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,1,4,balanced,0.3801651382446289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,1,4,balanced,0.3797529602050781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,1,4,balanced,0.38605697631835934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,1,4,balanced,0.37436031341552733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,1,4,balanced,5.18474853515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,1,4,balanced,0.3804991912841797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,1,4,balanced,0.3770073699951172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,1,4,balanced,0.3826700973510742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,1,4,balanced,0.38391422271728515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,1,4,balanced,0.3874777603149414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,1,4,balanced,0.40040958404541016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,1,4,balanced,0.45262081146240235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,1,4,balanced,0.4619110488891602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,1,4,balanced,0.5250713729858398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,1,4,balanced,0.591733741760254
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,4,1,power_law_1.2,1.7758937072753906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,1,4,balanced,0.8161203002929687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,1,4,balanced,7.8700976562500005
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,1,4,balanced,0.9490751647949219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,1,4,balanced,1.369713897705078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,1,4,balanced,1.7369139099121091
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,1,4,balanced,2.3658546447753905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,1,4,balanced,3.123468933105469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,1,4,balanced,10.705028076171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,4,1,power_law_1.2,2.701712646484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,1,4,balanced,0.027246079444885253
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,1,4,balanced,0.19627519607543947
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,1,4,balanced,0.19695743560791015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,1,4,balanced,0.04347648143768311
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,1,4,balanced,0.19718015670776368
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,1,4,balanced,0.19769983291625975
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,1,4,balanced,0.197706241607666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,1,4,balanced,0.05994495868682861
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,1,4,balanced,0.1975641632080078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,1,4,balanced,0.20011648178100586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,1,4,balanced,0.20091903686523435
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,1,4,balanced,0.20130687713623047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,1,4,balanced,0.20428672790527344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,1,4,balanced,0.2092889595031738
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,1,4,balanced,0.21456768035888674
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,1,4,balanced,0.2242304039001465
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,1,4,balanced,0.23328128814697266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,1,4,balanced,0.09469951629638672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,1,4,balanced,0.16269695281982421
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,1,4,balanced,0.19594751358032228
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,1,4,balanced,0.1959449577331543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,1,4,power_law_1.01,0.14361215591430665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,1,4,power_law_1.01,0.24878208160400392
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,1,4,power_law_1.01,0.2693734359741211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,1,4,power_law_1.01,0.3689689636230469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,1,4,power_law_1.01,0.37445888519287107
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,1,4,power_law_1.01,0.5928652954101563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,1,4,power_law_1.01,0.6322969436645508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,1,4,power_law_1.01,0.623432960510254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,1,4,power_law_1.01,0.6383385467529297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,1,4,power_law_1.01,0.6443302154541015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,1,4,power_law_1.01,0.6799065399169921
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,1,4,power_law_1.01,0.70308349609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,1,4,power_law_1.01,0.7326860809326172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,1,4,balanced,0.36467838287353516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,1,4,balanced,0.3423052978515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,1,4,power_law_1.01,0.7722061157226563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,1,4,balanced,0.505533447265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,1,4,balanced,0.657760009765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,1,4,power_law_1.01,0.8066073608398436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,1,4,power_law_1.01,0.835546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,1,4,power_law_1.01,0.8852761840820312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,1,4,balanced,0.8226802825927735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,1,4,power_law_1.01,1.0665126037597656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,1,4,power_law_1.01,1.2624703979492187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,1,4,power_law_1.01,1.4779008483886718
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,4,1,power_law_1.2,3.6026931762695313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,1,4,power_law_1.01,1.7569471740722655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,1,4,power_law_1.01,0.10653568267822267
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,1,4,balanced,1.2457164764404296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,1,4,power_law_1.01,2.4647013854980466
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,1,4,power_law_1.01,0.16150272369384766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,1,4,power_law_1.01,0.15726079940795898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,1,4,power_law_1.01,0.22739200592041015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,1,4,power_law_1.01,3.0179827880859373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,1,4,power_law_1.01,0.29609088897705077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,1,4,power_law_1.01,0.3378252792358398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,1,4,power_law_1.01,0.3208998489379883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,1,4,power_law_1.01,0.33890174865722655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,1,4,power_law_1.01,0.3651737594604492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,1,4,power_law_1.01,0.19978496551513672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,1,4,power_law_1.01,0.3633651351928711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,1,4,power_law_1.01,0.3563264083862305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,1,4,power_law_1.01,0.3602803039550781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,1,4,power_law_1.01,0.3587123107910156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,1,4,power_law_1.01,0.3558464050292969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,1,4,power_law_1.01,4.889784240722657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,1,4,power_law_1.01,0.3728550338745117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,1,4,power_law_1.01,0.3908953475952148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,1,4,power_law_1.01,0.4586419296264649
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,1,4,power_law_1.01,0.48244480133056644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,1,4,power_law_1.01,0.5905036926269531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,1,4,power_law_1.01,0.6827366638183594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,1,4,power_law_1.01,0.867914276123047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,1,4,power_law_1.01,5.859877319335938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,1,4,power_law_1.01,1.0615974426269532
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,1,4,power_law_1.01,0.06496384143829345
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,1,4,power_law_1.01,0.09756928443908691
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,1,4,power_law_1.01,1.4971994018554686
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,1,4,power_law_1.01,0.10078847885131836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,1,4,power_law_1.01,0.12656895637512205
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,1,4,power_law_1.01,0.1355840015411377
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,1,4,power_law_1.01,0.1787071990966797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,1,4,power_law_1.01,0.1830899238586426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,1,4,power_law_1.01,2.023390655517578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,1,4,power_law_1.01,0.18454399108886718
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,1,4,power_law_1.01,0.1866227149963379
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,1,4,power_law_1.01,0.18651391983032226
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,1,4,power_law_1.01,0.19538431167602538
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,1,4,power_law_1.01,0.19382144927978515
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,1,4,power_law_1.01,0.19453439712524415
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,1,4,power_law_1.01,0.19725824356079102
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,1,4,power_law_1.01,0.20393087387084963
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,1,4,power_law_1.01,0.20683263778686523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,1,4,power_law_1.01,2.761614074707031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,1,4,power_law_1.01,0.31509376525878907
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,1,4,power_law_1.01,0.3956492614746094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,1,4,power_law_1.01,8.609896850585937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,1,4,power_law_1.01,3.6879974365234376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,1,4,power_law_1.01,10.263134765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,1,4,power_law_1.01,0.2212416076660156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,1,4,power_law_1.01,0.24116096496582032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,1,4,power_law_1.2,0.14348287582397462
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,1,4,power_law_1.01,0.26699008941650393
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,1,4,power_law_1.2,0.15889280319213867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,1,4,power_law_1.01,0.46057216644287113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,1,4,power_law_1.2,0.3091929626464844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,1,4,power_law_1.2,0.3446937561035156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,1,4,power_law_1.01,0.67338623046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,1,4,power_law_1.2,0.3509491348266601
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,1,4,power_law_1.2,0.5830246353149414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,1,4,power_law_1.2,0.6012160110473632
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,1,4,power_law_1.01,0.8627110290527344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,1,4,power_law_1.2,0.6599282836914062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,1,4,power_law_1.2,0.6547904205322266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,1,4,power_law_1.2,0.6744908905029297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,1,4,power_law_1.2,0.7176064300537109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,1,4,power_law_1.2,0.7600921630859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,1,4,power_law_1.2,0.7847577667236327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,1,4,power_law_1.2,0.8137612915039062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,1,4,power_law_1.2,0.8413887786865235
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,1,4,power_law_1.01,1.179192352294922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,1,4,power_law_1.2,0.8510566711425781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,1,4,power_law_1.2,0.9397695922851563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,1,4,power_law_1.2,1.0734105682373047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,1,4,power_law_1.2,1.2094322967529298
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,1,4,power_law_1.01,1.554062042236328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,1,4,power_law_1.2,1.536776885986328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,1,4,power_law_1.2,1.8676185607910156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,1,4,power_law_1.2,0.10600447654724121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,1,4,power_law_1.2,0.10821887969970703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,1,4,power_law_1.2,2.741187744140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,1,4,power_law_1.2,0.15468799591064453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,1,4,power_law_1.2,0.18659711837768556
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,1,4,power_law_1.2,0.20871295928955078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,1,4,power_law_1.2,0.301584644317627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,1,4,power_law_1.2,3.62853515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,1,4,power_law_1.2,0.33563392639160156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,1,4,power_law_1.2,0.3371891021728516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,1,4,power_law_1.2,0.3401420974731445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,1,4,power_law_1.2,0.36177791595458986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,1,4,power_law_1.2,0.3697907257080078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,1,4,power_law_1.2,0.36805889129638675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,1,4,power_law_1.2,0.36801025390625003
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,1,4,power_law_1.2,0.3769715118408203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,1,4,power_law_1.2,0.38280448913574217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,1,4,power_law_1.2,0.38074623107910155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,1,4,power_law_1.2,5.555360107421874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,1,4,power_law_1.2,0.4097574234008789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,1,4,power_law_1.2,0.47231487274169925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,1,4,power_law_1.2,0.5153996658325195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,1,4,power_law_1.2,0.6084223937988281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,1,4,power_law_1.2,0.7053004455566406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,1,4,power_law_1.2,0.9281996917724609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,1,4,power_law_1.2,1.154856948852539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,1,4,power_law_1.2,1.682298889160156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,1,4,power_law_1.2,3.4350808715820316
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,1,4,power_law_1.2,0.06475264072418213
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,1,4,power_law_1.2,0.07048831939697266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,1,4,power_law_1.2,0.0944422435760498
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,1,4,power_law_1.2,0.11570303916931153
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,1,4,power_law_1.2,2.1406016540527344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,1,4,power_law_1.2,6.233922729492187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,1,4,power_law_1.2,0.12500224113464356
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,1,4,power_law_1.2,0.17412736892700195
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,1,4,power_law_1.2,0.1750579261779785
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,1,4,power_law_1.2,0.1688422393798828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,1,4,power_law_1.2,0.18943744659423828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,1,4,power_law_1.2,0.18980224609375002
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,1,4,power_law_1.2,0.19314815521240233
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,1,4,power_law_1.2,0.1929599952697754
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,1,4,power_law_1.2,0.20261119842529296
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,1,4,power_law_1.2,0.20225664138793947
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,1,4,power_law_1.2,0.23291135787963868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,1,4,power_law_1.2,9.520424804687499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,1,4,power_law_1.2,3.875554504394531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,1,4,power_law_1.2,14.401678466796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,1,4,power_law_1.01,2.0527859497070313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,1,4,power_law_1.2,0.2109312057495117
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,1,4,power_law_1.2,0.2183807945251465
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,1,4,power_law_1.2,0.2679897689819336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,1,4,power_law_1.2,0.3014131164550781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,1,4,power_law_1.2,0.3858662414550781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,2,2,balanced,0.07246719837188721
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,2,2,balanced,0.09022975921630859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,2,2,balanced,0.14737152099609374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,2,2,balanced,0.3130803108215332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,2,2,balanced,0.5054246520996093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,2,2,balanced,0.6486105346679688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,2,2,balanced,0.6512358093261719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,2,2,balanced,0.6531136322021485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,2,2,balanced,0.6543936157226563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,1,4,power_law_1.2,0.458996467590332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,2,2,balanced,0.7391423797607422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,2,2,balanced,0.7381657409667969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,1,4,power_law_1.2,0.6274380874633789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,2,2,balanced,0.7510233306884766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,2,2,balanced,0.7597299194335937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,2,2,balanced,0.7623628997802735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,2,2,balanced,0.7707839965820312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,2,2,balanced,0.7520256042480469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,2,2,balanced,0.7681600189208984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,2,2,balanced,0.8126374053955079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,2,2,balanced,0.8614335632324218
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,1,4,power_law_1.2,0.80890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,2,2,balanced,1.3375053405761719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,2,2,balanced,1.371751708984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,2,2,balanced,1.9918130493164063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,2,2,balanced,2.494398651123047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,2,2,balanced,3.303377990722656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,2,2,balanced,0.048857598304748534
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,2,2,balanced,0.06408959865570069
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,2,2,balanced,0.10520575523376466
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,2,2,balanced,4.330064697265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,2,2,balanced,0.17379840850830078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,2,2,balanced,0.2959244728088379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,2,2,balanced,0.3542899322509766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,2,2,balanced,0.35362174987792966
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,2,2,balanced,0.35468799591064454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,2,2,balanced,0.3563507080078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,2,2,balanced,0.3853772735595703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,2,2,balanced,0.38470016479492186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,2,2,balanced,0.3912652969360352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,2,2,balanced,0.38868736267089843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,2,2,balanced,0.39652481079101565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,2,2,balanced,6.161699829101563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,2,2,balanced,0.39429630279541017
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,2,2,balanced,0.3985228729248047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,2,2,balanced,0.41530879974365237
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,2,2,balanced,0.4351897430419922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,2,2,balanced,0.4767846298217774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,2,2,balanced,0.499238395690918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,2,2,balanced,0.5649267196655273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,2,2,balanced,0.836847381591797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,2,2,balanced,0.90281982421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,2,2,balanced,1.325896911621094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,2,2,balanced,8.648793334960938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,2,2,balanced,1.7033203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,2,2,balanced,2.272564544677734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,2,2,balanced,3.0099942016601564
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,1,4,power_law_1.2,1.29957763671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,1,4,power_law_1.2,1.463031005859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,1,4,power_law_1.2,2.8644082641601565
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,1,4,power_law_1.2,2.346782684326172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,2,2,balanced,0.03467904090881348
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,2,2,balanced,0.20877824783325197
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,2,2,balanced,0.2094361686706543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,2,2,balanced,0.046995201110839845
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,2,2,balanced,0.21119743347167969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,2,2,balanced,0.21097984313964843
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,2,2,balanced,0.211727352142334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,2,2,balanced,0.06617599964141846
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,2,2,balanced,0.21354623794555666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,2,2,balanced,0.2165542411804199
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,2,2,balanced,0.21781631469726564
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,2,2,balanced,0.22137088775634767
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,2,2,balanced,0.22870399475097658
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,2,2,balanced,0.23553535461425784
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,2,2,balanced,0.24998527526855469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,2,2,balanced,0.264451847076416
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,2,2,balanced,0.10040320396423338
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,2,2,balanced,0.17144704818725587
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,2,2,balanced,0.2076505661010742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,2,2,balanced,0.20820608139038085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,2,2,power_law_1.01,0.09100799560546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,2,2,power_law_1.01,0.13165568351745605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,2,2,power_law_1.01,0.17372800827026366
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,2,2,power_law_1.01,0.2936332893371582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,2,2,power_law_1.01,0.33126144409179686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,2,2,power_law_1.01,0.5089868927001954
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,2,2,balanced,0.2086489677429199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,2,2,power_law_1.01,0.527768325805664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,2,2,power_law_1.01,0.5582118225097656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,2,2,power_law_1.01,0.567327995300293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,2,2,power_law_1.01,0.6270323181152344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,2,2,power_law_1.01,0.6572415924072266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,2,2,power_law_1.01,0.6645491027832031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,2,2,power_law_1.01,0.6870732879638671
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,2,2,power_law_1.01,0.7383372497558593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,2,2,power_law_1.01,0.7637529754638672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,2,2,power_law_1.01,0.7911039733886719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,2,2,power_law_1.01,0.8153228759765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,2,2,power_law_1.01,0.920794906616211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,2,2,power_law_1.01,1.0622297668457032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,2,2,power_law_1.01,1.387034912109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,2,2,power_law_1.01,1.6180519104003905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,2,2,power_law_1.01,2.1415219116210937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,2,2,power_law_1.01,2.6933798217773437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,2,2,balanced,0.3861299133300781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,2,2,power_law_1.01,3.938236083984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,2,2,balanced,0.5891852951049804
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,2,2,balanced,0.41934078216552734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,2,2,power_law_1.01,0.06550399780273437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,2,2,power_law_1.01,0.08891776084899902
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,2,2,power_law_1.01,0.12081279754638671
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,2,2,power_law_1.01,0.16558719635009767
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,2,2,balanced,0.7721382141113281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,2,2,power_law_1.01,0.20543359756469726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,2,2,power_law_1.01,0.28821119308471677
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,2,2,power_law_1.01,0.30968959808349605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,2,2,power_law_1.01,4.761904602050781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,2,2,power_law_1.01,0.3194483184814453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,2,2,power_law_1.01,0.3157107162475586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,2,2,power_law_1.01,0.3616499328613281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,2,2,power_law_1.01,0.3437836837768555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,2,2,balanced,1.0485171508789062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,2,2,power_law_1.01,0.3624806213378906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,2,2,power_law_1.01,0.3675763320922852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,2,2,power_law_1.01,0.3626982498168945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,2,2,power_law_1.01,0.38182273864746097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,2,2,power_law_1.01,0.3769804763793945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,2,2,power_law_1.01,0.39305728912353516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,2,2,power_law_1.01,0.43572097778320307
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,2,2,power_law_1.01,0.4724761581420898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,2,2,power_law_1.01,0.5623129653930664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,2,2,power_law_1.01,0.6377881622314453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,2,2,power_law_1.01,0.8429888153076173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,2,2,power_law_1.01,0.9887232208251954
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,2,2,power_law_1.01,0.05474944114685058
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,2,2,power_law_1.01,1.4013261413574218
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,2,2,balanced,1.4307788085937498
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,2,2,power_law_1.01,0.06874752044677734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,2,2,power_law_1.01,0.07755263805389404
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,2,2,power_law_1.01,6.810105590820312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,2,2,power_law_1.01,0.11297408103942871
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,2,2,power_law_1.01,0.10217984199523926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,2,2,power_law_1.01,1.758408966064453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,2,2,power_law_1.01,0.1756070327758789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,2,2,power_law_1.01,0.18510208129882813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,2,2,power_law_1.01,0.18895103454589843
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,2,2,power_law_1.01,0.19503488540649414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,2,2,power_law_1.01,0.19933696746826174
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,2,2,power_law_1.01,0.20558208465576172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,2,2,power_law_1.01,0.21051263809204102
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,2,2,power_law_1.01,0.21096063613891602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,2,2,power_law_1.01,2.5261669921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,2,2,power_law_1.01,0.21831296920776366
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,2,2,power_law_1.01,0.2231884765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,2,2,power_law_1.01,0.16156671524047853
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,2,2,power_law_1.01,0.2770483207702637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,2,2,power_law_1.01,3.139808044433594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,2,2,power_law_1.01,8.861181640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,2,2,power_law_1.01,0.2409971237182617
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,2,2,power_law_1.01,0.3606975936889648
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,2,2,power_law_1.01,0.29700992584228514
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,2,2,power_law_1.01,0.41953407287597655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,2,2,power_law_1.2,0.09056768417358399
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,2,2,power_law_1.01,0.7357810974121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,2,2,power_law_1.01,0.5908838272094726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,2,2,power_law_1.2,0.12689151763916015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,2,2,power_law_1.2,0.17628799438476564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,2,2,power_law_1.2,0.2931507110595703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,2,2,power_law_1.2,0.34005630493164063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,2,2,power_law_1.2,0.5013542556762696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,2,2,power_law_1.2,0.5192895889282226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,2,2,power_law_1.2,0.5322009658813476
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,2,2,power_law_1.01,1.0293734741210936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,2,2,power_law_1.2,0.5670528030395507
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,2,2,power_law_1.2,0.6419776153564454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,2,2,power_law_1.2,0.6667763519287109
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,2,2,power_law_1.01,1.391065673828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,2,2,power_law_1.2,0.7166694641113281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,2,2,power_law_1.2,0.7127693176269532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,2,2,power_law_1.2,0.7798054504394532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,2,2,power_law_1.2,0.7965862274169921
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,2,2,power_law_1.2,0.8076825714111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,2,2,power_law_1.2,0.8662681579589844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,2,2,power_law_1.2,0.97272705078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,2,2,power_law_1.2,1.0909158325195312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,2,2,power_law_1.2,1.3756121826171874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,2,2,power_law_1.2,1.6665126037597655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,2,2,power_law_1.2,2.099939880371094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,2,2,power_law_1.2,0.06383872032165527
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,2,2,power_law_1.2,2.792341613769531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,2,2,power_law_1.2,0.11960960388183595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,2,2,power_law_1.2,0.15968383789062499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,2,2,power_law_1.2,0.07518080234527588
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,2,2,power_law_1.2,0.188670711517334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,2,2,power_law_1.2,4.293377380371093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,2,2,power_law_1.2,0.2886451148986816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,2,2,power_law_1.2,0.3006015968322754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,2,2,power_law_1.2,0.3073945617675781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,2,2,power_law_1.2,0.32879360198974605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,2,2,power_law_1.2,0.3561676788330078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,2,2,power_law_1.2,0.3612364959716797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,2,2,power_law_1.2,0.3717375946044922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,2,2,power_law_1.2,0.367718391418457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,2,2,power_law_1.2,0.3879347229003906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,2,2,power_law_1.2,0.3896857452392578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,2,2,power_law_1.2,0.3999347305297852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,2,2,power_law_1.2,0.4157696151733399
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,2,2,power_law_1.2,0.4553689575195312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,2,2,power_law_1.2,0.582666244506836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,2,2,power_law_1.2,7.138989868164063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,2,2,power_law_1.2,0.6657190704345703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,2,2,power_law_1.2,0.4981222534179688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,2,2,power_law_1.2,0.8557823944091798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,2,2,power_law_1.2,1.0686348724365236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,2,2,power_law_1.2,1.4358592224121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,2,2,power_law_1.2,4.903379211425781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,2,2,power_law_1.2,0.05419007778167725
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,2,2,power_law_1.2,0.05707263946533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,2,2,power_law_1.2,0.07272704124450684
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,2,2,power_law_1.2,0.08659711837768555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,2,2,power_law_1.2,2.7850418090820312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,2,2,power_law_1.2,0.10011391639709473
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,2,2,power_law_1.2,0.17430015563964846
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,2,2,power_law_1.2,0.1750387191772461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,2,2,power_law_1.2,0.1917375946044922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,2,2,power_law_1.2,0.1936396789550781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,2,2,power_law_1.2,1.832743682861328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,2,2,power_law_1.2,0.19805952072143557
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,2,2,power_law_1.2,0.20703359603881838
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,2,2,power_law_1.2,0.208407039642334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,2,2,power_law_1.2,0.21252864837646485
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,2,2,power_law_1.2,0.2157414436340332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,2,2,power_law_1.2,9.566458740234376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,2,2,power_law_1.2,3.384207458496094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,2,2,power_law_1.01,2.0909465026855467
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,2,2,power_law_1.01,2.5924197387695314
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,2,2,power_law_1.2,0.23023616790771487
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,2,2,power_law_1.2,0.2527347183227539
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,2,2,power_law_1.2,0.24368383407592775
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,2,2,power_law_1.2,0.29648895263671876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,2,2,power_law_1.2,0.33853313446044925
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,2,2,power_law_1.2,0.44320129394531255
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,4,1,balanced,0.08085887908935548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,4,1,balanced,0.10088064193725585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,4,1,balanced,0.1538035202026367
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,2,2,power_law_1.2,0.5074713516235352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,4,1,balanced,0.3217055892944336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,4,1,balanced,0.5162073516845703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,4,1,balanced,0.6630342102050781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,2,2,power_law_1.2,0.7200550079345703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,4,1,balanced,0.6631110382080079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,4,1,balanced,0.6652051544189452
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,4,1,balanced,0.66544189453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,4,1,balanced,0.6647910308837891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,4,1,balanced,0.6705305480957031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,4,1,balanced,0.6741350555419923
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,4,1,balanced,0.7425337219238282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,4,1,balanced,0.7634893035888671
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,4,1,balanced,0.7796345520019531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,4,1,balanced,0.7580358123779296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,4,1,balanced,0.7767526245117187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,4,1,balanced,0.8092537689208985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,4,1,balanced,0.881283187866211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,4,1,balanced,1.3170867919921876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,4,1,balanced,1.4129728698730468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,4,1,balanced,1.9505952453613282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,4,1,balanced,2.508135070800781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,4,1,balanced,3.188074951171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,4,1,balanced,4.2012197875976565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,4,1,balanced,0.06540160179138184
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,4,1,balanced,0.0775878381729126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,4,1,balanced,0.1257049560546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,4,1,balanced,0.1918400001525879
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,2,2,power_law_1.2,0.9818125152587891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,4,1,balanced,0.3299097442626953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,4,1,balanced,0.4090361785888672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,4,1,balanced,0.4031033706665039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,4,1,balanced,0.4019449615478516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,4,1,balanced,0.41034816741943364
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,4,1,balanced,0.4083219146728515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,4,1,balanced,6.0390039062500005
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,4,1,balanced,0.41316287994384765
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,2,2,power_law_1.2,1.5858816528320312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,4,1,balanced,0.4145568084716797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,4,1,balanced,0.46884414672851565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,4,1,balanced,0.46972606658935545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,4,1,balanced,0.4719136047363281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,4,1,balanced,0.47519744873046876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,4,1,balanced,0.4839142227172852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,4,1,balanced,0.5067903900146484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,4,1,balanced,0.5267071914672852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,4,1,balanced,0.5677638244628906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,4,1,balanced,0.6145683288574219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,4,1,balanced,0.9001203155517578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,2,2,power_law_1.2,3.233446350097656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,4,1,balanced,0.994163818359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,4,1,balanced,7.87774658203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,4,1,balanced,1.44238525390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,2,2,power_law_1.2,1.9745535278320312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,4,1,balanced,1.8417593383789062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,4,1,balanced,2.437727966308594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,2,2,power_law_1.2,3.889588623046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,4,1,balanced,3.19274169921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,4,1,balanced,0.06039999961853028
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,4,1,balanced,0.04785280227661133
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,4,1,balanced,0.07896448135375976
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,4,1,balanced,0.23880128860473632
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,4,1,balanced,0.2409734344482422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,4,1,balanced,0.11767807960510254
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,4,1,balanced,0.24202239990234373
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,4,1,balanced,0.1968351936340332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,4,1,balanced,0.242303352355957
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,4,1,balanced,0.2458803176879883
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,4,1,balanced,0.24872768402099607
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,4,1,balanced,0.2510092735290527
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,4,1,balanced,0.23660736083984374
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,4,1,balanced,0.2598419189453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,4,1,balanced,0.2729849624633789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,4,1,balanced,0.2849350357055664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,4,1,balanced,0.2537356758117676
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,4,1,balanced,0.308035831451416
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,4,1,balanced,0.3330060958862305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,4,1,balanced,0.23838911056518555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,4,1,power_law_1.01,0.08136639595031739
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,4,1,power_law_1.01,0.10131135940551758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,4,1,power_law_1.01,0.15374079704284668
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,4,1,balanced,0.23781696319580078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,4,1,power_law_1.01,0.2826297569274902
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,4,1,power_law_1.01,0.3393612670898437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,4,1,power_law_1.01,0.5112940979003906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,4,1,power_law_1.01,0.5421747207641602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,4,1,power_law_1.01,0.570775032043457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,4,1,power_law_1.01,0.5960236740112305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,4,1,power_law_1.01,0.6145331192016601
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,4,1,power_law_1.01,0.6391219329833985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,4,1,power_law_1.01,0.695257568359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,4,1,power_law_1.01,0.6981862640380859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,4,1,power_law_1.01,0.7603718566894531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,4,1,power_law_1.01,0.7780569458007813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,4,1,power_law_1.01,0.7789273834228515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,4,1,power_law_1.01,0.8136249542236328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,4,1,power_law_1.01,0.924176025390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,4,1,power_law_1.01,1.0217145538330077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,4,1,power_law_1.01,1.2587602996826173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,4,1,power_law_1.01,1.5007788085937501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,4,1,power_law_1.01,1.9631417846679688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,4,1,power_law_1.01,2.378626556396484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,4,1,power_law_1.01,3.3456512451171876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,4,1,power_law_1.01,4.2373809814453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,4,1,balanced,0.5116985702514649
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,4,1,power_law_1.01,0.05620863914489747
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,4,1,balanced,0.5818316650390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,4,1,power_law_1.01,0.07576064109802247
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,4,1,balanced,0.858243179321289
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,4,1,balanced,1.1139826965332031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,4,1,power_law_1.01,0.1609766387939453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,4,1,power_law_1.01,0.13024831771850587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,4,1,power_law_1.01,0.22664512634277342
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,4,1,power_law_1.01,6.105361328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,4,1,power_law_1.01,0.3333292770385742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,4,1,balanced,1.5233016967773438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,4,1,power_law_1.01,0.346506233215332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,4,1,power_law_1.01,0.35612670898437504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,4,1,power_law_1.01,0.36433406829833986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,4,1,power_law_1.01,0.37387134552001955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,4,1,power_law_1.01,0.3808294296264648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,4,1,power_law_1.01,0.39604991912841797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,4,1,power_law_1.01,0.43179969787597655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,4,1,power_law_1.01,0.4526835250854492
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,4,1,balanced,2.059784393310547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,4,1,power_law_1.01,0.4517055892944336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,4,1,power_law_1.01,0.4539097595214844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,4,1,power_law_1.01,0.47597183227539064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,4,1,power_law_1.01,0.5069625473022461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,4,1,power_law_1.01,0.5471161651611328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,4,1,power_law_1.01,0.6159385681152344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,4,1,power_law_1.01,0.7188224029541016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,4,1,power_law_1.01,0.8887558746337891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,4,1,power_law_1.01,1.0561939239501954
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,4,1,power_law_1.01,0.049203200340270994
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,4,1,power_law_1.01,1.3997747802734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,4,1,power_law_1.01,0.06637119770050048
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,4,1,power_law_1.01,1.7752108764648438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,4,1,power_law_1.01,0.07999872207641602
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,4,1,power_law_1.01,0.10429887771606446
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,4,1,power_law_1.01,0.13132160186767577
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,4,1,power_law_1.01,7.9345288085937495
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,4,1,power_law_1.01,0.19417919158935545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,4,1,power_law_1.01,2.4844761657714844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,4,1,power_law_1.01,0.21198848724365232
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,4,1,power_law_1.01,0.20503295898437499
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,4,1,power_law_1.01,0.21984191894531252
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,4,1,power_law_1.01,0.2241036796569824
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,4,1,power_law_1.01,0.23048320770263672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,4,1,power_law_1.01,0.23721343994140626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,4,1,power_law_1.01,0.2418003273010254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,4,1,power_law_1.01,3.224564514160156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,4,1,power_law_1.01,0.2572083282470703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,4,1,power_law_1.01,0.2648800086975098
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,4,1,power_law_1.01,0.2717984008789062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,4,1,power_law_1.01,0.29692480087280276
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,4,1,power_law_1.01,0.34851070404052736
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,4,1,power_law_1.01,0.39258880615234376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,4,1,power_law_1.01,0.49801025390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,4,1,power_law_1.01,0.6168441772460938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,4,1,power_law_1.2,0.08092351913452148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,4,1,power_law_1.2,0.10233535766601562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,4,1,power_law_1.2,0.15158528327941895
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,4,1,power_law_1.01,1.078751983642578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,4,1,power_law_1.2,0.24401344299316405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,4,1,power_law_1.2,0.33161857604980466
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,4,1,power_law_1.01,1.5723878479003905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,4,1,power_law_1.2,0.5069631958007813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,4,1,power_law_1.01,0.8370925140380858
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,4,1,power_law_1.2,0.5328102493286133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,4,1,power_law_1.2,0.5783596801757812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,4,1,power_law_1.2,0.6113017654418945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,4,1,power_law_1.2,0.6245580673217773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,4,1,power_law_1.2,0.6651347351074219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,4,1,power_law_1.2,0.6993004608154296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,4,1,power_law_1.2,0.7481107330322265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,4,1,power_law_1.2,0.7661721801757813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,4,1,power_law_1.2,0.7911373138427734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,4,1,power_law_1.2,0.8244147491455077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,4,1,power_law_1.2,0.8189183807373046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,4,1,power_law_1.2,0.9471437072753905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,4,1,power_law_1.2,1.0502496337890626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,4,1,power_law_1.2,1.2640255737304689
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,4,1,power_law_1.2,1.4987513732910156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,4,1,power_law_1.2,1.96214599609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,4,1,power_law_1.2,2.4664781188964846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,4,1,power_law_1.2,0.057866239547729494
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,4,1,power_law_1.2,0.07586112022399902
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,4,1,power_law_1.2,0.11840959548950196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,4,1,power_law_1.2,3.344849243164062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,4,1,power_law_1.2,0.150131196975708
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,4,1,power_law_1.2,0.22142784118652345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,4,1,power_law_1.2,0.32682880401611325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,4,1,power_law_1.2,0.3439980697631836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,4,1,power_law_1.2,0.35867134094238284
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,4,1,power_law_1.2,0.36927486419677735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,4,1,power_law_1.2,4.220751953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,4,1,power_law_1.2,0.3781184005737305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,4,1,power_law_1.2,0.3896409606933594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,4,1,power_law_1.2,0.40177791595458984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,4,1,power_law_1.2,0.45161216735839843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,4,1,power_law_1.2,0.45852672576904296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,4,1,power_law_1.2,0.46492862701416016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,4,1,power_law_1.2,0.4733369445800781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,4,1,power_law_1.2,0.489607048034668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,4,1,power_law_1.2,0.5198022460937499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,4,1,power_law_1.2,0.5560569763183594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,4,1,power_law_1.2,0.6357427215576171
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,4,1,power_law_1.2,0.7196403503417969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,4,1,power_law_1.2,0.8982150268554687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,4,1,power_law_1.2,1.0706610870361328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,4,1,power_law_1.2,1.4351820373535156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,4,1,power_law_1.2,6.170142822265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,4,1,power_law_1.2,1.7912237548828123
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,4,1,power_law_1.2,0.04904895782470703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,4,1,power_law_1.2,0.06407104015350343
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,4,1,power_law_1.2,0.07844480037689208
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,4,1,power_law_1.2,0.09858816146850587
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,4,1,power_law_1.2,0.12369728088378906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,4,1,power_law_1.2,0.19269311904907227
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,4,1,power_law_1.2,0.2035136032104492
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,4,1,power_law_1.2,0.2162227249145508
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,4,1,power_law_1.2,0.2230841636657715
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,4,1,power_law_1.2,0.2284409523010254
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,4,1,power_law_1.01,2.066366729736328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,4,1,power_law_1.2,0.24015104293823245
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,4,1,power_law_1.2,0.24510143280029295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,4,1,power_law_1.2,2.4988946533203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,4,1,power_law_1.2,0.25571264266967775
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,4,1,power_law_1.2,0.2668025588989258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,4,1,power_law_1.01,3.0899014282226562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,4,1,power_law_1.2,0.29978303909301757
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,4,1,power_law_1.01,4.162957458496093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,4,1,power_law_1.2,7.990801391601562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,4,1,power_law_1.2,3.1660980224609374
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,4,1,power_law_1.2,0.283874568939209
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,4,1,power_law_1.2,0.32982975006103515
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,4,1,power_law_1.2,0.46204673767089843
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,4,1,power_law_1.2,0.39625728607177735
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,4,1,power_law_1.2,0.6208531188964843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,4,balanced,0.07998208045959473
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,4,balanced,0.09847807884216309
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,4,balanced,0.14990976333618164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,4,balanced,0.30390527725219724
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,4,1,power_law_1.2,0.7898777770996094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,4,balanced,0.5250534439086915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,4,balanced,0.9478502655029297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,4,balanced,1.41091064453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,4,balanced,1.4085658264160155
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,4,1,power_law_1.2,1.1498438262939454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,4,balanced,1.4227635192871095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,4,balanced,1.429354248046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,4,balanced,1.5481036376953123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,4,balanced,1.5804658508300782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,4,balanced,1.5828147888183595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,4,balanced,1.5895794677734376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,4,balanced,1.6052313232421873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,4,balanced,1.6267776489257812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,4,balanced,1.6561882019042968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,4,balanced,1.7050009155273436
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,4,1,power_law_1.2,1.5366905212402344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,4,balanced,1.7501158142089843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,4,balanced,1.844969024658203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,4,balanced,1.9513381958007812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,4,balanced,2.1344627380371093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,4,balanced,3.1324057006835937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,4,balanced,3.4992959594726565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,4,balanced,0.04825215816497803
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,4,balanced,5.036491394042969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,4,balanced,0.06320511817932128
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,4,balanced,0.08878591537475586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,4,balanced,0.16069887161254884
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,4,balanced,0.28371328353881836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,4,balanced,0.4995571136474609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,4,balanced,7.129986572265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,4,balanced,0.7136576080322266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,4,balanced,0.7212671661376954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,4,balanced,0.7253670501708984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,4,balanced,0.7254732513427735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,4,balanced,0.7674457550048828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,4,balanced,0.7818918609619141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,4,balanced,0.7774374389648437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,4,balanced,0.7743846130371094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,4,balanced,0.8056333160400391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,4,balanced,0.8082918548583985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,4,balanced,0.8288985443115233
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,4,balanced,0.866030044555664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,4,balanced,0.8700697326660156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,4,balanced,0.9479872131347656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,4,balanced,11.237200927734374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,4,balanced,1.0170137786865234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,4,balanced,1.091320343017578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,4,balanced,1.2332825469970703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,4,balanced,1.4926988220214843
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,4,1,power_law_1.2,2.35619384765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,4,balanced,2.038146514892578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,4,balanced,2.4628005981445313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,4,1,power_law_1.2,3.191958312988281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,4,balanced,3.3524148559570315
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,4,1,power_law_1.2,4.9045458984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,4,1,power_law_1.2,6.53593994140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,1,4,balanced,0.03504256010055542
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,1,4,balanced,0.04755839824676514
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,1,4,balanced,0.06236159801483154
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,1,4,balanced,0.4263462448120118
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,1,4,balanced,0.42856193542480464
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,1,4,balanced,0.42832382202148434
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,1,4,balanced,0.4307686233520508
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,1,4,balanced,0.4322009658813476
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,1,4,balanced,0.15765631675720215
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,1,4,balanced,0.43058433532714846
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,1,4,balanced,0.43581695556640626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,1,4,balanced,0.09530752182006837
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,1,4,balanced,0.4369689559936523
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,1,4,balanced,0.4363225555419922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,1,4,balanced,0.44028800964355475
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,1,4,balanced,0.4491750335693359
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,1,4,balanced,0.45529087066650387
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,1,4,balanced,0.47356414794921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,1,4,balanced,0.48358783721923826
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,1,4,balanced,0.5059123229980469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,1,4,balanced,0.5526361465454102
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,1,4,balanced,0.2923904037475586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,4,power_law_1.01,0.146494722366333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,4,power_law_1.01,0.2947750473022461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,4,power_law_1.01,0.5203558349609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,4,power_law_1.01,0.548661766052246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,4,power_law_1.01,0.7265216064453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,4,power_law_1.01,0.9249356842041015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,4,power_law_1.01,1.0265792083740233
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,4,power_law_1.01,1.202676467895508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,4,power_law_1.01,1.1656409454345702
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,4,power_law_1.01,1.2028876495361327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,4,power_law_1.01,1.3689689636230469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,4,power_law_1.01,1.4057586669921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,4,power_law_1.01,1.2403187561035156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,4,power_law_1.01,1.2884658813476562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,4,power_law_1.01,1.3152255249023437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,4,power_law_1.01,1.3151449584960937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,1,4,balanced,0.42696193695068363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,4,power_law_1.01,1.384085693359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,4,power_law_1.01,1.5302566528320312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,4,power_law_1.01,1.7445465087890626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,4,power_law_1.01,1.9122470092773436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,4,power_law_1.01,2.288843536376953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,4,power_law_1.01,2.976856384277344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,4,power_law_1.01,3.3841522216796873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,4,power_law_1.01,4.544515991210938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,4,power_law_1.01,0.08745344161987305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,4,power_law_1.01,0.147706880569458
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,4,power_law_1.01,0.2637055969238281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,4,power_law_1.01,5.672894897460937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,4,power_law_1.01,0.26462080001831056
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,4,power_law_1.01,0.3923904037475586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,4,power_law_1.01,0.4671244812011719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,4,power_law_1.01,0.4845235061645508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,4,power_law_1.01,0.6358451080322266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,4,power_law_1.01,0.6648025512695312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,1,4,balanced,0.4278079986572266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,4,power_law_1.01,0.6072819137573242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,4,power_law_1.01,0.5620211029052735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,4,power_law_1.01,8.024442749023438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,4,power_law_1.01,0.5967424011230469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,4,power_law_1.01,0.6050559997558593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,4,power_law_1.01,0.6290035247802734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,4,power_law_1.01,0.6744742584228516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,4,power_law_1.01,0.6334220886230468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,4,power_law_1.01,0.6920345306396485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,4,power_law_1.01,0.7376140594482422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,4,power_law_1.01,0.7892479705810547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,4,power_law_1.01,0.8618994903564452
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,4,power_law_1.01,0.961902084350586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,4,power_law_1.01,10.816958007812499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,4,power_law_1.01,1.1614681243896485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,4,power_law_1.01,1.3489613342285156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,1,4,power_law_1.01,0.0689024019241333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,4,power_law_1.01,1.7286758422851562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,1,4,power_law_1.01,0.1025049591064453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,1,4,power_law_1.01,0.16761856079101561
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,4,power_law_1.01,2.1446258544921877
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,1,4,power_law_1.01,0.16377216339111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,1,4,power_law_1.01,0.21460479736328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,1,4,power_law_1.01,0.277894401550293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,1,4,power_law_1.01,0.2335103988647461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,4,power_law_1.01,2.9716070556640624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,1,4,power_law_1.01,0.3304537582397461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,1,4,power_law_1.01,0.3371993637084961
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,1,4,power_law_1.01,0.3199488067626953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,1,4,power_law_1.01,0.3082880020141602
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,1,4,power_law_1.01,0.3326950454711914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,1,4,power_law_1.01,0.347457275390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,4,power_law_1.01,3.666988830566406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,1,4,power_law_1.01,0.35841281890869137
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,1,4,power_law_1.01,0.37085567474365233
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,1,4,power_law_1.01,0.4048179244995117
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,1,4,power_law_1.01,0.4093363189697266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,1,4,power_law_1.01,0.40560768127441404
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,1,4,balanced,0.7901068878173828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,1,4,balanced,1.2645299530029297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,1,4,balanced,0.9258201599121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,1,4,power_law_1.01,0.43345790863037104
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,1,4,power_law_1.01,0.4875571060180664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,4,power_law_1.2,0.1437286376953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,1,4,power_law_1.01,0.6565081787109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,1,4,power_law_1.01,0.7516684722900391
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,1,4,power_law_1.01,0.5235494232177734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,4,power_law_1.2,0.29748863220214844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,4,power_law_1.2,0.3970751953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,4,power_law_1.2,0.5197427368164063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,4,power_law_1.2,0.6933542633056641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,4,power_law_1.2,0.8841343688964844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,1,4,power_law_1.01,0.9795289611816406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,1,4,power_law_1.01,1.1414157104492186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,4,power_law_1.2,0.9726579284667969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,4,power_law_1.2,1.2416844940185547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,4,power_law_1.2,1.1299699401855468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,4,power_law_1.2,1.168133087158203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,1,4,power_law_1.01,1.7294758605957032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,4,power_law_1.2,1.1176255798339843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,4,power_law_1.2,1.2461427307128905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,4,power_law_1.2,1.287704315185547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,4,power_law_1.2,1.2428428649902343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,4,power_law_1.2,1.2781350708007813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,4,power_law_1.2,1.4131852722167968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,4,power_law_1.2,1.4408499145507814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,4,power_law_1.2,1.5367884826660156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,4,power_law_1.2,1.7438450622558594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,4,power_law_1.2,2.0753292846679687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,4,power_law_1.2,2.511840057373047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,4,power_law_1.2,0.08956031799316407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,4,power_law_1.2,3.09362060546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,4,power_law_1.2,0.14920191764831542
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,4,power_law_1.2,0.22458879470825197
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,4,power_law_1.2,3.6379467773437497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,4,power_law_1.2,0.26386816024780274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,4,power_law_1.2,0.35359615325927735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,4,power_law_1.2,0.4636966323852539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,4,power_law_1.2,0.4665625762939453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,4,power_law_1.2,0.6555916595458984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,4,power_law_1.2,4.971247253417969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,4,power_law_1.2,0.6598886108398437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,4,power_law_1.2,0.649644775390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,4,power_law_1.2,0.5829862213134765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,4,power_law_1.2,0.6764915466308594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,4,power_law_1.2,0.6343552017211914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,4,power_law_1.2,0.6238438415527343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,4,power_law_1.2,0.6130227279663086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,4,power_law_1.2,6.443150024414062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,4,power_law_1.2,0.6911116790771484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,4,power_law_1.2,0.690540771484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,4,power_law_1.2,0.7578931427001953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,4,power_law_1.2,0.8239360046386718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,4,power_law_1.2,0.9034982299804687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,4,power_law_1.2,1.018888931274414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,4,power_law_1.2,1.2220722961425783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,4,power_law_1.2,8.98020751953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,1,4,power_law_1.2,0.06992512226104737
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,4,power_law_1.2,1.8544076538085938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,1,4,power_law_1.2,0.10264960289001465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,4,power_law_1.2,2.1927142333984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,1,4,power_law_1.2,0.14661888122558592
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,1,4,power_law_1.2,0.16062719345092774
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,1,4,power_law_1.2,0.2061644744873047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,1,4,power_law_1.2,0.2407923126220703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,4,power_law_1.2,3.193663940429688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,4,power_law_1.2,1.4402125549316405
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,1,4,power_law_1.2,0.33430400848388675
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,1,4,power_law_1.2,0.2959833526611328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,4,power_law_1.2,11.863404541015624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,1,4,power_law_1.2,0.38584449768066403
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,1,4,power_law_1.2,0.31923967361450195
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,1,4,power_law_1.2,0.3217753601074219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,1,4,power_law_1.2,0.2453363227844238
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,4,power_law_1.2,3.9681805419921874
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,1,4,power_law_1.2,0.3470374298095703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,1,4,power_law_1.2,0.35258113861083984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,1,4,power_law_1.2,0.36001407623291015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,1,4,power_law_1.2,0.4028364944458008
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,1,4,power_law_1.2,0.4189247894287109
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,1,4,power_law_1.2,0.432416000366211
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,1,4,power_law_1.01,2.1776089477539062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,1,4,power_law_1.2,0.46706176757812495
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,1,4,power_law_1.2,0.5311923217773438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,1,4,power_law_1.2,0.6083878326416016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,1,4,power_law_1.2,0.8955430603027343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,2,balanced,0.07590784072875977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,2,balanced,0.09882752418518066
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,1,4,power_law_1.2,0.7525094604492187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,2,balanced,0.14690176010131836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,2,balanced,0.298604793548584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,1,4,power_law_1.2,1.2783385467529296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,2,balanced,0.5235443115234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,2,balanced,0.9547277069091796
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,1,4,power_law_1.2,1.5661184692382812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,2,balanced,1.4060978698730469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,2,balanced,1.41114501953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,2,balanced,1.420668182373047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,2,balanced,1.4393869018554688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,2,balanced,1.4480589294433595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,2,balanced,1.4290982055664063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,2,balanced,1.5702310180664063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,2,balanced,1.5856306457519531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,2,balanced,1.5877708435058593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,2,balanced,1.425470733642578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,2,balanced,1.6160986328124998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,2,balanced,1.6784538269042968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,2,balanced,1.7166297912597657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,2,balanced,1.8309324645996092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,2,balanced,1.9337266540527345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,2,balanced,2.0936985778808594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,2,balanced,0.04488959789276123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,2,balanced,0.061674242019653325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,2,balanced,0.09063039779663086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,2,balanced,3.065871276855469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,2,balanced,0.1571737575531006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,2,balanced,3.470303955078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,2,balanced,0.2977177619934082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,2,balanced,0.518682861328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,2,balanced,0.7475161743164063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,2,balanced,4.656991882324219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,2,balanced,0.7465203094482422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,2,balanced,0.755440673828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,2,balanced,0.7495027160644531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,2,balanced,0.7698662567138672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,2,balanced,0.762798080444336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,2,balanced,0.7784909057617188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,2,balanced,0.8329344177246094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,2,balanced,6.215916748046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,2,balanced,0.848848648071289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,2,balanced,0.8708096313476563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,2,balanced,0.8832371520996094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,1,4,power_law_1.2,2.149427185058594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,2,balanced,0.891695327758789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,2,balanced,0.9554815673828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,2,balanced,0.9625933074951172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,2,balanced,1.0375552368164063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,2,balanced,1.1367743682861327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,2,balanced,1.214013442993164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,2,balanced,1.416441650390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,2,balanced,2.090498504638672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,2,balanced,2.4146546936035156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,2,balanced,9.781160888671874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,2,balanced,3.363978271484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,1,4,power_law_1.2,3.5981848144531248
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,2,2,balanced,0.05182591915130615
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,2,2,balanced,0.05956736087799073
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,2,2,balanced,0.45966335296630856
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,2,2,balanced,0.07498752117156983
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,2,2,balanced,0.4606284713745117
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,2,2,balanced,0.4628224182128906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,2,2,balanced,0.46275070190429685
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,2,2,balanced,0.10614144325256347
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,2,2,balanced,0.46480384826660154
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,2,2,balanced,0.46695167541503907
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,2,2,balanced,0.17322111129760742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,2,2,balanced,0.4665958404541016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,2,2,balanced,0.4711500930786133
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,2,2,balanced,0.4725811386108399
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,2,2,balanced,0.4827840042114258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,2,2,balanced,0.48842624664306644
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,2,2,balanced,0.5068492889404297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,2,2,balanced,0.3144588851928711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,2,2,balanced,0.5191296005249024
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,2,2,balanced,0.5567014312744141
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,2,2,balanced,0.5869311904907226
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,2,2,balanced,0.45971839904785156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,2,2,balanced,0.65372802734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,2,power_law_1.01,0.09831680297851562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,2,power_law_1.01,0.14131967544555663
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,2,power_law_1.01,0.2938252830505371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,2,power_law_1.01,0.36622081756591796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,2,power_law_1.01,0.5581913757324218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,2,power_law_1.01,0.7453874969482422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,2,power_law_1.01,0.859230728149414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,2,power_law_1.01,1.0516556549072267
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,2,power_law_1.01,1.0888832092285157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,2,power_law_1.01,1.1343615722656248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,2,power_law_1.01,1.1600717163085936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,2,power_law_1.01,1.2231462097167969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,2,power_law_1.01,1.1970944213867187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,2,power_law_1.01,1.3049317932128905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,2,power_law_1.01,1.3526361083984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,2,power_law_1.01,1.3797695922851563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,2,power_law_1.01,1.4453375244140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,2,power_law_1.01,1.5075570678710937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,2,power_law_1.01,1.6656063842773439
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,2,power_law_1.01,1.8804415893554687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,2,power_law_1.01,2.172419891357422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,2,power_law_1.01,2.6287155151367188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,2,power_law_1.01,3.1155841064453123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,2,power_law_1.01,3.9842980957031253
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,2,power_law_1.01,0.06836863994598388
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,2,power_law_1.01,0.09036031723022461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,2,power_law_1.01,0.15117568016052246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,2,power_law_1.01,4.981842041015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,2,power_law_1.01,0.19969663619995118
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,2,power_law_1.01,0.3241881561279297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,2,power_law_1.01,0.4064179229736328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,2,power_law_1.01,0.4529600143432617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,2,power_law_1.01,0.5661363220214843
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,2,2,balanced,0.4606809616088867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,2,power_law_1.01,0.565797119140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,2,power_law_1.01,6.9611328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,2,power_law_1.01,0.5831552124023437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,2,power_law_1.01,0.6008422470092774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,2,power_law_1.01,0.6154969787597656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,2,power_law_1.01,0.6434803009033203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,2,power_law_1.01,0.6863833618164062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,2,power_law_1.01,0.6943373107910157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,2,power_law_1.01,0.718097915649414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,2,power_law_1.01,0.7265280151367188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,2,power_law_1.01,0.7654029083251953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,2,power_law_1.01,0.831379165649414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,2,power_law_1.01,0.9298841857910156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,2,power_law_1.01,9.063435668945313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,2,power_law_1.01,0.9807500457763672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,2,power_law_1.01,1.1521625518798828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,2,power_law_1.01,1.335983428955078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,2,2,power_law_1.01,0.06803328037261963
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,2,power_law_1.01,1.737885437011719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,2,2,power_law_1.01,0.08461695671081543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,2,2,power_law_1.01,0.12282624244689941
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,2,power_law_1.01,2.078045501708984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,2,2,power_law_1.01,0.13369215965270995
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,2,2,power_law_1.01,0.1894041633605957
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,2,2,power_law_1.01,0.2390067291259766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,2,2,power_law_1.01,0.278210563659668
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,2,2,power_law_1.01,0.34782081604003906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,2,power_law_1.01,2.854912109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,2,2,power_law_1.01,0.3417536163330078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,2,2,power_law_1.01,0.3494374465942383
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,2,2,power_law_1.01,0.3627046585083008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,2,power_law_1.01,3.513852233886719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,2,2,power_law_1.01,0.3686975860595703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,2,2,power_law_1.01,0.4022809600830078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,2,2,power_law_1.01,0.41118335723876953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,2,2,power_law_1.01,0.41948928833007815
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,2,2,power_law_1.01,0.4133414459228516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,2,2,power_law_1.01,0.4215564727783203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,2,2,power_law_1.01,0.46374271392822264
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,2,2,balanced,0.9678873443603516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,2,2,balanced,1.5349836730957032
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,2,2,balanced,1.1237414550781248
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,2,2,power_law_1.01,0.49262847900390627
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,2,2,power_law_1.01,0.6169830322265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,2,2,power_law_1.01,0.7538086700439453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,2,2,power_law_1.01,0.5596505737304687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,2,power_law_1.2,0.10214400291442871
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,2,2,power_law_1.01,0.9306253051757812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,2,power_law_1.2,0.14117888450622557
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,2,power_law_1.2,0.1993484878540039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,2,power_law_1.2,0.3683737564086914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,2,power_law_1.2,0.5131673431396484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,2,power_law_1.2,0.7070207977294922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,2,power_law_1.2,0.8168627166748047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,2,power_law_1.2,1.0344844818115235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,2,power_law_1.2,1.084236831665039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,2,power_law_1.2,1.1115122985839843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,2,power_law_1.2,1.1766835021972657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,2,2,power_law_1.01,1.1645734405517578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,2,power_law_1.2,1.1962035369873045
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,2,2,power_law_1.01,1.4679487609863282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,2,power_law_1.2,1.280775604248047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,2,power_law_1.2,1.3493133544921876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,2,power_law_1.2,1.3603660583496093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,2,power_law_1.2,1.4624742126464843
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,2,2,power_law_1.01,2.1240538024902342
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,2,power_law_1.2,1.562329559326172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,2,power_law_1.2,1.7438067626953124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,2,power_law_1.2,1.950836486816406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,2,power_law_1.2,1.2015872192382813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,2,power_law_1.2,2.1900965881347654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,2,power_law_1.2,2.7163739013671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,2,power_law_1.2,0.06782976150512696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,2,power_law_1.2,0.09048192024230957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,2,power_law_1.2,3.2320602416992186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,2,power_law_1.2,0.13739775657653808
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,2,power_law_1.2,0.20968448638916018
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,2,power_law_1.2,0.37313022613525393
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,2,power_law_1.2,0.29846912384033203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,2,power_law_1.2,4.1041894531250005
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,2,power_law_1.2,0.4297792053222656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,2,power_law_1.2,0.5673523330688477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,2,power_law_1.2,0.5521484756469727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,2,power_law_1.2,0.6078796768188477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,2,power_law_1.2,5.189752197265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,2,power_law_1.2,0.5618431854248047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,2,power_law_1.2,0.5749708938598632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,2,power_law_1.2,0.6095436859130859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,2,power_law_1.2,0.6684646606445312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,2,power_law_1.2,0.7108787536621094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,2,power_law_1.2,0.7185984039306641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,2,power_law_1.2,0.7451277160644532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,2,power_law_1.2,0.779178237915039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,2,power_law_1.2,0.8316812896728516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,2,power_law_1.2,7.249346313476562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,2,power_law_1.2,0.9502528381347656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,2,power_law_1.2,1.0436300659179687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,2,power_law_1.2,1.2542015838623048
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,2,2,power_law_1.2,0.07286015987396241
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,2,power_law_1.2,1.4152627563476563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,2,2,power_law_1.2,0.08408320426940918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,2,2,power_law_1.2,0.10880895614624024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,2,power_law_1.2,1.8010456848144532
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,2,2,power_law_1.2,0.13323904037475587
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,2,2,power_law_1.2,0.18204032897949218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,2,power_law_1.2,2.1598118591308593
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,2,2,power_law_1.2,0.2196518325805664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,2,2,power_law_1.2,0.26102399826049805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,2,2,power_law_1.2,0.3376396942138672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,2,2,power_law_1.2,0.3350566482543945
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,2,2,power_law_1.2,0.3479667282104492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,2,power_law_1.2,3.034892883300781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,2,2,power_law_1.2,0.37640575408935545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,2,power_law_1.2,9.683887329101562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,2,2,power_law_1.2,0.39939456939697265
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,2,2,power_law_1.2,0.41291007995605467
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,2,2,power_law_1.2,0.40140670776367193
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,2,power_law_1.2,3.7008102416992186
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,2,2,power_law_1.2,0.41203968048095707
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,2,2,power_law_1.2,0.45013889312744143
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,2,2,power_law_1.2,0.36724224090576174
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,2,2,power_law_1.01,2.6619366455078124
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,2,2,power_law_1.2,0.49686016082763673
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,2,2,power_law_1.2,0.5369152069091797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,2,2,power_law_1.2,0.6362188720703126
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,2,2,power_law_1.2,0.7160332489013672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,4,1,balanced,0.08707455635070802
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,2,2,power_law_1.2,0.9771699523925781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,1,balanced,0.10605952262878418
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,2,2,power_law_1.2,1.1954431915283203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,1,balanced,0.14969599723815918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,1,balanced,0.3085312080383301
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,2,2,power_law_1.2,1.641518096923828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,1,balanced,0.5377260971069335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,1,balanced,0.9899468994140624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,1,balanced,1.4619815063476562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,1,balanced,1.4663711547851563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,1,balanced,1.4479788208007813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,1,balanced,1.4603015136718749
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,1,balanced,1.4855853271484374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,1,balanced,1.4986758422851563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,1,balanced,1.5153843688964845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,1,balanced,1.5055757141113282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,1,balanced,1.5445933532714844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,1,balanced,1.6040147399902345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,1,balanced,1.688709716796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,1,balanced,1.7337388610839846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,1,balanced,1.7704908752441404
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,1,balanced,1.8518194580078124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,1,balanced,1.9536230468749998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,1,balanced,2.1621881103515626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,1,balanced,3.124353332519531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,1,balanced,3.387115478515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,1,balanced,0.06462912082672119
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,1,balanced,0.08159104347229004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,1,balanced,0.11038016319274901
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,1,balanced,0.1825817680358887
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,2,2,power_law_1.2,2.1579942321777343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,1,balanced,0.32455230712890626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,1,balanced,0.5967705535888672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,1,balanced,4.611379699707031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,1,balanced,0.8773490905761718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,1,balanced,0.8801900482177734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,1,balanced,0.8840614318847656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,1,balanced,0.8873023986816406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,1,balanced,0.889246063232422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,1,balanced,0.8947481536865235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,1,balanced,0.9045037078857423
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,1,balanced,0.9081356811523438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,1,balanced,6.24302490234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,1,balanced,0.9135788726806642
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,1,balanced,0.9256601715087891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,1,balanced,1.024981155395508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,1,balanced,1.0654124450683593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,1,balanced,1.0816844940185546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,1,balanced,1.1417273712158202
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,1,balanced,1.1751398468017578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,1,balanced,1.2561074829101562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,1,balanced,1.369462432861328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,1,balanced,1.5775456237792969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,1,balanced,9.743963012695312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,1,balanced,2.2899987792968752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,1,balanced,2.673258972167969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,1,balanced,3.701262817382813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,2,2,power_law_1.2,3.2117681884765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,2,2,power_law_1.2,4.00693115234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,4,1,balanced,0.07803199768066407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,4,1,balanced,0.07812352180480957
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,4,1,balanced,0.08152192115783692
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,4,1,balanced,0.1280940818786621
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,4,1,balanced,0.07884031772613526
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,4,1,balanced,0.13615488052368163
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,4,1,balanced,0.1621459197998047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,4,1,balanced,0.08696191787719727
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,4,1,balanced,0.09801792144775391
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,4,1,balanced,0.10897472381591797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,4,1,balanced,0.12062335968017579
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,4,1,balanced,0.19730047225952146
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,4,1,balanced,0.21964160919189454
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,4,1,balanced,0.26483903884887694
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,4,1,balanced,0.32162750244140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,4,1,balanced,0.3652864074707031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,4,1,balanced,0.46113216400146484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,4,1,balanced,0.653070068359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,4,1,balanced,0.8508914947509766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,4,1,balanced,1.247744674682617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,4,1,power_law_1.01,0.08807807922363281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,1,power_law_1.01,0.10735744476318358
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,4,1,balanced,1.630043487548828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,1,power_law_1.01,0.1510470390319824
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,4,1,balanced,2.413277435302734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,1,power_law_1.01,0.3077017593383789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,1,power_law_1.01,0.4953836822509766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,1,power_law_1.01,0.7718118286132812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,1,power_law_1.01,0.6584518432617188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,4,1,balanced,3.2057791137695313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,1,power_law_1.01,0.9877120208740233
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,1,power_law_1.01,1.024238052368164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,1,power_law_1.01,1.0800819396972656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,1,power_law_1.01,1.1011609649658203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,1,power_law_1.01,1.1426566314697264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,1,power_law_1.01,1.1985753631591796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,1,power_law_1.01,1.30596923828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,4,1,balanced,4.743681335449219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,1,power_law_1.01,1.3683123779296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,1,power_law_1.01,1.5107334899902345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,1,power_law_1.01,1.4182266235351562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,1,power_law_1.01,1.5262783813476564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,1,power_law_1.01,1.6066578674316407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,1,power_law_1.01,1.8864869689941408
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,1,power_law_1.01,2.111516876220703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,1,power_law_1.01,2.569765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,1,power_law_1.01,3.0436651611328123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,1,power_law_1.01,0.06374015808105468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,1,power_law_1.01,0.07928192138671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,1,power_law_1.01,3.895692138671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,1,power_law_1.01,0.10969792366027833
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,1,power_law_1.01,0.18280128479003904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,1,power_law_1.01,0.3104038429260254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,1,power_law_1.01,4.714085083007812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,1,power_law_1.01,0.4393056106567383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,1,power_law_1.01,0.527136001586914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,1,power_law_1.01,0.6357126235961914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,1,power_law_1.01,0.6608966064453126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,1,power_law_1.01,0.6852941131591797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,1,power_law_1.01,0.7064678192138671
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,1,power_law_1.01,6.558340454101563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,1,power_law_1.01,0.7381938934326172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,1,power_law_1.01,0.7131168365478515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,1,power_law_1.01,0.7608589172363281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,1,power_law_1.01,0.7665337371826172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,1,power_law_1.01,8.344578857421876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,1,power_law_1.01,0.7937689971923828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,1,power_law_1.01,0.8765222167968749
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,1,power_law_1.01,0.9097785949707031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,1,power_law_1.01,0.9593267059326172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,1,power_law_1.01,1.0414176177978516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,1,power_law_1.01,1.1301964569091798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,1,power_law_1.01,1.3080044555664063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,1,power_law_1.01,1.4898521423339843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,1,power_law_1.01,1.88406982421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,1,power_law_1.01,2.2292787170410158
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,4,1,power_law_1.01,0.07821375846862792
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,1,power_law_1.01,2.978818054199219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,4,1,power_law_1.01,0.07915455818176269
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,4,1,power_law_1.01,0.07806911945343017
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,4,1,power_law_1.01,0.0817356777191162
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,4,1,power_law_1.01,0.08495488166809081
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,1,power_law_1.01,3.7090988159179688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,4,1,power_law_1.01,0.09865216255187989
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,4,1,power_law_1.01,0.1083833599090576
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,4,1,power_law_1.01,0.11992128372192383
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,4,1,power_law_1.01,0.12273664474487304
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,4,1,power_law_1.01,0.1363980770111084
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,4,1,power_law_1.01,0.1555001640319824
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,4,1,power_law_1.01,0.21922367095947265
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,4,1,power_law_1.01,0.2650412750244141
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,4,1,power_law_1.01,0.32133247375488283
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,4,1,power_law_1.01,0.1977280044555664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,4,1,power_law_1.01,0.36585216522216796
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,4,1,power_law_1.01,0.45933696746826175
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,4,1,power_law_1.01,0.650057601928711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,4,1,power_law_1.01,0.8418022155761719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,4,1,power_law_1.01,2.412234191894531
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,4,1,power_law_1.01,4.745177001953126
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,4,1,power_law_1.01,1.2516569519042968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,4,1,power_law_1.01,1.6345382690429688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,4,1,power_law_1.2,0.08643008232116699
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,1,power_law_1.2,0.10684864044189453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,1,power_law_1.2,0.3086150360107422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,1,power_law_1.2,0.429747200012207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,1,power_law_1.2,0.6100223922729493
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,4,1,power_law_1.01,3.1846270751953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,1,power_law_1.2,0.7127436828613282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,1,power_law_1.2,0.9582617950439453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,1,power_law_1.2,0.991167984008789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,1,power_law_1.2,1.0183679962158203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,1,power_law_1.2,1.1406521606445312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,1,power_law_1.2,1.1732224273681642
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,1,power_law_1.2,1.2730566406250001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,1,power_law_1.2,1.4979936218261718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,1,power_law_1.2,1.4570719909667968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,1,power_law_1.2,0.14648960113525392
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,1,power_law_1.2,1.6265005493164062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,1,power_law_1.2,1.0646284484863282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,1,power_law_1.2,1.7188461303710938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,1,power_law_1.2,1.9158047485351564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,1,power_law_1.2,2.1575482177734377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,1,power_law_1.2,1.3615199279785155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,1,power_law_1.2,2.5929855346679687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,1,power_law_1.2,3.107532653808594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,1,power_law_1.2,0.05789120197296142
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,1,power_law_1.2,0.0793939208984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,1,power_law_1.2,0.11035776138305664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,1,power_law_1.2,3.957218627929687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,1,power_law_1.2,0.1835923194885254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,1,power_law_1.2,0.4207974243164062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,1,power_law_1.2,4.880240783691407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,1,power_law_1.2,0.28138559341430663
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,1,power_law_1.2,0.4894611358642578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,1,power_law_1.2,0.6216243362426758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,1,power_law_1.2,0.6467001342773437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,1,power_law_1.2,6.566995849609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,1,power_law_1.2,0.6646572875976562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,1,power_law_1.2,0.6853094482421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,1,power_law_1.2,0.7087276458740235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,1,power_law_1.2,8.370599975585938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,1,power_law_1.2,0.723443832397461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,1,power_law_1.2,0.7503884887695312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,1,power_law_1.2,0.7664473724365235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,1,power_law_1.2,0.7876818847656251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,1,power_law_1.2,0.8783071899414063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,1,power_law_1.2,1.0730099487304687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,1,power_law_1.2,0.9322815704345704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,1,power_law_1.2,0.9812000274658204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,1,power_law_1.2,1.1802079772949219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,1,power_law_1.2,1.367538604736328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,1,power_law_1.2,1.5676730346679686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,1,power_law_1.2,1.9204447937011717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,1,power_law_1.2,2.3165798950195313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,1,power_law_1.2,3.0463040161132815
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,4,1,power_law_1.2,0.07826303958892822
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,4,1,power_law_1.2,0.07970367908477784
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,4,1,power_law_1.2,0.08687935829162598
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,1,power_law_1.2,3.796693725585938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,4,1,power_law_1.2,0.09711872100830078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,4,1,power_law_1.2,0.10813055992126466
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,4,1,power_law_1.2,0.12849727630615235
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,4,1,power_law_1.2,0.13973888397216797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,4,1,power_law_1.2,0.16048320770263672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,4,1,power_law_1.2,0.2200409507751465
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,4,1,power_law_1.2,0.19778751373291015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,4,1,power_law_1.2,0.26424575805664063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,4,1,power_law_1.2,0.3657139205932617
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,4,1,power_law_1.2,0.0786956787109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,4,1,power_law_1.2,0.08128191947937011
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,4,1,power_law_1.2,0.11756735801696779
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,4,1,power_law_1.2,0.3207718276977539
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,4,1,power_law_1.2,1.6299481201171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,4,1,power_law_1.2,0.46012542724609373
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,4,1,power_law_1.2,0.6520352172851562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,4,1,power_law_1.2,0.8511212921142578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,4,1,power_law_1.2,1.2484582519531249
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,4,1,power_law_1.2,2.410364227294922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,4,1,power_law_1.2,3.196654663085938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,4,balanced,0.06243072032928467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,4,balanced,0.09790719985961914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,4,balanced,0.08496767997741698
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,4,balanced,0.17761152267456054
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,4,balanced,0.33376766204833985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,4,balanced,0.334543342590332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,4,balanced,0.35291774749755855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,4,balanced,0.33500415802001954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,4,balanced,0.3548735809326172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,4,balanced,0.3611686325073242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,4,balanced,0.36684417724609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,4,balanced,0.3623001480102539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,4,balanced,0.3719065475463867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,4,balanced,0.39048065185546876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,4,balanced,0.39955841064453124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,4,balanced,0.4251699066162109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,4,balanced,0.44928638458251957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,4,balanced,0.5037171173095703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,4,balanced,0.554337272644043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,4,balanced,0.7800819396972656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,4,balanced,0.8803033447265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,4,balanced,1.2346112060546877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,4,balanced,1.5829158020019531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,4,balanced,0.026479361057281492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,4,balanced,0.027619841098785403
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,4,balanced,2.462024993896484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,4,balanced,0.043424000740051275
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,4,balanced,0.07293056011199951
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,4,balanced,0.10963456153869629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,4,balanced,0.18993408203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,4,balanced,0.028892159461975098
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,4,balanced,0.37900543212890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,4,balanced,0.18770559310913087
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,4,balanced,3.4125158691406248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,4,balanced,0.18939647674560547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,4,balanced,0.21277952194213867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,4,balanced,0.2116032028198242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,4,balanced,0.21134208679199218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,4,balanced,0.21077375411987304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,4,balanced,0.20976383209228516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,4,balanced,0.21224063873291016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,4,balanced,0.21484544754028317
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,4,balanced,0.21764608383178713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,4,balanced,0.2526604843139649
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,4,balanced,0.2563225555419922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,4,balanced,0.29796480178833007
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,4,balanced,0.2092390441894531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,4,balanced,0.33577598571777345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,4,balanced,0.3766400146484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,4,balanced,0.46095230102539064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,4,balanced,0.638133773803711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,4,balanced,0.7553459167480469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,4,balanced,1.0841356658935548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,4,balanced,1.4070886230468749
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,4,1,power_law_1.2,4.749926452636719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,1,4,balanced,0.035057919025421144
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,1,4,balanced,0.11316351890563965
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,1,4,balanced,0.1130406379699707
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,1,4,balanced,0.11327360153198243
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,1,4,balanced,0.11326720237731933
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,1,4,balanced,0.11355263710021972
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,1,4,balanced,0.11377535820007326
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,1,4,balanced,0.1144371223449707
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,1,4,balanced,0.11449600219726563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,1,4,balanced,0.1149401569366455
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,1,4,balanced,0.04736767768859863
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,1,4,balanced,0.11522047996520997
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,1,4,balanced,0.11514880180358886
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,1,4,balanced,0.11730560302734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,1,4,balanced,0.11880319595336915
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,1,4,balanced,0.12270719528198241
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,1,4,balanced,0.1263526439666748
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,1,4,balanced,0.13132672309875487
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,1,4,balanced,0.14021632194519043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,1,4,balanced,0.02785279989242554
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,1,4,balanced,0.14909184455871582
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,1,4,balanced,0.02664704084396362
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,1,4,balanced,0.09306879997253417
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,4,power_law_1.01,0.05883776187896729
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,4,power_law_1.01,0.09522432327270508
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,1,4,balanced,0.23268096923828124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,4,power_law_1.01,0.1522316837310791
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,1,4,balanced,0.21299968719482423
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,1,4,balanced,0.3240729522705078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,1,4,balanced,0.4121459197998047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,4,power_law_1.01,0.21384191513061523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,4,power_law_1.01,0.2332326316833496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,4,power_law_1.01,0.3140121650695801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,4,power_law_1.01,0.2691776084899902
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,4,power_law_1.01,0.32270465850830077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,4,power_law_1.01,0.3291212844848633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,4,power_law_1.01,0.3444249725341797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,4,power_law_1.01,0.34211071014404293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,4,power_law_1.01,0.3526387023925781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,4,power_law_1.01,0.1731839942932129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,4,power_law_1.01,0.264902400970459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,4,power_law_1.01,0.3949568176269531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,4,power_law_1.01,0.36672000885009765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,4,power_law_1.01,0.4450905609130859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,4,power_law_1.01,0.48981376647949215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,4,power_law_1.01,0.5775807952880859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,4,power_law_1.01,0.3701785659790039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,4,power_law_1.01,0.6748582458496093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,4,power_law_1.01,0.849323501586914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,4,power_law_1.01,1.0509798431396484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,4,power_law_1.01,0.042790398597717286
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,4,power_law_1.01,0.06997632026672364
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,4,power_law_1.01,1.4608256530761718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,4,power_law_1.01,0.08386431694030762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,4,power_law_1.01,1.8626188659667968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,4,power_law_1.01,0.10468480110168457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,4,power_law_1.01,0.12409472465515137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,4,power_law_1.01,0.1588044834136963
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,4,power_law_1.01,0.16128128051757812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,4,power_law_1.01,0.17071744918823242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,4,power_law_1.01,0.18018560409545897
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,4,power_law_1.01,2.713406677246094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,4,power_law_1.01,0.18518272399902344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,4,power_law_1.01,0.18181247711181642
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,4,power_law_1.01,0.1898828887939453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,4,power_law_1.01,0.19388799667358397
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,4,power_law_1.01,0.19682815551757812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,4,power_law_1.01,0.20029312133789062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,4,power_law_1.01,0.2006912040710449
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,4,power_law_1.01,0.23790975570678713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,4,power_law_1.01,0.2536767959594727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,4,power_law_1.01,0.3090060806274414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,4,power_law_1.01,0.330041618347168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,4,power_law_1.01,0.41611904144287115
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,4,power_law_1.01,3.495173034667969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,4,power_law_1.01,0.49693439483642576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,4,power_law_1.01,0.6559436798095704
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,1,4,power_law_1.01,0.0344870400428772
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,1,4,power_law_1.01,0.049570560455322266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,4,power_law_1.01,0.8514982604980469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,4,power_law_1.01,1.1906163024902345
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,1,4,power_law_1.01,0.052761597633361815
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,1,4,power_law_1.01,0.06641536235809327
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,1,4,power_law_1.01,0.07620992183685302
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,4,power_law_1.01,1.5397926330566407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,1,4,power_law_1.01,0.0756326389312744
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,1,4,power_law_1.01,0.09980159759521484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,1,4,power_law_1.01,0.09500288009643555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,1,4,power_law_1.01,0.0924403190612793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,1,4,power_law_1.01,0.09851648330688476
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,1,4,power_law_1.01,0.09641983985900879
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,1,4,power_law_1.01,0.10171135902404785
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,1,4,power_law_1.01,0.1087564754486084
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,1,4,power_law_1.01,0.11185279846191407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,1,4,power_law_1.01,0.1159936046600342
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,1,4,power_law_1.01,0.12477696418762207
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,1,4,power_law_1.01,0.0997977638244629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,4,power_law_1.01,0.1423475170135498
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,1,4,power_law_1.01,0.10515456199645996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,1,4,power_law_1.01,0.13577983856201173
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,1,4,power_law_1.01,0.1570995235443115
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,1,4,power_law_1.01,0.17667455673217775
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,1,4,power_law_1.01,0.2263564872741699
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,1,4,power_law_1.01,0.28405248641967773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,4,power_law_1.2,0.06043776035308838
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,4,power_law_1.2,0.0924351978302002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,4,power_law_1.2,0.11939200401306152
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,4,power_law_1.2,0.16745344161987305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,4,power_law_1.2,0.17447423934936523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,4,power_law_1.2,0.21994752883911134
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,4,power_law_1.2,0.28756223678588866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,4,power_law_1.2,0.2714534378051758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,4,power_law_1.2,0.3150668716430664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,4,power_law_1.2,0.32765697479248046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,4,power_law_1.2,0.33359489440917967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,4,power_law_1.2,0.34471553802490235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,4,power_law_1.2,0.3584896087646484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,4,power_law_1.2,0.3647078323364258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,4,power_law_1.2,0.41001728057861325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,4,power_law_1.2,0.4472473526000977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,4,power_law_1.2,0.6037516784667969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,4,power_law_1.2,0.7081318664550781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,4,power_law_1.2,0.9000179290771484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,4,power_law_1.2,1.135132141113281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,4,power_law_1.2,0.32506881713867186
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,1,4,power_law_1.01,0.37662208557128907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,4,power_law_1.2,1.5883775329589844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,4,power_law_1.2,1.8855322265624999
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,1,4,power_law_1.01,0.4891225433349609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,4,power_law_1.2,0.3874726486206055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,4,power_law_1.2,0.06948607921600342
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,1,4,power_law_1.01,0.6610111999511719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,4,power_law_1.2,0.0969331169128418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,4,power_law_1.2,2.8213491821289063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,4,power_law_1.2,0.10135295867919922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,4,power_law_1.2,0.16532480239868164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,4,power_law_1.2,0.5050406265258789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,4,power_law_1.2,0.1616281509399414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,4,power_law_1.2,0.18816511154174803
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,1,4,power_law_1.01,0.9756825256347657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,4,power_law_1.2,0.17359231948852538
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,4,power_law_1.2,3.718696899414063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,4,power_law_1.2,0.17305599212646486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,4,power_law_1.2,0.17990783691406248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,4,power_law_1.2,0.18543231964111329
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,4,power_law_1.2,0.19824384689331054
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,4,power_law_1.2,0.1960063934326172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,4,power_law_1.2,0.13873023986816407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,4,power_law_1.2,0.20353023529052736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,4,power_law_1.2,0.20466688156127927
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,4,power_law_1.2,0.23871360778808595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,4,power_law_1.2,0.2625676727294922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,4,power_law_1.2,0.31229951858520505
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,4,power_law_1.2,0.34435329437255857
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,4,power_law_1.2,0.4315481567382813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,4,power_law_1.2,0.5301862335205078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,4,power_law_1.2,0.713655014038086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,1,4,power_law_1.2,0.034623999595642094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,4,power_law_1.2,0.9093119812011718
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,1,4,power_law_1.2,0.04927360057830811
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,1,4,power_law_1.2,0.052080640792846675
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,1,4,power_law_1.2,0.06120448112487793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,1,4,power_law_1.2,0.07690368175506593
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,1,4,power_law_1.2,0.06782464027404786
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,1,4,power_law_1.2,0.09373824119567871
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,1,4,power_law_1.2,0.0906214427947998
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,1,4,power_law_1.2,0.09354240417480468
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,1,4,power_law_1.2,0.09651071548461913
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,1,4,power_law_1.2,0.10194175720214844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,1,4,power_law_1.2,0.09897343635559082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,1,4,power_law_1.2,0.10600576400756836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,1,4,power_law_1.2,0.10142335891723633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,1,4,power_law_1.2,0.11050111770629882
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,1,4,power_law_1.2,0.11639807701110841
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,1,4,power_law_1.2,0.12559616088867187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,1,4,power_law_1.2,0.14379776000976563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,4,power_law_1.2,1.255251235961914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,4,power_law_1.2,1.5548760986328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,4,power_law_1.2,0.04116608142852783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,4,power_law_1.2,0.07325439929962159
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,1,4,power_law_1.2,0.14942975997924807
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,1,4,power_law_1.2,0.18679040908813477
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,1,4,power_law_1.2,0.21466751098632814
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,1,4,power_law_1.2,0.3276224136352539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,2,2,balanced,0.02744064092636108
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,1,4,power_law_1.2,0.5510310363769532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,2,2,balanced,0.04419583797454834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,2,2,balanced,0.0583129596710205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,2,2,balanced,0.09761664390563965
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,1,4,power_law_1.2,0.3845248031616211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,2,2,balanced,0.17748352050781252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,2,2,balanced,0.32901119232177733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,2,2,balanced,0.3258943939208984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,2,2,balanced,0.3267212677001953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,2,2,balanced,0.32573566436767576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,2,2,balanced,0.3299545669555664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,2,2,balanced,0.3540633773803711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,2,2,balanced,0.35618560791015624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,2,2,balanced,0.3284288024902344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,2,2,balanced,0.3602547073364258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,2,2,balanced,0.3662220764160156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,2,2,balanced,0.37034496307373044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,2,2,balanced,0.37462913513183593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,2,2,balanced,0.3955187225341797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,2,2,balanced,0.4160512161254883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,2,2,balanced,0.46328063964843746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,2,2,balanced,0.5122150421142578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,2,2,balanced,0.7049088287353515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,2,2,balanced,0.8432486724853515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,2,2,balanced,1.1091302490234374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,2,2,balanced,0.03176320075988769
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,2,2,balanced,1.4281843566894532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,2,2,balanced,0.04428671836853028
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,2,2,balanced,0.07563007831573486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,2,2,balanced,0.12323967933654785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,2,2,balanced,0.20256511688232423
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,2,2,balanced,0.19986560821533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,2,2,balanced,2.090154266357422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,2,2,balanced,0.1995955276489258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,2,2,balanced,0.20189695358276366
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,2,2,balanced,0.20095743179321288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,2,2,balanced,0.2037708854675293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,2,2,balanced,0.22924287796020507
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,2,2,balanced,0.23026559829711912
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,2,2,balanced,2.7307315063476563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,2,2,balanced,0.2323711967468262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,2,2,balanced,0.23311744689941408
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,2,2,balanced,0.23583488464355468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,2,2,balanced,0.24360960006713867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,1,4,power_law_1.2,0.7233548736572265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,2,2,balanced,0.2668544006347656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,2,2,balanced,0.281212158203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,2,2,balanced,0.31745792388916017
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,1,4,power_law_1.2,1.1066073608398437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,2,2,balanced,0.4204352188110351
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,2,2,balanced,0.23048063278198244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,2,2,balanced,0.3679129409790039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,2,2,balanced,0.6110028839111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,2,2,balanced,0.7071244812011719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,2,2,balanced,1.316143341064453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,2,2,balanced,1.0078079986572266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,2,2,balanced,0.028051199913024905
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,1,4,power_law_1.2,1.5510374450683595
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,2,2,balanced,0.036200959682464603
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,2,2,balanced,0.038195199966430664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,2,2,balanced,0.04428287982940674
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,2,2,balanced,0.07706111907958985
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,2,2,balanced,0.05527935981750488
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,2,2,balanced,0.1235366439819336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,2,2,balanced,0.1233894443511963
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,2,2,balanced,0.12347264289855957
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,2,2,balanced,0.1239129638671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,2,2,balanced,0.12616191864013673
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,2,2,balanced,0.1269926357269287
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,2,2,balanced,0.12698111534118653
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,2,2,balanced,0.12861696243286133
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,2,2,balanced,0.13701120376586914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,2,2,balanced,0.14301952362060547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,2,2,balanced,0.15145983695983886
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,2,2,balanced,0.12413311958312989
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,2,2,balanced,0.12344703674316407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,2,2,balanced,0.16486143112182616
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,2,2,balanced,0.17733375549316405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,2,2,power_law_1.01,0.046003198623657225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,2,2,power_law_1.01,0.05893248081207275
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,2,2,power_law_1.01,0.08049023628234864
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,2,2,balanced,0.12341376304626465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,2,2,power_law_1.01,0.1260812759399414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,2,2,power_law_1.01,0.19611391067504882
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,2,2,power_law_1.01,0.25198591232299805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,2,2,power_law_1.01,0.2621785545349121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,2,2,power_law_1.01,0.26022144317626955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,2,2,power_law_1.01,0.267511043548584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,2,2,power_law_1.01,0.2766694450378418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,2,2,power_law_1.01,0.3267942428588867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,2,2,power_law_1.01,0.322371826171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,2,2,power_law_1.01,0.332542724609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,2,2,power_law_1.01,0.33787902832031247
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,2,2,power_law_1.01,0.3460812759399414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,2,2,power_law_1.01,0.37474433898925785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,2,2,power_law_1.01,0.40687744140625004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,2,2,power_law_1.01,0.43931648254394534
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,2,2,power_law_1.01,0.17006336212158205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,2,2,power_law_1.01,0.524158706665039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,2,2,power_law_1.01,0.6009049606323242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,2,2,power_law_1.01,0.7590054321289063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,2,2,power_law_1.01,0.93193603515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,2,2,power_law_1.01,1.2635533142089845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,2,2,power_law_1.01,1.7231091308593751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,2,2,power_law_1.01,2.4251788330078123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,2,2,power_law_1.01,0.03177599906921387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,2,2,power_law_1.01,0.04746240139007568
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,2,2,power_law_1.01,3.0687603759765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,2,2,power_law_1.01,0.061236481666564944
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,2,2,power_law_1.01,0.1186393642425537
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,2,2,power_law_1.01,0.08339072227478027
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,2,2,power_law_1.01,0.14018943786621094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,2,2,power_law_1.01,0.16494207382202147
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,2,2,power_law_1.01,0.16882688522338868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,2,2,power_law_1.01,0.1657676887512207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,2,2,power_law_1.01,0.16977664947509766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,2,2,power_law_1.01,0.179814395904541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,2,2,power_law_1.01,0.20027135848999023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,2,2,power_law_1.01,0.20174848556518557
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,2,2,power_law_1.01,0.2095462417602539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,2,2,power_law_1.01,0.21291648864746096
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,2,2,balanced,0.2600204849243164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,2,2,power_law_1.01,0.21138559341430666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,2,2,power_law_1.01,0.21985023498535156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,2,2,power_law_1.01,0.25895679473876954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,2,2,power_law_1.01,0.23765375137329103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,2,2,power_law_1.01,0.2925632095336914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,2,2,power_law_1.01,0.3281024169921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,2,2,balanced,0.290643196105957
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,2,2,balanced,0.40851070404052736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,2,2,power_law_1.01,0.4045939254760742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,2,2,balanced,0.5275276947021484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,2,2,power_law_1.01,0.4772825622558594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,2,2,power_law_1.01,0.6332761764526367
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,2,2,power_law_1.01,0.03803647994995117
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,2,2,power_law_1.01,0.04444928169250488
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,2,2,power_law_1.01,0.053390078544616706
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,2,2,power_law_1.01,0.062054400444030766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,2,2,power_law_1.01,0.07293439865112304
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,2,2,power_law_1.01,0.07973504066467285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,2,2,power_law_1.01,0.7959027099609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,2,2,power_law_1.01,0.09797375679016114
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,2,2,power_law_1.01,0.10536447525024414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,2,2,power_law_1.01,1.1103308868408202
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,2,2,power_law_1.01,0.10103679656982423
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,2,2,power_law_1.01,0.1057215976715088
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,2,2,power_law_1.01,1.4218931579589844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,2,2,power_law_1.01,0.10766847610473633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,2,2,power_law_1.01,0.10872832298278809
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,2,2,power_law_1.01,0.11478528022766113
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,2,2,power_law_1.01,0.12118271827697753
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,2,2,power_law_1.01,0.12568448066711427
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,2,2,power_law_1.01,0.13264896392822265
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,2,2,power_law_1.01,0.22312448501586912
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,2,2,power_law_1.01,0.11812479972839356
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,2,2,power_law_1.01,0.19128063201904297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,2,2,power_law_1.01,0.14739839553833006
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,2,2,power_law_1.01,0.16058496475219725
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,2,2,power_law_1.01,0.2770944023132324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,2,2,power_law_1.2,0.04687744140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,2,2,power_law_1.01,0.36551422119140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,2,2,power_law_1.2,0.07108992099761963
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,2,2,power_law_1.2,0.10996992111206054
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,2,2,power_law_1.2,0.15369343757629395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,2,2,power_law_1.2,0.18059520721435546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,2,2,power_law_1.2,0.2512422370910644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,2,2,power_law_1.2,0.25644031524658206
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,2,2,power_law_1.01,0.5177254486083984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,2,2,power_law_1.2,0.26503423690795896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,2,2,power_law_1.2,0.26622079849243163
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,2,2,power_law_1.2,0.3151398468017578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,2,2,power_law_1.2,0.32400638580322266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,2,2,power_law_1.2,0.27584512710571285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,2,2,power_law_1.2,0.3365887832641602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,2,2,power_law_1.2,0.3434380722045899
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,2,2,power_law_1.2,0.351596794128418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,2,2,power_law_1.2,0.3801830291748047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,2,2,power_law_1.2,0.41308158874511725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,2,2,power_law_1.2,0.4628006362915039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,2,2,power_law_1.2,0.5384678268432618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,2,2,power_law_1.2,0.060175361633300783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,2,2,power_law_1.2,0.6277171325683594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,2,2,power_law_1.2,0.7911961364746094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,2,2,power_law_1.2,0.9493939208984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,2,2,power_law_1.2,1.27272705078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,2,2,power_law_1.2,1.7211672973632812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,2,2,power_law_1.2,2.5273703002929686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,2,2,power_law_1.2,0.02981760025024414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,2,2,power_law_1.2,0.046845440864562986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,2,2,power_law_1.2,0.059782400131225585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,2,2,power_law_1.2,0.08155776023864746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,2,2,power_law_1.2,0.11104895591735839
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,2,2,power_law_1.2,0.13321727752685547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,2,2,power_law_1.2,0.163768310546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,2,2,power_law_1.2,3.094685363769531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,2,2,power_law_1.2,0.16605440139770508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,2,2,power_law_1.2,0.1738175964355469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,2,2,power_law_1.2,0.1706188774108887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,2,2,power_law_1.2,0.17612031936645506
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,2,2,power_law_1.2,0.19733760833740235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,2,2,power_law_1.2,0.20215808868408205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,2,2,power_law_1.2,0.20962047576904297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,2,2,power_law_1.2,0.2175551986694336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,2,2,power_law_1.2,0.21731071472167968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,2,2,power_law_1.2,0.2381747245788574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,2,2,power_law_1.2,0.2703590393066406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,2,2,power_law_1.2,0.20372095108032226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,2,2,power_law_1.2,0.30921855926513675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,2,2,power_law_1.2,0.34227455139160157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,2,2,power_law_1.2,0.4238118362426758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,2,2,power_law_1.2,0.49546241760253906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,2,2,power_law_1.2,0.6627597045898438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,2,2,power_law_1.01,0.9369664001464845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,2,2,power_law_1.2,0.826033935546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,2,2,power_law_1.2,0.03811840057373047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,2,2,power_law_1.2,1.1306483459472656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,2,2,power_law_1.2,0.04690688133239746
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,2,2,power_law_1.2,0.0512499189376831
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,2,2,power_law_1.01,0.6144793701171876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,2,2,power_law_1.2,0.06023039817810059
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,2,2,power_law_1.2,0.06893439769744873
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,2,2,power_law_1.2,0.07686783790588378
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,2,2,power_law_1.2,0.10033535957336426
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,2,2,power_law_1.2,0.09809023857116698
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,2,2,power_law_1.2,0.10081791877746582
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,2,2,power_law_1.2,0.1036518383026123
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,2,2,power_law_1.01,1.2284659576416015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,2,2,power_law_1.2,0.10731519699096678
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,2,2,power_law_1.2,1.4563827514648438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,2,2,power_law_1.2,0.11049344062805175
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,2,2,power_law_1.2,0.11583999633789062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,2,2,power_law_1.2,0.12060159683227538
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,2,2,power_law_1.2,0.12616319656372071
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,2,2,power_law_1.2,0.13081343650817873
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,2,2,power_law_1.2,0.1381056022644043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,2,2,power_law_1.2,0.18298240661621096
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,2,2,power_law_1.2,0.16665088653564453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,2,2,power_law_1.2,0.2331635284423828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,2,2,power_law_1.2,0.2845939254760742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,2,2,power_law_1.2,0.38120193481445314
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,2,2,power_law_1.2,0.5036428833007813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,4,1,balanced,0.05362304210662842
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,4,1,balanced,0.07024576187133788
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,4,1,balanced,0.03531519889831543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,4,1,balanced,0.10395520210266114
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,4,1,balanced,0.34496833801269533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,4,1,balanced,0.34438976287841794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,4,1,balanced,0.18788736343383788
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,4,1,balanced,0.34553089141845705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,4,1,balanced,0.3462028884887695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,4,1,balanced,0.34557758331298827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,4,1,balanced,0.34797054290771484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,4,1,balanced,0.34861633300781253
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,4,1,balanced,0.3494367980957031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,4,1,balanced,0.3533536148071289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,4,1,balanced,0.37176063537597653
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,4,1,balanced,0.3732870483398437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,4,1,balanced,0.38264190673828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,4,1,balanced,0.3995500946044922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,4,1,balanced,0.4168672180175781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,4,1,balanced,0.4571583938598633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,4,1,balanced,0.503770866394043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,4,1,balanced,0.7086112213134765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,4,1,balanced,0.771063003540039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,4,1,balanced,1.1146623992919922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,4,1,balanced,1.3888595581054688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,4,1,balanced,2.0680268859863284
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,4,1,balanced,0.03824896097183227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,4,1,balanced,0.042275199890136714
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,4,1,balanced,0.05733376026153565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,4,1,balanced,0.09117440223693848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,4,1,balanced,0.14130687713623047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,4,1,balanced,2.7109362792968748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,4,1,balanced,0.24469055175781249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,4,1,balanced,0.24927616119384766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,4,1,balanced,0.24480127334594726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,4,1,balanced,0.24826431274414062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,4,1,balanced,0.2484832000732422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,4,1,balanced,0.24920896530151365
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,4,1,balanced,0.2513612747192383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,4,1,balanced,0.25320575714111326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,4,1,balanced,0.25634111404418947
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,4,1,balanced,0.2917100715637207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,4,1,balanced,0.2942092704772949
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,4,1,balanced,0.29518463134765627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,4,1,balanced,0.3053824043273926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,4,1,balanced,0.3172908782958984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,4,1,balanced,0.3326374435424805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,4,1,balanced,0.3516774368286133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,4,1,balanced,0.3963737487792969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,2,2,power_law_1.2,0.7414720153808594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,4,1,balanced,0.4392454528808594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,2,2,power_law_1.2,0.9801010894775392
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,4,1,balanced,0.6517030334472655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,4,1,balanced,0.7402694702148438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,4,1,balanced,1.0532723236083985
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,2,2,power_law_1.2,1.6334783935546873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,4,1,balanced,1.3708044433593751
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,2,2,power_law_1.2,2.3907020568847654
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,4,1,balanced,0.05783743858337402
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,4,1,balanced,0.06694143772125244
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,4,1,balanced,0.15707455635070802
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,4,1,balanced,0.0784780788421631
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,4,1,balanced,0.15764736175537108
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,4,1,balanced,0.10351872444152832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,4,1,balanced,0.15831871986389162
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,4,1,balanced,0.15636032104492187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,4,1,balanced,0.15918399810791015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,4,1,balanced,0.1601055908203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,4,1,balanced,0.16036991119384764
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,4,1,balanced,0.1581599998474121
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,4,1,balanced,0.16251455307006837
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,4,1,balanced,0.1646463966369629
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,4,1,balanced,0.16839679718017578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,4,1,balanced,0.05870463848114014
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,4,1,balanced,0.17542272567749023
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,4,1,balanced,0.18216064453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,4,1,balanced,0.16564416885375977
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,4,1,balanced,0.1932659149169922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,4,1,balanced,0.20673280715942383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,4,1,power_law_1.01,0.0353056001663208
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,4,1,balanced,0.25804800033569336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,4,1,power_law_1.01,0.056384639739990236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,4,1,power_law_1.01,0.07022016048431397
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,4,1,power_law_1.01,0.10526592254638671
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,4,1,power_law_1.01,0.17226175308227537
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,4,1,power_law_1.01,0.1968262481689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,4,1,power_law_1.01,0.26179712295532226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,4,1,power_law_1.01,0.270450553894043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,4,1,power_law_1.01,0.2716102409362793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,4,1,power_law_1.01,0.27299968719482426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,4,1,power_law_1.01,0.2840518379211426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,4,1,power_law_1.01,0.30568704605102537
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,4,1,power_law_1.01,0.31507776260375975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,4,1,power_law_1.01,0.3362112045288086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,4,1,power_law_1.01,0.3504518508911133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,4,1,power_law_1.01,0.35513919830322266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,4,1,power_law_1.01,0.3721952056884766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,4,1,power_law_1.01,0.41637119293212893
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,4,1,power_law_1.01,0.44044864654541016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,4,1,power_law_1.01,0.5136729431152344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,4,1,power_law_1.01,0.5730483245849609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,4,1,power_law_1.01,0.7078374481201172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,4,1,power_law_1.01,0.8666770935058594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,4,1,power_law_1.01,1.1664383697509766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,4,1,power_law_1.01,1.4902291870117188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,4,1,power_law_1.01,0.03800447940826416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,4,1,power_law_1.01,0.041034879684448244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,4,1,power_law_1.01,2.1360365295410153
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,4,1,power_law_1.01,0.05829696178436279
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,4,1,power_law_1.01,0.09488320350646973
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,4,1,power_law_1.01,0.11889984130859374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,4,1,power_law_1.01,2.794569091796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,4,1,power_law_1.01,0.15632960319519043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,4,1,power_law_1.01,0.19486143112182616
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,4,1,power_law_1.01,0.20049919128417967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,4,1,power_law_1.01,0.20593343734741212
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,4,1,power_law_1.01,0.20811519622802735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,4,1,power_law_1.01,0.2173446464538574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,4,1,power_law_1.01,0.22353919982910156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,4,1,power_law_1.01,0.2256697654724121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,4,1,power_law_1.01,0.23054847717285157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,4,1,power_law_1.01,0.26034112930297854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,4,1,power_law_1.01,0.26291263580322266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,4,1,power_law_1.01,0.2736031913757324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,4,1,power_law_1.01,0.28319679260253905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,4,1,power_law_1.01,0.30742464065551756
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,4,1,power_law_1.01,0.33763137817382816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,4,1,power_law_1.01,0.3756307220458984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,4,1,power_law_1.01,0.4492927932739258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,4,1,power_law_1.01,0.5233689498901367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,4,1,power_law_1.01,0.6777958679199219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,4,1,power_law_1.01,0.840360336303711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,4,1,power_law_1.01,1.1516505432128905
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,4,1,power_law_1.01,0.05698048114776612
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,4,1,power_law_1.01,1.4625599670410157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,4,1,power_law_1.01,0.058768639564514155
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,4,1,power_law_1.01,0.06999872207641603
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,4,1,balanced,0.3848332977294922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,4,1,power_law_1.01,0.0792307186126709
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,4,1,power_law_1.01,0.09214336395263671
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,4,1,balanced,0.4385273742675781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,4,1,power_law_1.01,0.10478912353515626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,4,1,power_law_1.01,0.13264448165893555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,4,1,balanced,0.8575878143310547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,4,1,balanced,0.6252479934692383
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,4,1,power_law_1.01,0.13441280364990232
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,4,1,power_law_1.01,0.1373145580291748
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,4,1,power_law_1.01,0.1431462383270264
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,4,1,power_law_1.01,0.1467251205444336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,4,1,power_law_1.01,0.12631168365478515
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,4,1,power_law_1.01,0.1500486373901367
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,4,1,power_law_1.01,0.1551238441467285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,4,1,power_law_1.01,0.1674131202697754
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,4,1,power_law_1.01,0.1618124771118164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,4,1,power_law_1.01,0.17839935302734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,4,1,balanced,0.23068096160888674
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,4,1,power_law_1.01,0.22687871932983397
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,4,1,power_law_1.01,0.2783091163635254
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,4,1,power_law_1.01,0.20067840576171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,4,1,power_law_1.2,0.03491199970245361
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,4,1,power_law_1.01,0.33148414611816407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,4,1,power_law_1.2,0.05590911865234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,4,1,power_law_1.01,0.43829566955566407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,4,1,power_law_1.2,0.07119999885559082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,4,1,power_law_1.2,0.10193984031677246
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,4,1,power_law_1.01,0.5546489715576172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,4,1,power_law_1.01,0.7957113647460938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,4,1,power_law_1.2,0.19381759643554688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,4,1,power_law_1.2,0.26094015121459957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,4,1,power_law_1.2,0.27490943908691406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,4,1,power_law_1.2,0.27073152542114254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,4,1,power_law_1.2,0.28482879638671876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,4,1,power_law_1.2,0.15676671981811524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,4,1,power_law_1.2,0.31832256317138674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,4,1,power_law_1.2,0.3477471923828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,4,1,power_law_1.2,0.26777919769287106
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,4,1,power_law_1.2,0.3530668640136719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,4,1,power_law_1.2,0.35893375396728516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,4,1,power_law_1.2,0.38263038635253904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,4,1,power_law_1.2,0.3041971206665039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,4,1,power_law_1.01,2.0886163330078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,4,1,power_law_1.2,0.4236800003051758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,4,1,power_law_1.2,0.44746559143066406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,4,1,power_law_1.2,0.5234003067016602
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,4,1,power_law_1.01,1.555773468017578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,4,1,power_law_1.2,0.5853772735595703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,4,1,power_law_1.01,1.0344287872314453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,4,1,power_law_1.2,0.7258144378662109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,4,1,power_law_1.2,0.03788991928100586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,4,1,power_law_1.2,0.06018112182617188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,4,1,power_law_1.2,0.8859910583496093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,4,1,power_law_1.2,0.08861568450927734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,4,1,power_law_1.2,1.1993977355957033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,4,1,power_law_1.2,0.1147270393371582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,4,1,power_law_1.2,1.5023001098632813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,4,1,power_law_1.2,0.15592512130737304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,4,1,power_law_1.2,0.1911564826965332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,4,1,power_law_1.2,0.20122240066528319
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,4,1,power_law_1.2,0.2023308753967285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,4,1,power_law_1.2,0.2125555229187012
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,4,1,power_law_1.2,0.21648767471313474
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,4,1,power_law_1.2,0.22229375839233398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,4,1,power_law_1.2,0.22694847106933597
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,4,1,power_law_1.2,2.131395263671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,4,1,power_law_1.2,0.23446016311645507
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,4,1,power_law_1.2,0.2655014419555664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,4,1,power_law_1.2,0.2689068794250488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,4,1,power_law_1.2,0.2747923278808594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,4,1,power_law_1.2,0.29966911315917966
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,4,1,power_law_1.2,0.3182156753540039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,4,1,power_law_1.2,0.35594432830810546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,4,1,power_law_1.2,0.3852876663208008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,4,1,power_law_1.2,0.46355648040771485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,4,1,power_law_1.2,0.5472870254516602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,4,1,power_law_1.2,0.687733154296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,4,1,power_law_1.2,0.8573222351074218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,4,1,power_law_1.2,1.1688947296142578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,4,1,power_law_1.2,0.05664832115173339
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,4,1,power_law_1.2,0.05928319931030274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,4,1,power_law_1.2,1.4818194580078123
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,4,1,power_law_1.2,0.06944767951965332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,4,1,power_law_1.2,0.07792511940002442
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,4,1,power_law_1.2,0.0377510404586792
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,4,1,power_law_1.2,0.08870207786560058
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,4,1,power_law_1.2,0.10149824142456054
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,4,1,power_law_1.2,0.1254918384552002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,4,1,power_law_1.2,2.76963134765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,4,1,power_law_1.2,0.13387392044067384
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,4,1,power_law_1.2,0.13783040046691894
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,4,1,power_law_1.2,0.14852736473083497
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,4,1,power_law_1.2,0.14357312202453615
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,4,1,power_law_1.2,0.15241215705871583
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,4,1,power_law_1.2,0.1306220817565918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,4,1,power_law_1.2,0.1610233688354492
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,4,1,power_law_1.2,0.16942783355712893
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,4,1,power_law_1.2,0.1770400047302246
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,4,1,power_law_1.2,0.19496383666992187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,4,1,power_law_1.2,0.3497740936279297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,4,1,power_law_1.2,0.2314041519165039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,4,1,power_law_1.2,0.2697900772094727
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,4,1,power_law_1.2,0.438403205871582
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,4,1,power_law_1.2,0.6242649459838867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,4,1,power_law_1.2,0.8287718200683594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,4,1,power_law_1.2,1.7458189392089842
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,4,balanced,0.04241536140441894
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,4,1,power_law_1.2,1.2707206726074218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,4,balanced,0.16831615447998047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,4,balanced,0.1784832000732422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,4,balanced,0.17404672622680664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,4,balanced,0.18017919540405275
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,4,balanced,0.21121536254882814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,4,balanced,0.13859583854675292
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,4,balanced,0.2154278373718262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,4,balanced,0.20166271209716796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,4,balanced,0.20692607879638675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,4,balanced,0.20151424407958984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,4,balanced,0.20797695159912108
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,4,balanced,0.22389759063720702
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,4,balanced,0.237893123626709
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,4,balanced,0.2733683204650879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,4,balanced,0.31448959350585937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,4,balanced,0.3547903823852539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,4,balanced,0.39891456604003905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,4,balanced,0.03709952116012573
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,4,balanced,0.42379135131835943
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,4,balanced,0.6229350280761718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,4,balanced,0.6940377807617187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,4,balanced,0.06441343784332275
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,4,balanced,0.979968032836914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,4,balanced,1.1253734588623048
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,4,balanced,0.19977472305297853
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,4,balanced,0.0887667179107666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,4,1,power_law_1.2,2.7600799560546876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,4,1,power_law_1.2,3.857667236328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,1,4,balanced,0.02391040086746216
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,1,4,balanced,0.07247744083404542
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,1,4,balanced,0.023975679874420168
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,1,4,balanced,0.028096001148223877
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,1,4,balanced,0.04140031814575195
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,1,4,balanced,0.07274367809295654
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,1,4,balanced,0.07247615814208984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,1,4,balanced,0.07283071994781494
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,1,4,balanced,0.07380352020263672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,1,4,balanced,0.0740774393081665
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,1,4,balanced,0.07436416149139405
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,1,4,balanced,0.07589248180389405
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,1,4,balanced,0.08008319854736327
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,1,4,balanced,0.08264320373535157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,1,4,balanced,0.08607104301452637
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,1,4,balanced,0.05931392192840577
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,1,4,balanced,0.07369728088378906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,1,4,balanced,0.07351935863494873
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,1,4,balanced,0.07240320205688476
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,1,4,balanced,0.08353023529052735
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,1,4,balanced,0.09796480178833009
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,4,power_law_1.01,0.07333119869232177
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,4,power_law_1.01,0.08232319831848145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,4,power_law_1.01,0.04730112075805664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,4,power_law_1.01,0.11455360412597657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,4,power_law_1.01,0.15827072143554688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,4,power_law_1.01,0.162608642578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,4,power_law_1.01,0.17213695526123046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,4,power_law_1.01,0.17404800415039062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,1,4,balanced,0.11900032043457032
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,1,4,balanced,0.17325439453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,4,power_law_1.01,0.18585983276367188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,1,4,balanced,0.12772607803344727
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,1,4,balanced,0.1895782470703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,4,power_law_1.01,0.19987840652465821
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,4,power_law_1.01,0.18708223342895508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,4,power_law_1.01,0.20020736694335936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,4,power_law_1.01,0.13665535926818848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,4,power_law_1.01,0.2220748710632324
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,1,4,balanced,0.288088321685791
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,4,power_law_1.01,0.3101107215881348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,4,power_law_1.01,0.3846220779418945
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,1,4,balanced,0.3558694458007813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,4,power_law_1.01,0.6446387481689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,4,power_law_1.01,0.20127872467041014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,4,power_law_1.01,0.8438886260986328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,4,power_law_1.01,0.1946726417541504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,4,power_law_1.01,0.19059072494506837
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,1,4,power_law_1.01,0.02790271997451782
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,1,4,power_law_1.01,0.040538878440856935
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,1,4,power_law_1.01,0.03927808046340943
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,1,4,power_law_1.01,0.04927616119384766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,1,4,power_law_1.01,0.06195968151092529
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,4,power_law_1.01,1.4243942260742188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,1,4,power_law_1.01,0.06207615852355956
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,1,4,power_law_1.01,0.071528959274292
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,1,4,power_law_1.01,0.06392447948455811
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,1,4,power_law_1.01,0.04666751861572266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,4,power_law_1.01,0.22902015686035154
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,1,4,power_law_1.01,0.0679091215133667
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,1,4,power_law_1.01,0.06761600017547607
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,1,4,power_law_1.01,0.07148799896240235
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,1,4,power_law_1.01,0.0696345615386963
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,1,4,power_law_1.01,0.07172736167907715
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,1,4,power_law_1.01,0.07415679931640626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,1,4,power_law_1.01,0.0638323211669922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,4,power_law_1.01,0.28783359527587893
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,4,power_law_1.01,0.4259238433837891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,4,power_law_1.01,0.5296089553833008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,4,power_law_1.01,1.0928972625732423
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,1,4,power_law_1.01,0.0795084810256958
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,1,4,power_law_1.01,0.0924351978302002
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,1,4,power_law_1.01,0.08842368125915527
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,1,4,power_law_1.01,0.13498496055603026
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,1,4,power_law_1.01,0.11158528327941894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,4,power_law_1.2,0.04722815990447998
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,1,4,power_law_1.01,0.18374656677246093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,4,power_law_1.2,0.10835200309753419
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,4,power_law_1.2,0.13468671798706056
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,4,power_law_1.2,0.1594278430938721
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,4,power_law_1.2,0.16729728698730467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,4,power_law_1.2,0.17325439453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,4,power_law_1.2,0.17733888626098632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,4,power_law_1.2,0.06255616188049316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,4,power_law_1.2,0.20327295303344725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,4,power_law_1.2,0.08384384155273437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,4,power_law_1.2,0.20545791625976562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,4,power_law_1.2,0.19772159576416015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,4,power_law_1.2,0.19689983367919922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,4,power_law_1.2,0.197076473236084
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,1,4,power_law_1.01,0.2228505516052246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,4,power_law_1.2,0.21017728805541994
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,4,power_law_1.2,0.20101247787475587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,4,power_law_1.2,0.23824384689331052
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,4,power_law_1.2,0.3101529693603516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,4,power_law_1.2,0.3012518310546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,4,power_law_1.2,0.4008921432495117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,4,power_law_1.2,0.4500223922729492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,4,power_law_1.2,0.7126809692382812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,4,power_law_1.2,0.8531545257568359
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,1,4,power_law_1.01,0.3194675254821777
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,1,4,power_law_1.2,0.02765183925628662
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,1,4,power_law_1.2,0.032641279697418216
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,1,4,power_law_1.2,0.040501761436462405
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,1,4,power_law_1.2,0.04637824058532715
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,1,4,power_law_1.2,0.04736127853393555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,1,4,power_law_1.2,0.06157824039459229
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,4,power_law_1.2,1.1598822021484376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,1,4,power_law_1.2,0.06237567901611328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,1,4,power_law_1.2,0.07044991970062256
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,1,4,power_law_1.2,0.06403200149536133
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,1,4,power_law_1.2,0.06801536083221435
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,1,4,power_law_1.2,0.06764927864074707
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,1,4,power_law_1.2,0.06946176052093506
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,1,4,power_law_1.2,0.0719475221633911
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,1,4,power_law_1.2,0.07517695903778077
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,1,4,power_law_1.2,0.07677567958831787
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,4,power_law_1.2,0.18505855560302736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,4,power_law_1.2,1.627308807373047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,1,4,power_law_1.2,0.1135923194885254
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,1,4,power_law_1.2,0.13634559631347656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,4,power_law_1.2,0.5305075073242188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,1,4,power_law_1.01,0.3727462387084961
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,1,4,power_law_1.01,0.5733030319213868
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,1,4,power_law_1.01,0.8103334045410155
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,1,4,power_law_1.2,0.07827839851379395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,1,4,power_law_1.2,0.08603263854980468
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,1,4,power_law_1.2,0.09102975845336914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,1,4,power_law_1.2,0.16498943328857424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,2,balanced,0.02941440105438232
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,2,balanced,0.037843201160430905
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,1,4,power_law_1.2,0.24193279266357423
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,2,balanced,0.12897279739379883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,2,balanced,0.1697983932495117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,2,balanced,0.1719820785522461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,2,balanced,0.1720204734802246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,2,balanced,0.051997442245483395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,2,balanced,0.17248640060424805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,2,balanced,0.17438207626342775
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,1,4,power_law_1.2,0.2719231986999512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,2,balanced,0.1776947212219238
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,2,balanced,0.18101247787475586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,2,balanced,0.17868928909301757
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,1,4,power_law_1.2,0.4723379135131836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,1,4,power_law_1.2,0.6036185455322266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,2,balanced,0.08354816436767579
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,2,balanced,0.23930240631103517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,2,balanced,0.3223040008544922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,2,balanced,0.338287353515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,2,balanced,0.3952703857421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,2,balanced,0.3987865447998047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,2,balanced,0.6182963180541992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,2,balanced,0.6834355163574218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,2,balanced,0.20176256179809568
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,2,balanced,0.9882316589355469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,2,balanced,0.20845823287963866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,2,balanced,0.19990272521972657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,2,balanced,0.2182489585876465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,2,balanced,1.101286392211914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,2,balanced,0.20216320037841795
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,1,4,power_law_1.2,0.78
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,1,4,power_law_1.2,1.2589273834228516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,2,2,balanced,0.02928895950317383
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,2,2,balanced,0.08123007774353028
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,2,2,balanced,0.06486527919769287
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,2,2,balanced,0.08205951690673828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,2,2,balanced,0.07998720169067383
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,2,2,balanced,0.08038144111633301
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,2,2,balanced,0.028546559810638427
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,2,2,balanced,0.08166144371032716
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,2,2,balanced,0.08123519897460937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,2,2,balanced,0.08219136238098144
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,2,2,balanced,0.08319104194641114
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,2,2,balanced,0.08370816230773925
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,2,2,balanced,0.08493696212768555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,2,2,balanced,0.08786304473876953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,2,2,balanced,0.09166720390319824
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,2,2,balanced,0.0956339168548584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,2,2,balanced,0.1016767978668213
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,2,2,balanced,0.03462656021118164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,2,2,balanced,0.048000001907348634
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,2,2,balanced,0.08070143699645996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,2,power_law_1.01,0.02932735919952393
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,2,2,balanced,0.08152192115783692
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,2,power_law_1.01,0.04642687797546387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,2,power_law_1.01,0.057148160934448244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,2,power_law_1.01,0.10628479957580565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,2,power_law_1.01,0.15064319610595703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,2,power_law_1.01,0.15477760314941408
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,2,power_law_1.01,0.1563967990875244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,2,power_law_1.01,0.1600192070007324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,2,power_law_1.01,0.1638387107849121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,2,power_law_1.01,0.16615423202514648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,2,power_law_1.01,0.08507391929626465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,2,power_law_1.01,0.17261823654174804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,2,power_law_1.01,0.18027904510498047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,2,2,balanced,0.16100608825683593
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,2,2,balanced,0.25088384628295896
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,2,2,balanced,0.22695039749145507
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,2,2,balanced,0.38659198760986324
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,2,2,balanced,0.47509887695312497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,2,power_law_1.01,0.19446783065795897
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,2,power_law_1.01,0.19661184310913088
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,2,power_law_1.01,0.3773196792602539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,2,power_law_1.01,0.5030118560791015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,2,power_law_1.01,0.24708736419677732
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,2,power_law_1.01,0.20690303802490234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,2,power_law_1.01,0.23001472473144532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,2,power_law_1.01,0.39687934875488284
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,2,2,power_law_1.01,0.02931328058242798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,2,power_law_1.01,0.7609190368652344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,2,2,power_law_1.01,0.03871615886688232
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,2,2,power_law_1.01,0.046196479797363285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,2,power_law_1.01,1.0323673248291017
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,2,2,power_law_1.01,0.053569278717041015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,2,2,power_law_1.01,0.07178368091583251
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,2,2,power_law_1.01,0.0688921594619751
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,2,2,power_law_1.01,0.07209983825683594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,2,power_law_1.01,0.3328857421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,2,2,power_law_1.01,0.07453184127807617
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,2,2,power_law_1.01,0.07452415943145752
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,2,2,power_law_1.01,0.03528064012527466
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,2,2,power_law_1.01,0.07591167926788331
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,2,2,power_law_1.01,0.07708928108215332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,2,2,power_law_1.01,0.07866623878479004
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,2,2,power_law_1.01,0.08084863662719727
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,2,2,power_law_1.01,0.08354047775268555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,2,2,power_law_1.01,0.08505472183227539
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,2,2,power_law_1.01,0.1407923221588135
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,2,power_law_1.01,1.401263427734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,2,power_law_1.01,0.6334323120117188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,2,power_law_1.01,0.21075584411621096
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,2,2,power_law_1.01,0.08948863983154297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,2,2,power_law_1.01,0.09995264053344727
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,2,2,power_law_1.01,0.1131545639038086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,2,2,power_law_1.01,0.3932876968383789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,2,2,power_law_1.01,0.1684761619567871
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,2,2,power_law_1.01,0.22492544174194337
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,2,2,power_law_1.01,0.2759884834289551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,2,power_law_1.2,0.04232063770294189
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,2,power_law_1.2,0.05893631935119629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,2,power_law_1.2,0.08099840164184571
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,2,power_law_1.2,0.10190976142883301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,2,power_law_1.2,0.13453439712524412
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,2,power_law_1.2,0.1495859241485596
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,2,power_law_1.2,0.15844096183776854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,2,power_law_1.2,0.16211967468261718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,2,power_law_1.2,0.1636070442199707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,2,power_law_1.2,0.16846847534179688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,2,power_law_1.2,0.17590784072875976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,2,power_law_1.2,0.18577280044555664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,2,power_law_1.2,0.19795583724975585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,2,power_law_1.2,0.2149427223205566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,2,power_law_1.2,0.21230079650878908
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,2,power_law_1.2,0.23604095458984373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,2,power_law_1.2,0.2587276840209961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,2,power_law_1.2,0.3505369567871094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,2,power_law_1.2,0.3877824020385742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,2,power_law_1.2,0.4173734283447265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,2,power_law_1.2,0.029283840656280518
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,2,power_law_1.2,0.509040641784668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,2,power_law_1.2,0.6425933074951172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,2,power_law_1.2,1.0541849517822266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,2,2,power_law_1.2,0.029982719421386715
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,2,2,power_law_1.2,0.03330431938171387
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,2,2,power_law_1.2,0.03855232000350952
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,2,2,power_law_1.2,0.04553599834442139
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,2,2,power_law_1.2,0.05255296230316162
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,2,power_law_1.2,0.20119039535522462
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,2,2,power_law_1.2,0.06941184043884277
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,2,2,power_law_1.2,0.07181695938110352
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,2,2,power_law_1.2,0.07461631774902344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,2,2,power_law_1.2,0.07540863990783692
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,2,2,power_law_1.2,0.07658495903015136
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,2,2,power_law_1.2,0.07734784126281738
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,2,2,power_law_1.2,0.08064640045166016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,2,2,power_law_1.2,0.0819711971282959
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,2,2,power_law_1.2,0.08593791961669922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,2,power_law_1.2,0.778919677734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,2,power_law_1.2,1.436327667236328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,2,2,power_law_1.2,0.11681535720825195
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,2,2,power_law_1.01,0.5259840011596679
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,2,2,power_law_1.01,0.7878105926513672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,2,2,power_law_1.01,0.999362564086914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,2,2,power_law_1.2,0.08702848434448243
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,2,2,power_law_1.2,0.09100288391113281
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,2,2,power_law_1.2,0.09700736045837402
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,2,2,power_law_1.2,0.12796799659729002
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,2,2,power_law_1.2,0.1564902400970459
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,2,2,power_law_1.2,0.2033126449584961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,4,1,balanced,0.029203839302062988
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,2,2,power_law_1.2,0.2565439987182617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,4,1,balanced,0.03738111972808838
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,4,1,balanced,0.0575219202041626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,4,1,balanced,0.08733504295349122
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,4,1,balanced,0.15643327713012695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,2,2,power_law_1.2,0.34440959930419923
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,4,1,balanced,0.1852070426940918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,2,2,power_law_1.2,0.5492966461181641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,4,1,balanced,0.1856377601623535
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,2,2,power_law_1.2,0.7460441589355469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,4,1,balanced,0.18968000411987304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,4,1,balanced,0.18112895965576173
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,2,2,power_law_1.2,1.119599380493164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,4,1,balanced,0.19303039550781248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,4,1,balanced,0.18956480026245118
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,4,1,balanced,0.25046079635620117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,4,1,balanced,0.21071231842041013
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,4,1,balanced,0.3057472038269043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,4,1,balanced,0.18418943405151367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,4,1,balanced,0.18839168548583984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,4,1,balanced,0.2799033546447754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,4,1,balanced,0.18815616607666014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,4,1,balanced,0.2969804763793945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,4,1,balanced,0.39230464935302733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,4,1,balanced,0.4175980758666992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,4,1,balanced,0.6514905548095703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,4,1,balanced,0.7010514831542969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,4,1,power_law_1.01,0.029155840873718263
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,4,1,balanced,1.0329542541503907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,4,1,balanced,1.1447647857666017
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,4,1,power_law_1.01,0.10254719734191894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,4,1,balanced,0.19008832931518554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,4,1,power_law_1.01,0.16090944290161133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,4,1,power_law_1.01,0.16576448440551758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,4,1,power_law_1.01,0.03742784023284912
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,2,2,power_law_1.2,1.3350515747070313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,4,1,power_law_1.01,0.17157184600830078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,4,1,power_law_1.01,0.17161600112915038
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,4,1,power_law_1.01,0.17915647506713867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,4,1,power_law_1.01,0.1865177536010742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,4,1,balanced,0.2374655914306641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,4,1,power_law_1.01,0.2077356719970703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,4,1,power_law_1.01,0.21393215179443362
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,4,1,power_law_1.01,0.22443967819213867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,4,1,power_law_1.01,0.24661439895629883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,4,1,power_law_1.01,0.28687551498413083
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,4,1,power_law_1.01,0.08178239822387695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,4,1,power_law_1.01,0.3078739166259766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,4,1,power_law_1.01,0.45132415771484374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,4,1,power_law_1.01,0.5261721420288086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,4,1,power_law_1.01,0.14836416244506836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,4,1,power_law_1.01,0.17649791717529298
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,4,1,power_law_1.01,0.6441069030761719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,4,1,power_law_1.01,0.05781184196472168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,4,1,power_law_1.01,0.7612294769287109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,4,1,power_law_1.01,1.025912322998047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,4,1,power_law_1.2,0.02902719974517822
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,4,1,power_law_1.2,0.05700479984283448
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,4,1,power_law_1.01,0.3488550567626953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,4,1,power_law_1.2,0.15066752433776856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,4,1,power_law_1.01,0.2614451217651367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,4,1,power_law_1.2,0.1694988822937012
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,4,1,power_law_1.01,1.2694886779785155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,4,1,power_law_1.2,0.1724403190612793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,4,1,power_law_1.2,0.17320255279541014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,4,1,power_law_1.2,0.1789695930480957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,4,1,power_law_1.2,0.18323583602905275
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,4,1,power_law_1.2,0.1931283187866211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,4,1,power_law_1.2,0.21229888916015627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,4,1,power_law_1.2,0.21836671829223633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,4,1,power_law_1.2,0.22913024902343748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,4,1,power_law_1.2,0.16111743927001954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,4,1,power_law_1.2,0.2493087959289551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,4,1,power_law_1.2,0.2947187232971191
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,4,1,power_law_1.2,0.35396991729736327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,4,1,power_law_1.2,0.46335617065429685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,4,1,power_law_1.2,0.040113282203674314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,4,1,power_law_1.2,0.07399551868438721
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,4,1,power_law_1.2,0.6508614349365234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,4,balanced,0.0872486400604248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,4,1,power_law_1.2,0.7757536315917969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,4,1,power_law_1.2,0.31315967559814456
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,4,1,power_law_1.2,0.2625145530700684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,4,1,power_law_1.2,0.5334995269775391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,4,balanced,0.484527359008789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,4,1,power_law_1.2,1.0451308441162108
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,4,balanced,0.26559360504150387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,4,balanced,0.1572275161743164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,4,balanced,1.252495346069336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,4,balanced,1.3403392028808594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,4,balanced,0.8865817260742187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,4,1,power_law_1.2,0.09609408378601074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,4,balanced,1.3588800048828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,4,balanced,1.3640127563476563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,4,1,power_law_1.2,1.3026074218750001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,4,balanced,1.3277081298828126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,4,balanced,1.2520486450195312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,4,balanced,1.4945318603515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,4,balanced,1.2333964538574218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,4,balanced,1.566604766845703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,4,balanced,1.2734015655517577
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,4,balanced,1.652033233642578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,4,balanced,2.1450303649902347
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,4,balanced,1.8023513793945312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,4,balanced,1.8948825073242186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,4,balanced,1.4754328918457031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,4,balanced,2.6111270141601564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,4,balanced,2.6698944091796877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,4,balanced,0.05826816082000732
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,4,balanced,0.30518016815185545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,4,balanced,4.565461730957031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,4,balanced,0.5024307250976563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,4,balanced,4.958202819824218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,4,balanced,1.471463623046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,4,balanced,0.7031219482421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,4,balanced,0.7132569885253905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,4,balanced,0.09778559684753418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,4,balanced,0.17266687393188476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,4,balanced,0.7259814453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,4,balanced,0.7434009552001953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,4,balanced,0.7115596771240235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,4,balanced,0.7468608093261719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,4,balanced,0.7866022491455078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,4,balanced,6.979254760742188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,4,balanced,0.8006604766845703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,4,balanced,0.828485107421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,4,balanced,0.8631871795654297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,4,balanced,0.8455884552001953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,4,balanced,7.681641235351563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,4,balanced,0.9604710388183595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,4,balanced,1.0345830535888672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,4,balanced,1.1215821075439454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,4,balanced,1.2418252563476562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,4,balanced,1.8808242797851562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,4,balanced,2.103626251220703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,4,balanced,0.9015052795410157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,4,balanced,3.004322509765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,4,balanced,3.491677551269531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,4,balanced,5.185628051757813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,4,balanced,6.442938842773438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,1,4,balanced,0.05387392044067383
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,1,4,balanced,0.07150080204010009
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,1,4,balanced,0.43356800079345703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,1,4,balanced,0.10641023635864258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,1,4,balanced,0.43454334259033206
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,1,4,balanced,0.4398732757568359
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,1,4,balanced,0.44104576110839844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,1,4,balanced,0.442737922668457
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,1,4,balanced,0.3133542442321777
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,1,4,balanced,0.4453247833251953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,1,4,balanced,0.44696575164794916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,1,4,balanced,0.18085376739501952
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,1,4,balanced,0.4489126586914063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,1,4,balanced,0.4324607849121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,1,4,balanced,0.4563046264648437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,1,4,balanced,0.4706521606445312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,1,4,balanced,0.4739328002929687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,1,4,balanced,0.5032614517211914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,1,4,balanced,0.5205683135986329
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,1,4,balanced,0.43127681732177736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,4,power_law_1.01,0.25276031494140627
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,1,4,balanced,0.432616958618164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,4,power_law_1.01,0.4745792007446289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,4,power_law_1.01,0.6322444915771485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,4,power_law_1.01,0.7863948822021485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,4,power_law_1.01,0.44408958435058593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,4,power_law_1.01,1.072442855834961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,4,power_law_1.01,1.1417395019531251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,4,power_law_1.01,1.0874444580078124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,4,power_law_1.01,1.1042559814453123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,4,power_law_1.01,1.1825676727294923
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,4,power_law_1.01,1.1923404693603517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,4,power_law_1.01,1.2298982238769531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,4,power_law_1.01,1.2653887939453123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,4,power_law_1.01,1.3752024841308592
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,4,power_law_1.01,1.3783949279785157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,4,power_law_1.01,1.4109709167480469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,4,power_law_1.01,1.6743026733398438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,1,4,balanced,0.78123779296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,1,4,balanced,0.8917056274414061
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,1,4,balanced,1.2073881530761719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,4,power_law_1.01,1.472711639404297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,1,4,balanced,1.3516018676757813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,1,4,balanced,2.1639257812499997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,4,power_law_1.01,2.0031333923339845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,4,power_law_1.01,2.231357421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,1,4,balanced,2.6094091796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,4,power_law_1.01,2.9747198486328124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,4,power_law_1.01,3.4308990478515624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,4,power_law_1.01,0.2690009689331055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,4,power_law_1.01,0.2863411140441895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,4,power_law_1.01,0.15692416191101075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,4,power_law_1.01,0.382158088684082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,4,power_law_1.01,4.361930236816407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,4,power_law_1.01,0.4422937774658203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,4,power_law_1.01,0.6019865417480469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,4,power_law_1.01,0.6223551940917968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,4,power_law_1.01,0.607715835571289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,4,power_law_1.01,0.6150758361816406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,4,power_law_1.01,0.6358911895751953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,4,power_law_1.01,2.1071142578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,4,power_law_1.01,0.6595558166503906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,4,power_law_1.01,0.7007615661621094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,4,power_law_1.01,0.7223744201660156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,4,power_law_1.01,0.7840716552734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,4,power_law_1.01,0.7837862396240235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,4,power_law_1.01,5.354854125976563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,4,power_law_1.01,0.7739545440673827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,4,power_law_1.01,0.9364749145507811
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,4,power_law_1.01,7.321911010742187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,4,power_law_1.01,1.018593292236328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,4,power_law_1.01,1.2172966766357423
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,4,power_law_1.01,1.4163955688476562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,4,power_law_1.01,1.8066380310058594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,1,4,power_law_1.01,0.11118207931518556
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,4,power_law_1.01,9.164494018554688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,1,4,power_law_1.01,0.17592960357666015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,4,power_law_1.01,2.241008605957031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,1,4,power_law_1.01,0.1821683120727539
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,1,4,power_law_1.01,0.21795711517333985
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,1,4,power_law_1.01,0.2734783935546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,4,power_law_1.01,3.1200421142578127
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,1,4,power_law_1.01,0.37729663848876954
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,1,4,power_law_1.01,0.3498099136352539
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,1,4,power_law_1.01,0.36312065124511717
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,1,4,power_law_1.01,0.36577663421630857
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,4,power_law_1.01,3.8175003051757814
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,1,4,power_law_1.01,0.3896614456176758
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,1,4,power_law_1.01,0.39547904968261716
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,1,4,power_law_1.01,0.40851200103759766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,1,4,power_law_1.01,0.4123545455932617
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,1,4,power_law_1.01,0.4218265533447266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,1,4,power_law_1.01,0.43977344512939454
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,1,4,power_law_1.01,0.44982143402099606
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,1,4,power_law_1.01,0.5247705459594727
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,1,4,power_law_1.01,0.577204475402832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,4,power_law_1.01,5.4990820312499995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,4,power_law_1.01,0.8452761840820312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,4,power_law_1.01,7.1800537109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,1,4,power_law_1.01,0.4690956878662109
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,1,4,power_law_1.01,0.6616973114013672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,1,4,power_law_1.01,0.8025138854980469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,4,power_law_1.2,0.2578163146972656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,1,4,power_law_1.01,1.2869760131835937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,1,4,power_law_1.01,1.0377677154541014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,4,power_law_1.2,0.4833599853515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,1,4,power_law_1.01,1.9050201416015624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,4,power_law_1.2,0.7705689239501953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,4,power_law_1.2,1.0618777465820313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,1,4,power_law_1.01,2.312482604980469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,4,power_law_1.2,0.34791423797607424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,4,power_law_1.2,1.1295410919189455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,4,power_law_1.2,0.5967347335815429
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,4,power_law_1.2,1.1368141174316406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,4,power_law_1.2,1.2241958618164062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,4,power_law_1.2,1.262835235595703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,1,4,power_law_1.01,3.5966912841796876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,4,power_law_1.2,1.3068684387207032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,4,power_law_1.2,1.4215245056152344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,4,power_law_1.2,1.41212158203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,4,power_law_1.2,1.515530242919922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,4,power_law_1.2,1.0403084564208984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,4,power_law_1.2,1.5399807739257814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,4,power_law_1.2,1.736873016357422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,4,power_law_1.2,2.2073638916015623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,4,power_law_1.2,2.1062669372558593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,4,power_law_1.2,2.432469787597656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,4,power_law_1.2,3.1103219604492187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,4,power_law_1.2,1.1888256072998047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,4,power_law_1.2,0.16216320037841797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,4,power_law_1.2,3.7224435424804687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,4,power_law_1.2,0.2132736015319824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,4,power_law_1.2,0.26857088088989256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,4,power_law_1.2,0.37345664978027343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,4,power_law_1.2,0.41729919433593754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,4,power_law_1.2,0.6119756698608398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,4,power_law_1.2,0.5911206436157227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,4,power_law_1.2,0.6083622360229493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,4,power_law_1.2,4.648423767089843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,4,power_law_1.2,0.6286822509765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,4,power_law_1.2,5.730057983398437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,4,power_law_1.2,0.658845443725586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,4,power_law_1.2,0.7448089599609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,4,power_law_1.2,0.7403890991210937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,4,power_law_1.2,0.8114572906494141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,4,power_law_1.2,0.7986752319335937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,4,power_law_1.2,0.8337471771240235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,4,power_law_1.2,7.589869995117188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,4,power_law_1.2,0.9033830261230469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,4,power_law_1.2,0.9539571380615234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,4,power_law_1.2,1.0585497283935548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,4,power_law_1.2,1.3066482543945312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,4,power_law_1.2,1.4974592590332032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,4,power_law_1.2,1.8492787170410157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,4,power_law_1.2,2.3251046752929687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,1,4,power_law_1.2,0.11013504028320313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,1,4,power_law_1.2,0.14559231758117677
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,4,power_law_1.2,0.6272076797485352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,4,power_law_1.2,4.131528930664063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,1,4,power_law_1.2,0.21708288192749023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,4,power_law_1.2,3.338965759277344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,4,power_law_1.2,9.708572387695312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,1,4,power_law_1.2,0.17919872283935548
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,1,4,power_law_1.2,0.2586252784729004
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,1,4,power_law_1.2,0.37384193420410156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,1,4,power_law_1.2,0.36262657165527346
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,1,4,power_law_1.2,0.37699710845947265
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,1,4,power_law_1.2,0.39286911010742187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,1,4,power_law_1.2,0.39074047088623043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,1,4,power_law_1.2,0.40674686431884766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,1,4,power_law_1.2,0.41161342620849606
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,1,4,power_law_1.2,0.42099456787109374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,4,power_law_1.2,5.833634643554687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,1,4,power_law_1.2,0.44520832061767573
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,1,4,power_law_1.2,0.4538675308227539
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,1,4,power_law_1.01,4.596355895996093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,4,power_law_1.2,7.423111572265626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,1,4,power_law_1.2,0.4748876953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,1,4,power_law_1.2,0.48116096496582034
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,1,4,power_law_1.2,0.6516019439697265
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,1,4,power_law_1.2,0.9174374389648439
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,1,4,power_law_1.2,0.7681855773925781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,1,4,power_law_1.2,0.5578598403930665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,2,balanced,0.07991551876068115
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,2,balanced,0.15127424240112303
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,2,balanced,0.2513830375671387
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,1,4,power_law_1.2,1.2915367126464843
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,1,4,power_law_1.2,1.8342259216308594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,2,balanced,0.4775052642822265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,2,balanced,0.869073944091797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,2,balanced,1.2612044525146484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,2,balanced,1.269128952026367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,2,balanced,1.2792217254638671
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,2,balanced,1.284734649658203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,2,balanced,1.2899749755859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,2,balanced,1.2532685089111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,2,balanced,1.3146701049804688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,2,balanced,1.3339865112304687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,2,balanced,1.4532978820800782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,2,balanced,1.4435571289062499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,2,balanced,1.4614579772949219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,2,balanced,1.4938546752929687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,2,balanced,1.6670938110351563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,2,balanced,2.3246797180175784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,2,balanced,2.3959397888183593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,2,balanced,2.5949798583984376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,2,balanced,3.0965069580078124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,2,balanced,1.5659878540039063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,2,balanced,4.292652893066406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,2,balanced,0.06521088123321533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,2,balanced,0.09678976058959961
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,1,4,power_law_1.2,2.5142451477050782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,2,balanced,0.16615680694580076
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,2,balanced,4.578031311035156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,2,balanced,0.2950745582580566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,2,balanced,0.5141439819335938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,1,4,power_law_1.2,3.1925747680664065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,2,balanced,0.7264012908935547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,2,balanced,0.7307507324218749
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,2,balanced,0.7400204467773437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,2,balanced,0.7564940643310547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,2,balanced,7.1147570800781255
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,2,balanced,0.7878489685058594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,2,balanced,0.790277099609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,1,4,power_law_1.2,5.069548950195313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,2,balanced,0.8192115020751952
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,2,balanced,0.8703974151611329
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,2,balanced,0.7633650970458985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,2,balanced,8.470374145507813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,2,balanced,0.9263295745849609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,2,balanced,0.9752627563476561
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,2,balanced,0.899651870727539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,2,balanced,1.0965055847167968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,2,balanced,0.7503807830810547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,2,balanced,1.2023999786376953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,1,4,power_law_1.2,6.554506225585937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,2,balanced,0.9948429107666016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,2,balanced,1.9025804138183595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,2,balanced,2.0762611389160157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,2,balanced,2.9561190795898438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,2,balanced,3.3664190673828123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,2,balanced,6.362150268554688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,2,balanced,5.077611389160156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,2,2,balanced,0.08728320121765136
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,2,2,balanced,0.07412735939025879
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,2,2,balanced,0.11941120147705078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,2,2,balanced,0.46524673461914057
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,2,2,balanced,0.4701567840576172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,2,2,balanced,0.4710784149169922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,2,2,balanced,0.1903615951538086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,2,2,balanced,0.4725363159179688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,2,2,balanced,0.47900543212890623
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,2,2,balanced,0.48532352447509763
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,2,2,balanced,0.4812633514404297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,2,2,balanced,0.49164543151855467
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,2,2,balanced,0.3295116806030273
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,2,2,balanced,0.4994137573242187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,2,2,balanced,0.527380485534668
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,2,2,balanced,0.5333311843872071
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,2,2,balanced,0.4629926300048828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,2,2,balanced,0.5601663970947266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,2,2,balanced,0.5984806442260743
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,2,2,balanced,0.46452480316162104
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,2,power_law_1.01,0.14994688034057618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,2,power_law_1.01,0.24967296600341796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,2,power_law_1.01,0.33680641174316406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,2,power_law_1.01,0.5156262588500977
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,2,2,balanced,0.46519424438476564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,2,power_law_1.01,0.6667033386230469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,2,power_law_1.01,1.0674803161621094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,2,power_law_1.01,1.0857369232177736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,2,power_law_1.01,1.0987622070312502
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,2,power_law_1.01,0.9732569885253906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,2,power_law_1.01,1.1583999633789062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,2,power_law_1.01,1.2287155151367188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,2,power_law_1.01,1.1761203002929688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,2,power_law_1.01,1.2659251403808593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,2,power_law_1.01,1.3918234252929689
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,2,power_law_1.01,1.4605824279785158
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,2,power_law_1.01,1.443583984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,2,power_law_1.01,1.3777407836914062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,2,power_law_1.01,1.6556083679199218
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,2,2,balanced,0.9415219116210938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,2,power_law_1.01,1.7401011657714842
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,2,2,balanced,1.0002662658691406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,2,2,balanced,1.474929962158203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,2,2,balanced,1.6585165405273437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,2,2,balanced,2.518219451904297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,2,power_law_1.01,2.3723199462890627
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,2,2,balanced,3.1139749145507816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,2,power_law_1.01,2.6017791748046877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,2,power_law_1.01,0.09552512168884278
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,2,power_law_1.01,3.039422607421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,2,power_law_1.01,0.16471168518066406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,2,power_law_1.01,0.19487104415893555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,2,power_law_1.01,0.38953086853027347
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,2,power_law_1.01,0.5904576110839843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,2,power_law_1.01,3.173143005371094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,2,power_law_1.01,4.251481628417968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,2,power_law_1.01,0.6011328125000001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,2,power_law_1.01,0.6186969757080079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,2,power_law_1.01,0.6560537719726562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,2,power_law_1.01,5.170975952148437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,2,power_law_1.01,0.6840038299560547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,2,power_law_1.01,0.6679360198974609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,2,power_law_1.01,0.6912371063232422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,2,power_law_1.01,0.7129689788818359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,2,power_law_1.01,0.7622105407714843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,2,power_law_1.01,0.7988992309570312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,2,power_law_1.01,0.8284889221191406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,2,power_law_1.01,0.31215232849121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,2,power_law_1.01,0.8638540649414063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,2,power_law_1.01,0.9676582336425781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,2,power_law_1.01,7.0865673828124995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,2,power_law_1.01,1.0514905548095703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,2,power_law_1.01,1.2633817291259766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,2,power_law_1.01,1.4020095825195313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,2,power_law_1.01,1.7802674865722659
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,2,power_law_1.01,8.8964111328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,2,2,power_law_1.01,0.09513855934143066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,2,power_law_1.01,2.1648832702636716
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,2,2,power_law_1.01,0.1319155216217041
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,2,2,power_law_1.01,0.15561856269836424
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,2,2,power_law_1.01,0.19638143539428712
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,2,2,power_law_1.01,0.24270336151123048
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,2,2,power_law_1.01,0.3593779373168945
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,2,2,power_law_1.01,0.3836198425292969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,2,power_law_1.01,2.953566589355469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,2,2,power_law_1.01,0.38923519134521484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,2,2,power_law_1.01,0.3919334411621094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,2,2,power_law_1.01,0.4065395355224609
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,2,2,power_law_1.01,0.4237222290039062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,2,2,power_law_1.01,0.4259276962280273
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,2,2,power_law_1.01,0.4374591827392578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,2,2,power_law_1.01,0.45557247161865233
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,2,2,power_law_1.01,0.478600959777832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,2,2,power_law_1.01,0.4878732681274414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,2,2,power_law_1.01,0.6474201965332032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,2,power_law_1.01,3.6656549072265627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,2,power_law_1.01,5.1961779785156255
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,2,power_law_1.01,6.676055297851562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,2,2,power_law_1.01,0.5163366317749023
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,2,2,power_law_1.01,0.5768998336791993
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,2,2,power_law_1.01,0.9093670654296876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,2,2,power_law_1.01,0.806228485107422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,2,power_law_1.2,0.14894975662231444
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,2,2,power_law_1.01,1.2155007934570312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,2,power_law_1.2,0.22681343078613284
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,2,power_law_1.2,0.33422847747802736
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,2,2,power_law_1.01,1.5712934875488282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,2,power_law_1.2,0.4697151947021484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,2,2,power_law_1.01,2.2199999999999998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,2,power_law_1.2,1.0276274871826172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,2,power_law_1.2,1.0847923278808593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,2,power_law_1.2,1.0891455841064452
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,2,power_law_1.2,1.1425753784179686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,2,power_law_1.2,1.1769177246093752
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,2,2,power_law_1.01,2.7147930908203124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,2,power_law_1.2,1.243441925048828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,2,power_law_1.2,1.2975526428222657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,2,power_law_1.2,0.6304716873168945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,2,power_law_1.2,1.4193370056152343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,2,power_law_1.2,1.7957618713378907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,2,power_law_1.2,1.056848602294922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,2,power_law_1.2,1.4076249694824219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,2,power_law_1.2,2.558830108642578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,2,power_law_1.2,1.5078182983398438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,2,power_law_1.2,1.5147442626953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,2,power_law_1.2,1.7212902832031252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,2,power_law_1.2,2.690729064941406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,2,power_law_1.2,3.070592041015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,2,power_law_1.2,3.2974707031249997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,2,power_law_1.2,0.09763456344604492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,2,power_law_1.2,0.19754880905151367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,2,power_law_1.2,0.2836185646057129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,2,power_law_1.2,0.3696665573120117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,2,power_law_1.2,4.329730529785157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,2,power_law_1.2,5.481066284179688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,2,power_law_1.2,0.5814400100708008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,2,power_law_1.2,0.6483340454101563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,2,power_law_1.2,0.13634688377380372
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,2,power_law_1.2,0.6376179122924805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,2,power_law_1.2,0.6304256057739257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,2,power_law_1.2,0.6468531036376953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,2,power_law_1.2,0.6761484527587891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,2,power_law_1.2,7.527211303710938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,2,power_law_1.2,0.7094694519042969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,2,power_law_1.2,0.7290662384033203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,2,power_law_1.2,0.7758118438720703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,2,power_law_1.2,0.8354316711425781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,2,power_law_1.2,0.8973286437988282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,2,power_law_1.2,0.9116121673583985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,2,power_law_1.2,0.9903679656982423
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,2,power_law_1.2,1.084860153198242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,2,power_law_1.2,1.2792179107666015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,2,power_law_1.2,1.4930406188964844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,2,power_law_1.2,1.7891404724121096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,2,power_law_1.2,2.2342233276367187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,2,2,power_law_1.2,0.09939200401306152
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,2,2,power_law_1.2,0.11641983985900879
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,2,2,power_law_1.01,4.204115295410157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,2,power_law_1.2,9.095603027343751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,2,power_law_1.2,3.0178765869140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,2,2,power_law_1.2,0.1475263977050781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,2,2,power_law_1.2,0.1845644760131836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,2,2,power_law_1.2,0.24868480682373048
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,2,2,power_law_1.2,0.37449600219726564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,2,power_law_1.2,3.721363220214844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,2,2,power_law_1.2,0.3797452926635742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,2,2,power_law_1.01,5.6521240234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,2,2,power_law_1.2,0.3927590560913086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,2,2,power_law_1.2,0.40498046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,2,2,power_law_1.2,0.4053862380981445
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,2,2,power_law_1.2,0.4337791824340821
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,2,2,power_law_1.2,0.4521318435668945
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,2,2,power_law_1.2,0.4545548629760742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,2,2,power_law_1.2,0.47251838684082037
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,2,2,power_law_1.2,0.49996673583984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,2,2,power_law_1.2,0.5234560012817383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,2,power_law_1.2,5.3202404785156245
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,2,2,power_law_1.2,0.6385292816162109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,2,power_law_1.2,6.879586791992187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,2,2,power_law_1.2,0.5623731231689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,2,2,power_law_1.2,0.7295999908447266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,2,2,power_law_1.2,0.9418701171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,2,2,power_law_1.2,1.1111257934570313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,2,2,power_law_1.2,1.488684844970703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,2,2,power_law_1.2,2.021429748535156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,4,1,balanced,0.09900032043457031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,2,2,power_law_1.2,3.0065728759765626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,4,1,balanced,0.15976192474365233
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,4,1,balanced,0.47520702362060546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,4,1,balanced,0.2968204879760742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,4,1,balanced,1.287084197998047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,2,2,power_law_1.2,3.9168984985351565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,4,1,balanced,0.8668498992919922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,4,1,balanced,1.3166688537597655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,4,1,balanced,1.3069036865234376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,4,1,balanced,1.3896070861816407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,4,1,balanced,1.2873875427246095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,4,1,balanced,1.3817689514160156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,4,1,balanced,1.3041523742675782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,4,1,balanced,1.3136441040039064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,4,1,balanced,1.6772108459472654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,4,1,balanced,1.3405580139160156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,4,1,balanced,1.3673843383789062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,4,1,balanced,1.5773036193847656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,4,1,balanced,2.4541651916503904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,4,1,balanced,1.6757656860351564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,4,1,balanced,1.9825209045410155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,4,1,balanced,2.090693817138672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,4,1,balanced,3.275343933105469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,4,1,balanced,2.970928039550781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,2,2,power_law_1.2,5.961419677734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,4,1,balanced,0.12368895530700683
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,4,1,balanced,0.19526079177856445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,4,1,balanced,0.5863123321533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,4,1,balanced,0.08009535789489745
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,4,1,balanced,0.8497964477539062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,4,1,balanced,5.086156921386719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,4,1,balanced,0.8504307556152344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,4,1,balanced,0.8672268676757813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,4,1,balanced,5.40851318359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,4,1,balanced,0.3257683181762695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,4,1,balanced,0.884625244140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,4,1,balanced,0.8877490997314453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,4,1,balanced,0.8734598541259766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,4,1,balanced,0.9044159698486329
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,4,1,balanced,0.9024857330322267
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,4,1,balanced,0.9263359832763672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,4,1,balanced,0.9208723449707031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,4,1,balanced,0.9240979003906251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,4,1,balanced,0.9461248016357422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,4,1,balanced,1.1679046630859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,4,1,balanced,1.1950041961669922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,4,1,balanced,1.2980184936523438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,4,1,balanced,1.394060821533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,4,1,balanced,2.205699768066406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,4,1,balanced,2.3890771484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,4,1,balanced,7.829956665039063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,4,1,balanced,3.339786987304687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,2,2,power_law_1.2,8.325228881835937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,4,1,balanced,8.467034912109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,4,1,balanced,3.7495712280273437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,4,1,balanced,5.765179443359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,4,1,balanced,6.971441650390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,4,1,balanced,0.11365568161010742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,4,1,balanced,0.11392000198364258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,4,1,balanced,0.11426495552062987
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,4,1,balanced,0.11919103622436525
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,4,1,balanced,0.11725376129150392
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,4,1,balanced,0.14984512329101562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,4,1,balanced,0.12755776405334474
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,4,1,balanced,0.15836031913757326
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,4,1,balanced,0.17332160949707032
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,4,1,balanced,0.13440256118774413
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,4,1,balanced,0.1429203224182129
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,4,1,balanced,0.2120153617858887
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,4,1,balanced,0.25738880157470706
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,4,1,balanced,0.22702079772949219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,4,1,balanced,0.3089254379272461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,4,1,balanced,0.3403827285766602
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,4,1,balanced,0.42579776763916016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,4,1,balanced,0.58720703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,4,1,balanced,0.7535270690917969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,4,1,balanced,1.0841030120849608
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,4,1,power_law_1.01,0.09943552017211914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,4,1,balanced,1.4132876586914063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,4,1,balanced,2.0733325195312498
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,4,1,balanced,2.736968994140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,4,1,power_law_1.01,0.4325196838378906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,4,1,balanced,4.052404479980469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,4,1,power_law_1.01,0.2927008056640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,4,1,power_law_1.01,0.15370431900024414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,4,1,power_law_1.01,1.0185011291503907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,4,1,power_law_1.01,0.583936653137207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,4,1,power_law_1.01,0.9760492706298829
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,4,1,balanced,5.403602905273438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,4,1,power_law_1.01,1.0678777313232422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,4,1,power_law_1.01,1.1050975799560547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,4,1,power_law_1.01,1.1311827087402344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,4,1,power_law_1.01,1.2689190673828126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,4,1,power_law_1.01,1.441724853515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,4,1,power_law_1.01,1.2980351257324219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,4,1,power_law_1.01,1.526581115722656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,4,1,power_law_1.01,1.7229568481445312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,4,1,power_law_1.01,1.1889523315429689
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,4,1,power_law_1.01,2.195763244628906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,4,1,power_law_1.01,2.525038757324219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,4,1,power_law_1.01,1.8757600402832033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,4,1,power_law_1.01,2.2707122802734374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,4,1,power_law_1.01,1.8600370788574216
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,4,1,power_law_1.01,3.742151794433594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,4,1,power_law_1.01,0.08090239524841308
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,4,1,power_law_1.01,0.1183135986328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,4,1,power_law_1.01,4.626265563964844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,4,1,power_law_1.01,5.588098754882813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,4,1,power_law_1.01,0.19319936752319336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,4,1,power_law_1.01,0.2963936042785645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,4,1,power_law_1.01,0.4212243270874024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,4,1,power_law_1.01,0.6535270690917969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,4,1,power_law_1.01,0.7009017944335938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,4,1,power_law_1.01,0.7288896179199218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,4,1,power_law_1.01,2.942396240234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,4,1,power_law_1.01,7.389608764648438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,4,1,power_law_1.01,0.7639027404785156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,4,1,power_law_1.01,0.7896320343017578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,4,1,power_law_1.01,0.8272383880615234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,4,1,power_law_1.01,0.7538470458984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,4,1,power_law_1.01,0.8425055694580077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,4,1,power_law_1.01,0.8535724639892578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,4,1,power_law_1.01,0.8544671630859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,4,1,power_law_1.01,0.8969983673095703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,4,1,power_law_1.01,1.1372134399414062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,4,1,power_law_1.01,0.9418009948730468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,4,1,power_law_1.01,9.181763916015624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,4,1,power_law_1.01,1.2174720001220702
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,4,1,power_law_1.01,1.5898643493652345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,4,1,power_law_1.01,1.4072160339355468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,4,1,power_law_1.01,2.0043431091308594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,4,1,power_law_1.01,2.368161315917969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,4,1,power_law_1.01,3.16275634765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,4,1,power_law_1.01,0.11401663780212404
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,4,1,power_law_1.01,0.11395135879516602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,4,1,power_law_1.01,3.9092813110351563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,4,1,power_law_1.01,0.11451264381408692
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,4,1,power_law_1.01,0.1145798397064209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,4,1,power_law_1.01,5.538145751953126
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,4,1,power_law_1.01,0.11992832183837891
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,4,1,power_law_1.01,0.12743231773376465
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,4,1,power_law_1.01,0.1354419231414795
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,4,1,power_law_1.01,0.1429075241088867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,4,1,power_law_1.01,7.096519165039062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,4,1,power_law_1.01,0.15103424072265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,4,1,power_law_1.01,0.15828672409057618
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,4,1,power_law_1.01,0.17382976531982422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,4,1,power_law_1.01,0.22613567352294922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,4,1,power_law_1.01,0.25693952560424804
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,4,1,power_law_1.01,0.3089823913574219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,4,1,power_law_1.01,0.2103923225402832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,4,1,power_law_1.01,0.34049728393554685
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,4,1,power_law_1.01,0.4246540832519531
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,4,1,power_law_1.01,0.5878124618530274
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,4,1,power_law_1.01,0.7526131439208984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,4,1,power_law_1.01,1.0844774627685547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,4,1,power_law_1.01,1.4205087280273436
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,4,1,power_law_1.01,2.0741363525390626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,4,1,power_law_1.2,0.15501503944396972
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,4,1,power_law_1.01,2.73236083984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,4,1,power_law_1.01,4.057130737304687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,4,1,power_law_1.2,0.09930047988891602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,4,1,power_law_1.2,0.4052652740478516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,4,1,power_law_1.2,0.2892198371887207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,4,1,power_law_1.2,1.013570556640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,4,1,power_law_1.01,5.376842651367188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,4,1,power_law_1.2,1.0544505310058594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,4,1,power_law_1.2,0.9622310638427735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,4,1,power_law_1.2,0.5262425613403321
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,4,1,power_law_1.2,1.151430435180664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,4,1,power_law_1.2,1.2902125549316406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,4,1,power_law_1.2,1.2091270446777345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,4,1,power_law_1.2,1.3500685119628906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,4,1,power_law_1.2,1.49826171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,4,1,power_law_1.2,1.6264576721191406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,4,1,power_law_1.2,1.785846405029297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,4,1,power_law_1.2,1.9928076171874998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,4,1,power_law_1.2,2.373363189697266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,4,1,power_law_1.2,1.9400051879882814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,4,1,power_law_1.2,1.1010336303710937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,4,1,power_law_1.2,2.3115916442871094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,4,1,power_law_1.2,2.527196807861328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,4,1,power_law_1.2,3.809981994628906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,4,1,power_law_1.2,0.08129728317260743
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,4,1,power_law_1.2,4.697153930664062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,4,1,power_law_1.2,0.1187507152557373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,4,1,power_law_1.2,0.19258176803588867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,4,1,power_law_1.2,0.2793337631225586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,4,1,power_law_1.2,0.3859750366210938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,4,1,power_law_1.2,0.6432300567626953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,4,1,power_law_1.2,0.6994617462158204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,4,1,power_law_1.2,5.718413085937501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,4,1,power_law_1.2,3.050340576171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,4,1,power_law_1.2,0.7203705596923828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,4,1,power_law_1.2,0.7514201354980469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,4,1,power_law_1.2,0.7679424285888672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,4,1,power_law_1.2,0.8008614349365235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,4,1,power_law_1.2,0.8202374267578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,4,1,power_law_1.2,0.838031997680664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,4,1,power_law_1.2,0.8848588562011719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,4,1,power_law_1.2,0.8986796569824218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,4,1,power_law_1.2,0.9178854370117188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,4,1,power_law_1.2,0.983663330078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,4,1,power_law_1.2,1.194522247314453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,4,1,power_law_1.2,1.2840121459960936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,4,1,power_law_1.2,1.4688255310058593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,4,1,power_law_1.2,1.6449932861328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,4,1,power_law_1.2,2.0508102416992187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,4,1,power_law_1.2,7.543391723632813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,4,1,power_law_1.2,2.416913299560547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,4,1,power_law_1.2,3.1861285400390624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,4,1,power_law_1.2,0.11426943778991698
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,4,1,power_law_1.2,0.11290304183959961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,4,1,power_law_1.2,3.9285516357421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,4,1,power_law_1.2,0.11367808341979982
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,4,1,power_law_1.2,0.11552576065063476
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,4,1,power_law_1.2,0.11928319931030273
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,4,1,power_law_1.2,0.12695487976074218
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,4,1,power_law_1.2,0.1340358352661133
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,4,1,power_law_1.2,0.14264127731323242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,4,1,power_law_1.2,0.15049728393554687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,4,1,power_law_1.2,0.15797504425048828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,4,1,power_law_1.2,5.574711303710937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,4,1,power_law_1.2,0.17391424179077147
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,4,1,power_law_1.2,0.21129087448120115
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,4,1,power_law_1.2,0.22613567352294922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,4,1,power_law_1.2,0.30834239959716797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,4,1,power_law_1.2,9.224256591796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,4,1,power_law_1.2,7.1659521484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,4,1,power_law_1.2,0.42433025360107424
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,4,1,power_law_1.2,0.34226112365722655
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,4,1,power_law_1.2,0.2573043251037598
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,4,1,power_law_1.2,0.5873433685302734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,4,1,power_law_1.2,0.7523590087890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,4,1,power_law_1.2,1.4126911926269532
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,4,1,power_law_1.2,4.052338562011719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,1,8,balanced,0.16418560028076173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,1,8,balanced,0.16001152038574218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,1,8,balanced,0.15994239807128907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,1,8,balanced,0.2841958427429199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,1,8,balanced,0.2846092796325684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,1,8,balanced,0.2846937561035156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,1,8,balanced,0.28527360916137695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,1,8,balanced,0.28435071945190427
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,1,8,balanced,0.2866022491455078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,1,8,balanced,0.286746883392334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,1,8,balanced,0.28690687179565433
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,1,8,balanced,0.2924275207519531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,1,8,balanced,0.2930879974365234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,1,8,balanced,0.3342758560180664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,4,1,power_law_1.2,1.0832371520996094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,1,8,balanced,0.3353126525878906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,1,8,balanced,0.344318733215332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,1,8,balanced,0.3625190353393555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,1,8,balanced,0.39185791015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,1,8,balanced,0.4627276611328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,1,8,balanced,0.5144908905029297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,1,8,balanced,0.7085836791992188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,1,8,balanced,0.9056678771972656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,1,8,balanced,1.288476104736328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,1,8,balanced,1.7362547302246092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,1,8,balanced,2.6612145996093752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,1,8,balanced,0.09142271995544433
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,1,8,balanced,0.0916812801361084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,1,8,balanced,0.09019904136657715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,1,8,balanced,0.09887743949890136
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,1,8,balanced,0.09910143852233887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,1,8,balanced,0.29991039276123044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,1,8,balanced,0.09915904045104981
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,1,8,balanced,0.09715071678161621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,1,8,balanced,0.09939200401306152
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,1,8,balanced,0.09740415573120118
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,1,8,balanced,0.10067199707031249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,1,8,balanced,3.8518823242187503
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,1,8,balanced,0.10076031684875489
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,1,8,balanced,0.10646656036376953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,1,8,balanced,0.1025984001159668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,1,8,balanced,0.10621439933776855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,1,8,balanced,0.1056499195098877
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,4,1,power_law_1.2,2.082610626220703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,1,8,balanced,0.10917119979858399
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,1,8,balanced,0.11196288108825683
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,1,8,balanced,0.13817983627319336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,1,8,balanced,0.16589567184448242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,1,8,balanced,0.17810176849365233
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,1,8,balanced,0.2197657585144043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,1,8,balanced,0.263187198638916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,1,8,balanced,0.3824319839477539
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,4,1,power_law_1.2,2.7337985229492188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,1,8,balanced,0.4668019104003906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,1,8,balanced,0.7051929473876953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,1,8,balanced,0.9118566131591797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,1,8,balanced,0.13005696296691896
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,4,1,power_law_1.2,5.372266845703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,1,8,balanced,0.04117631912231445
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,1,8,balanced,0.04124032020568848
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,1,8,balanced,0.04109824180603027
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,1,8,balanced,0.04124032020568848
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,1,8,balanced,0.0413375997543335
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,1,8,balanced,0.041635842323303224
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,1,8,balanced,0.04016128063201904
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,1,8,balanced,0.041748480796813967
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,1,8,balanced,0.04124800205230713
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,1,8,balanced,0.04245888233184815
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,1,8,balanced,0.042049279212951665
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,1,8,balanced,0.04067200183868408
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,1,8,balanced,0.04154240131378174
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,1,8,balanced,0.04321280002593994
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,1,8,balanced,0.05436160087585449
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,1,8,balanced,0.04328832149505615
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,1,8,balanced,0.043265280723571775
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,1,8,balanced,0.05315711975097657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,1,8,balanced,0.05291776180267334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,1,8,balanced,0.06438144207000732
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,1,8,balanced,0.0765772819519043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,1,8,balanced,0.1263308811187744
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,1,8,balanced,0.10461055755615234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,1,8,power_law_1.01,0.15976063728332518
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,1,8,power_law_1.01,0.284323844909668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,1,8,power_law_1.01,0.2845939254760742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,1,8,power_law_1.01,0.2869913673400879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,1,8,power_law_1.01,0.2821017646789551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,1,8,power_law_1.01,0.28390016555786135
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,1,8,power_law_1.01,0.2976486396789551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,1,8,power_law_1.01,0.28693504333496095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,1,8,power_law_1.01,0.3115135955810547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,1,8,power_law_1.01,0.33586303710937504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,1,8,power_law_1.01,0.3410764694213867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,1,8,power_law_1.01,0.34685440063476564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,1,8,power_law_1.01,0.2915545654296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,1,8,power_law_1.01,0.3696844863891601
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,1,8,power_law_1.01,0.40230785369873046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,1,8,power_law_1.01,0.4837503814697266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,1,8,power_law_1.01,0.44517375946044924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,1,8,power_law_1.01,0.670904312133789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,1,8,power_law_1.01,0.8558233642578126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,1,8,power_law_1.01,1.24919677734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,1,8,power_law_1.01,1.6094323730468751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,1,8,power_law_1.01,2.9913818359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,1,8,power_law_1.01,2.5418663024902344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,1,8,power_law_1.01,5.4000244140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,1,8,power_law_1.01,0.0915993595123291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,1,8,power_law_1.01,8.059710693359374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,1,8,power_law_1.01,0.09817728042602539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,1,8,power_law_1.01,0.09937408447265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,1,8,power_law_1.01,0.09983360290527343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,1,8,power_law_1.01,0.09685759544372559
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,1,8,power_law_1.01,0.10085503578186035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,1,8,power_law_1.01,0.1025113582611084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,1,8,power_law_1.01,0.10392064094543456
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,1,8,power_law_1.01,0.10477184295654296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,1,8,power_law_1.01,0.10838527679443359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,1,8,power_law_1.01,0.09987968444824219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,1,8,power_law_1.01,0.11636223793029785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,1,8,power_law_1.01,0.12488832473754882
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,1,8,power_law_1.01,0.13319168090820313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,1,8,power_law_1.01,0.14947199821472168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,1,8,power_law_1.01,0.15677824020385742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,1,8,power_law_1.01,0.17224319458007814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,1,8,power_law_1.01,0.22045951843261719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,1,8,power_law_1.01,0.20887296676635742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,1,8,power_law_1.01,0.3499827194213867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,1,8,power_law_1.01,0.41091583251953123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,1,8,power_law_1.01,0.7401945495605469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,1,8,power_law_1.01,0.6386419296264648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,1,8,power_law_1.01,11.981201171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,1,8,power_law_1.01,1.1510809326171876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,1,8,power_law_1.01,1.5761459350585938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,1,8,power_law_1.01,0.04513792037963867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,1,8,balanced,0.233175048828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,1,8,balanced,0.44827903747558595
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,1,8,power_law_1.01,0.04033408164978027
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,1,8,power_law_1.01,0.039777278900146484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,1,8,power_law_1.01,0.04073088169097901
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,1,8,power_law_1.01,0.039741439819335936
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,1,8,power_law_1.01,0.0413043212890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,1,8,power_law_1.01,0.04036223888397217
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,1,8,power_law_1.01,0.04069759845733643
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,1,8,power_law_1.01,0.042455039024353027
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,1,8,power_law_1.01,0.041333761215209965
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,1,8,power_law_1.01,0.04214784145355225
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,1,8,power_law_1.01,0.04851327896118164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,1,8,power_law_1.01,0.05264639854431152
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,1,8,balanced,0.17669248580932617
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,1,8,balanced,0.3462886428833008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,1,8,power_law_1.01,2.493510437011719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,1,8,power_law_1.01,0.11260288238525391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,1,8,power_law_1.01,19.770533447265628
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,1,8,power_law_1.01,3.3098547363281248
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,1,8,power_law_1.01,0.059264001846313474
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,1,8,power_law_1.01,0.047316479682922366
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,1,8,power_law_1.01,0.06129536151885986
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,1,8,power_law_1.01,0.0688102388381958
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,1,8,power_law_1.01,0.09833344459533691
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,1,8,power_law_1.01,0.16619647979736327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,1,8,power_law_1.2,0.16151296615600585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,1,8,power_law_1.2,0.280577278137207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,1,8,power_law_1.2,0.2840627288818359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,1,8,power_law_1.2,0.2845401573181152
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,1,8,power_law_1.01,0.22201215744018557
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,1,8,power_law_1.2,0.2860249519348145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,1,8,power_law_1.2,0.2877350425720215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,1,8,power_law_1.2,0.2925222396850586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,1,8,power_law_1.2,0.2859148788452148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,1,8,power_law_1.2,0.30614015579223636
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,1,8,power_law_1.2,0.3211993789672852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,1,8,power_law_1.2,0.3367641448974609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,1,8,power_law_1.2,0.3466124725341797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,1,8,power_law_1.2,0.3529126358032227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,1,8,power_law_1.2,0.37345409393310547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,1,8,power_law_1.2,0.42478591918945313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,1,8,power_law_1.01,0.31159807205200196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,1,8,power_law_1.2,0.4526668930053711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,1,8,power_law_1.2,0.49552768707275396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,1,8,power_law_1.2,0.6905817413330079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,1,8,power_law_1.2,0.8718886566162108
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,1,8,power_law_1.2,1.2289024353027345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,1,8,power_law_1.2,1.6752281188964844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,1,8,power_law_1.2,2.607575073242187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,1,8,power_law_1.2,3.299225463867187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,1,8,power_law_1.2,5.69886474609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,1,8,power_law_1.2,11.0297607421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,1,8,power_law_1.2,0.0889408016204834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,1,8,power_law_1.2,0.09871871948242188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,1,8,power_law_1.01,0.40862464904785156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,1,8,power_law_1.2,0.09948160171508788
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,1,8,power_law_1.01,0.6188684844970703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,1,8,power_law_1.2,0.10466815948486327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,1,8,power_law_1.2,0.09945728302001952
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,1,8,power_law_1.2,0.11425151824951171
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,1,8,power_law_1.2,0.10797056198120117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,1,8,power_law_1.2,0.10254079818725585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,1,8,power_law_1.2,0.10747648239135743
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,1,8,power_law_1.2,0.10874367713928224
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,1,8,power_law_1.2,0.10701184272766114
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,1,8,power_law_1.2,0.11721856117248534
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,1,8,power_law_1.01,0.7979571533203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,1,8,power_law_1.2,0.12655872344970703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,1,8,power_law_1.2,0.13561216354370118
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,1,8,power_law_1.2,0.1730624008178711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,1,8,power_law_1.2,0.15438464164733887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,1,8,power_law_1.2,0.2292799949645996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,1,8,power_law_1.2,0.21217792510986327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,1,8,power_law_1.2,0.35394302368164066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,1,8,power_law_1.2,0.5911027145385742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,1,8,power_law_1.01,1.3368946838378908
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,1,8,power_law_1.2,0.44183166503906246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,1,8,power_law_1.2,9.06248291015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,1,8,power_law_1.2,0.7990220642089844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,1,8,power_law_1.2,0.04147840023040771
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,1,8,power_law_1.2,1.2724928283691406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,1,8,power_law_1.2,0.04184703826904297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,1,8,power_law_1.2,0.042302718162536626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,1,8,power_law_1.2,0.0420300817489624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,1,8,power_law_1.2,0.04253952026367187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,1,8,power_law_1.2,0.04169087886810303
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,1,8,power_law_1.2,0.04232319831848145
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,1,8,power_law_1.2,0.04275712013244629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,1,8,power_law_1.2,0.16108671188354493
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,1,8,power_law_1.2,0.04286464214324951
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,1,8,power_law_1.2,0.04296703815460205
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,1,8,power_law_1.2,0.0419814395904541
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,1,8,power_law_1.2,0.04754687786102295
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,1,8,power_law_1.2,0.050228481292724614
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,1,8,power_law_1.2,0.05280896186828613
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,1,8,power_law_1.2,0.05997183799743653
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,1,8,power_law_1.2,0.0616051197052002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,1,8,power_law_1.2,2.5805133056640623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,1,8,power_law_1.2,18.502336425781248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,1,8,power_law_1.2,1.7128933715820314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,1,8,power_law_1.2,3.5710961914062502
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,1,8,power_law_1.01,1.9362252807617188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,1,8,power_law_1.2,0.07228799819946288
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,1,8,power_law_1.2,0.11662848472595215
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,1,8,power_law_1.2,0.102225923538208
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,1,8,power_law_1.2,0.1662656021118164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,2,4,balanced,0.0961292839050293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,2,4,balanced,0.09553791999816894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,2,4,balanced,0.11508864402770995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,2,4,balanced,0.18313983917236326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,2,4,balanced,0.1850150489807129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,2,4,balanced,0.18452735900878908
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,2,4,balanced,0.18644096374511718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,2,4,balanced,0.18624895095825195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,2,4,balanced,0.18862207412719728
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,2,4,balanced,0.18871040344238282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,2,4,balanced,0.1892313575744629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,2,4,balanced,0.19295103073120118
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,2,4,balanced,0.19437952041625978
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,2,4,balanced,0.20237823486328127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,2,4,balanced,0.2408844757080078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,1,8,power_law_1.2,0.8961318206787109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,2,4,balanced,0.24336000442504885
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,2,4,balanced,0.2829747200012207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,2,4,balanced,0.2852582359313965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,2,4,balanced,0.38780033111572265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,2,4,balanced,0.4840716934204101
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,2,4,balanced,0.663641586303711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,2,4,balanced,0.8576204681396484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,2,4,balanced,0.24894208908081056
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,2,4,balanced,1.2898291015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,2,4,balanced,1.650697021484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,1,8,power_law_1.2,0.2292505645751953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,2,4,balanced,0.05292416095733643
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,2,4,balanced,0.05346303939819336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,1,8,power_law_1.2,0.2856831932067871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,2,4,balanced,0.06773375988006591
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,2,4,balanced,0.07296127796173095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,2,4,balanced,0.07377535820007323
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,2,4,balanced,0.07387392044067383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,2,4,balanced,0.07420544147491455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,2,4,balanced,2.607466125488281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,2,4,balanced,0.07598336219787598
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,2,4,balanced,0.07795584201812744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,2,4,balanced,0.077292799949646
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,2,4,balanced,0.08071040153503418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,2,4,balanced,0.07861120223999023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,2,4,balanced,0.08284543991088868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,2,4,balanced,0.08456831932067871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,2,4,balanced,0.08823936462402344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,2,4,balanced,0.07428607940673829
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,2,4,balanced,0.09209983825683594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,2,4,balanced,0.11477375984191895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,2,4,balanced,0.10986240386962891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,2,4,balanced,0.13668095588684082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,2,4,balanced,0.16390655517578126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,2,4,balanced,0.22578176498413086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,2,4,balanced,3.509744567871094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,2,4,balanced,0.27263359069824217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,2,4,balanced,0.39599487304687503
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,2,4,balanced,0.515475196838379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,2,4,balanced,0.7918988800048827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,2,4,balanced,0.9714176177978515
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,1,8,power_law_1.2,0.3957555389404297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,1,8,power_law_1.2,0.6641177368164063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,2,4,balanced,0.022337279319763183
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,2,4,balanced,0.037100799083709717
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,2,4,balanced,0.03747967958450317
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,2,4,balanced,0.022370560169219973
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,2,4,balanced,0.03724031925201416
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,2,4,balanced,0.037489919662475585
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,2,4,balanced,0.037611520290374754
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,2,4,balanced,0.03750911951065063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,2,4,balanced,0.03746943950653076
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,2,4,balanced,0.03795968055725098
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,2,4,balanced,0.03666431903839111
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,2,4,balanced,0.03764991998672486
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,2,4,balanced,0.03811072111129761
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,2,4,balanced,0.03871104001998901
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,2,4,balanced,0.036104960441589354
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,2,4,balanced,0.03554816007614135
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,1,8,power_law_1.2,1.303909149169922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,2,4,balanced,0.03647104024887085
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,2,4,balanced,0.048172798156738285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,2,4,balanced,0.05056767940521241
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,2,4,balanced,0.06431871891021729
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,2,4,balanced,0.07569407939910888
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,2,4,power_law_1.01,0.09856384277343751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,2,4,power_law_1.01,0.114202880859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,2,4,power_law_1.01,0.1820070457458496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,2,4,power_law_1.01,0.18370559692382812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,2,4,power_law_1.01,0.1845452880859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,2,4,power_law_1.01,0.18498176574707031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,2,4,power_law_1.01,0.18673664093017578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,2,4,power_law_1.01,0.18904832839965818
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,1,8,power_law_1.2,1.8990296936035158
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,2,4,power_law_1.01,0.19026559829711914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,2,4,power_law_1.01,0.21536895751953128
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,2,4,power_law_1.01,0.2145267105102539
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,2,4,balanced,0.09872384071350097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,2,4,power_law_1.01,0.2312179183959961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,2,4,power_law_1.01,0.23335296630859376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,2,4,power_law_1.01,0.25766143798828123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,2,4,power_law_1.01,0.26653440475463863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,2,4,power_law_1.01,0.291778564453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,2,4,power_law_1.01,0.33235198974609376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,2,4,balanced,0.12358400344848633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,2,4,power_law_1.01,0.4128678512573242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,2,4,power_law_1.01,0.5548492813110352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,2,4,power_law_1.01,0.6865843200683595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,2,4,power_law_1.01,0.8829183959960938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,2,4,balanced,0.17608320236206054
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,2,4,power_law_1.01,1.8083891296386718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,2,4,power_law_1.01,2.3164454650878907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,2,4,power_law_1.01,0.058764801025390626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,2,4,power_law_1.01,1.438561248779297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,2,4,power_law_1.01,0.0677184009552002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,2,4,power_law_1.01,0.07334911823272705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,2,4,power_law_1.01,0.07425024032592772
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,2,4,power_law_1.01,0.0755020809173584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,2,4,power_law_1.01,0.07473279953002929
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,2,4,power_law_1.01,0.0758028793334961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,2,4,power_law_1.01,0.07661056041717529
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,2,4,power_law_1.01,0.07848063945770264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,2,4,power_law_1.01,0.07710720062255859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,2,4,power_law_1.01,0.07997824192047119
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,2,4,power_law_1.01,0.08669568061828613
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,2,4,power_law_1.01,0.09267583847045899
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,2,4,power_law_1.01,0.09896320343017578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,2,4,power_law_1.01,0.10604415893554689
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,2,4,power_law_1.01,0.11032704353332519
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,2,4,power_law_1.01,0.12067839622497559
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,2,4,power_law_1.01,0.15069184303283692
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,2,4,power_law_1.01,3.4665765380859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,2,4,power_law_1.01,0.1474931240081787
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,2,4,power_law_1.01,0.22687999725341798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,2,4,power_law_1.01,0.279052791595459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,2,4,power_law_1.01,0.5050751876831054
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,2,4,power_law_1.01,0.7186803436279297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,2,4,power_law_1.01,0.418656005859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,2,4,power_law_1.01,1.801666564941406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,2,4,power_law_1.01,1.0257714843749999
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,2,4,power_law_1.01,0.034284799098968505
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,2,4,power_law_1.01,0.03354752063751221
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,2,4,power_law_1.01,0.036102399826049805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,2,4,power_law_1.01,0.03632895946502686
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,2,4,power_law_1.01,0.036514561176300045
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,2,4,power_law_1.01,0.03683712005615235
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,2,4,power_law_1.01,0.03674751996994018
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,2,4,power_law_1.01,0.03715712070465088
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,2,4,power_law_1.01,0.03709439992904663
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,2,4,power_law_1.01,0.03765120029449463
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,2,4,balanced,0.22223743438720703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,2,4,balanced,0.31912191390991207
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,2,4,power_law_1.01,0.04190720081329345
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,2,4,power_law_1.01,0.04337152004241943
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,2,4,balanced,0.42477695465087895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,2,4,power_law_1.01,1.622941436767578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,2,4,power_law_1.01,8.43365478515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,2,4,power_law_1.01,8.835851440429687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,2,4,power_law_1.01,0.026835200786590574
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,2,4,power_law_1.01,0.03987328052520752
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,2,4,power_law_1.01,0.04757887840270996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,2,4,power_law_1.01,0.05378943920135498
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,2,4,power_law_1.01,0.047608318328857424
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,2,4,power_law_1.01,0.08882304191589355
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,2,4,power_law_1.01,0.1117363166809082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,2,4,power_law_1.01,0.07474815845489502
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,2,4,power_law_1.2,0.09798399925231933
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,2,4,power_law_1.2,0.10911487579345702
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,2,4,power_law_1.2,0.18185216903686524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,2,4,power_law_1.2,0.18398975372314452
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,2,4,power_law_1.2,0.18416639328002932
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,2,4,power_law_1.2,0.18528383255004882
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,2,4,power_law_1.2,0.18685056686401366
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,2,4,power_law_1.2,0.18963071823120117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,2,4,power_law_1.2,0.19402496337890623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,2,4,power_law_1.2,0.19988479614257812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,2,4,power_law_1.2,0.21273216247558593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,2,4,power_law_1.2,0.2327244758605957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,2,4,power_law_1.2,0.24123903274536135
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,2,4,power_law_1.2,0.25644031524658206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,2,4,power_law_1.2,0.27747583389282227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,2,4,power_law_1.2,0.29126911163330077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,2,4,power_law_1.2,0.33515392303466796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,2,4,power_law_1.2,0.4461568069458008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,2,4,power_law_1.2,0.5750668716430665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,2,4,power_law_1.2,0.687768325805664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,2,4,power_law_1.2,0.8938931274414064
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,2,4,power_law_1.01,0.13170432090759276
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,2,4,power_law_1.01,0.19255552291870118
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,2,4,power_law_1.2,1.5202380371093749
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,2,4,power_law_1.2,1.4023628234863281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,2,4,power_law_1.2,2.5127565002441408
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,2,4,power_law_1.01,0.33549057006835936
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,2,4,power_law_1.01,0.2529151916503906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,2,4,power_law_1.2,0.058855681419372564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,2,4,power_law_1.2,0.06529280185699463
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,2,4,power_law_1.2,0.07404032230377197
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,2,4,power_law_1.2,0.0741260814666748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,2,4,power_law_1.2,0.0755673599243164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,2,4,power_law_1.01,0.5480780792236328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,2,4,power_law_1.2,0.08064767837524414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,2,4,power_law_1.2,0.07775360107421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,2,4,power_law_1.2,0.07916031837463379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,2,4,power_law_1.2,0.08050047874450683
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,2,4,power_law_1.2,0.07543295860290528
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,2,4,power_law_1.2,0.08731776237487793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,2,4,power_law_1.2,0.0937177562713623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,2,4,power_law_1.2,0.10002559661865233
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,2,4,power_law_1.2,0.10614656448364257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,2,4,power_law_1.2,0.11101823806762696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,2,4,power_law_1.2,0.12059647560119628
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,2,4,power_law_1.2,0.14772864341735842
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,2,4,power_law_1.2,5.667587890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,2,4,power_law_1.2,0.07189760208129883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,2,4,power_law_1.2,0.14467712402343752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,2,4,power_law_1.2,0.22756736755371093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,2,4,power_law_1.2,3.7255615234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,2,4,power_law_1.2,0.2782975959777832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,2,4,power_law_1.2,0.3713792037963867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,2,4,power_law_1.2,0.4908108901977539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,2,4,power_law_1.2,0.7260224151611328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,2,4,power_law_1.2,0.027453439235687254
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,2,4,power_law_1.2,0.03256831884384155
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,2,4,power_law_1.2,0.03083775997161865
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,2,4,power_law_1.2,0.03635967969894409
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,2,4,power_law_1.2,0.037063679695129394
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,2,4,power_law_1.2,0.03660543918609619
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,2,4,power_law_1.2,0.03794559955596924
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,2,4,power_law_1.2,0.03704576015472412
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,2,4,power_law_1.2,0.037871360778808594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,2,4,power_law_1.2,0.03714943885803222
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,2,4,power_law_1.2,0.038255360126495365
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,2,4,power_law_1.2,0.040181760787963865
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,2,4,power_law_1.2,0.041900801658630374
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,2,4,power_law_1.2,0.048312320709228515
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,2,4,power_law_1.2,0.04509696006774903
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,2,4,power_law_1.01,0.7617100524902344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,2,4,power_law_1.2,0.052547841072082525
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,2,4,power_law_1.2,1.527552032470703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,2,4,power_law_1.2,1.0622758483886718
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,2,4,power_law_1.2,0.15596927642822264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,2,4,power_law_1.2,8.18966796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,2,4,power_law_1.2,2.0781478881835938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,2,4,power_law_1.2,0.056143360137939455
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,2,4,power_law_1.2,0.07868288040161134
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,2,4,power_law_1.2,0.34944385528564453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,2,4,power_law_1.2,0.06928256034851074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,4,2,balanced,0.06001023769378662
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,4,2,balanced,0.072641282081604
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,4,2,balanced,0.09233407974243164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,4,2,balanced,0.13614208221435548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,4,2,balanced,0.13706496238708496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,4,2,balanced,0.1381939220428467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,4,2,balanced,0.1401036834716797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,4,2,balanced,0.1409177589416504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,4,2,balanced,0.1423859214782715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,4,2,balanced,0.14314751625061034
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,4,2,balanced,0.14499839782714843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,4,2,balanced,0.14615296363830566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,4,2,balanced,0.1486207962036133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,4,2,balanced,0.151975679397583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,4,2,balanced,0.20270208358764646
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,4,2,balanced,0.20756223678588864
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,4,2,balanced,0.16777984619140623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,4,2,balanced,0.23754112243652345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,4,2,balanced,0.27319807052612305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,4,2,balanced,0.37861759185791016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,4,2,balanced,0.4670502471923828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,2,4,power_law_1.01,1.2905932617187499
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,2,4,power_law_1.2,0.11111295700073243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,4,2,balanced,0.6614630126953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,4,2,balanced,0.8825049591064452
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,4,2,balanced,1.2948133850097656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,4,2,balanced,1.6978367614746095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,4,2,balanced,0.03607295989990235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,4,2,balanced,0.04508543968200683
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,4,2,balanced,0.05883008003234863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,4,2,balanced,0.06149631977081299
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,4,2,balanced,0.061736960411071774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,4,2,balanced,0.06286719799041748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,4,2,balanced,2.599942321777344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,4,2,balanced,0.06325119972229004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,4,2,balanced,0.06704512119293213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,4,2,balanced,0.06564095973968506
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,4,2,balanced,0.0676966381072998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,4,2,balanced,0.06795775890350342
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,4,2,balanced,0.07110655784606934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,4,2,balanced,0.0729318380355835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,4,2,balanced,0.0752294397354126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,4,2,balanced,0.0796608018875122
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,4,2,balanced,0.08588288307189942
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,4,2,balanced,0.07826816082000733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,4,2,balanced,0.10115327835083007
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,4,2,balanced,0.11109760284423828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,4,2,balanced,3.407729797363281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,4,2,balanced,0.14844927787780762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,4,2,balanced,0.1778495979309082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,4,2,balanced,0.25706239700317385
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,4,2,balanced,0.32391807556152347
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,4,2,balanced,0.47303550720214843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,4,2,balanced,0.5970316696166992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,4,2,balanced,0.8808934020996094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,4,2,balanced,1.153292770385742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,2,4,power_law_1.2,0.2059174346923828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,2,4,power_law_1.2,0.23937408447265626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,4,2,balanced,0.02062079906463623
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,4,2,balanced,0.034517760276794436
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,4,2,balanced,0.03453439950942993
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,4,2,balanced,0.020665600299835205
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,4,2,balanced,0.034891519546508785
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,4,2,balanced,0.034781439304351805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,4,2,balanced,0.03536767959594726
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,4,2,balanced,0.03648256063461304
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,4,2,balanced,0.03646719932556152
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,4,2,balanced,0.03695616006851196
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,4,2,balanced,0.036172800064086914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,4,2,balanced,0.037067520618438723
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,4,2,balanced,0.03839871883392334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,4,2,balanced,0.03778048038482666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,4,2,balanced,0.03901823997497559
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,2,4,power_law_1.2,0.4589913558959961
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,4,2,balanced,0.0346726393699646
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,4,2,balanced,0.03563136100769043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,2,4,power_law_1.2,0.686033935546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,4,2,balanced,0.05178880214691162
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,4,2,balanced,0.05171455860137939
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,4,2,balanced,0.06524672031402587
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,4,2,balanced,0.07746816158294678
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,4,2,power_law_1.01,0.06654463768005371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,4,2,power_law_1.01,0.07950719833374023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,4,2,power_law_1.01,0.08156800270080566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,4,2,power_law_1.01,0.1364070415496826
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,4,2,power_law_1.01,0.13623167991638183
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,4,2,power_law_1.01,0.13789952278137207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,4,2,power_law_1.01,0.141080322265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,4,2,power_law_1.01,0.1414848041534424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,4,2,power_law_1.01,0.14277631759643555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,4,2,power_law_1.01,0.14475135803222655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,4,2,power_law_1.01,0.1472640037536621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,4,2,power_law_1.01,0.14985856056213379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,4,2,power_law_1.01,0.16189056396484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,4,2,power_law_1.01,0.19947263717651367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,4,2,power_law_1.01,0.21379583358764648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,4,2,power_law_1.01,0.23236736297607422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,4,2,power_law_1.01,0.25601152420043943
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,4,2,power_law_1.01,0.3081548881530762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,4,2,power_law_1.01,0.35338367462158204
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,4,2,balanced,0.10468223571777344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,4,2,power_law_1.01,0.5015654373168945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,4,2,power_law_1.01,0.6524633789062501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,4,2,power_law_1.01,0.9310284423828126
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,4,2,balanced,0.1279206371307373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,4,2,power_law_1.01,1.2357529449462892
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,2,4,power_law_1.2,1.1203839874267578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,4,2,power_law_1.01,2.42455810546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,4,2,power_law_1.01,1.7322496032714845
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,4,2,balanced,0.17969280242919922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,4,2,power_law_1.01,0.044544000625610355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,4,2,power_law_1.01,0.05301631927490234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,4,2,power_law_1.01,0.054018559455871584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,4,2,power_law_1.01,0.06173312187194824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,4,2,power_law_1.01,0.06301695823669433
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,4,2,power_law_1.01,0.06286848068237305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,4,2,power_law_1.01,0.06442240238189698
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,4,2,power_law_1.01,0.06531583786010742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,4,2,power_law_1.01,0.06695295810699463
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,4,2,power_law_1.01,0.06793344020843506
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,4,2,power_law_1.01,0.0697331190109253
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,4,2,power_law_1.01,0.07181312084197998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,4,2,power_law_1.01,0.07872000217437744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,4,2,power_law_1.01,0.06329599857330323
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,4,2,power_law_1.01,0.09292032241821288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,4,2,power_law_1.01,0.08897791862487793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,4,2,power_law_1.01,0.1092249584197998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,4,2,power_law_1.01,0.1281446361541748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,4,2,power_law_1.01,0.0954751968383789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,4,2,balanced,0.2364761543273926
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,4,2,balanced,0.3477824020385742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,4,2,power_law_1.01,0.22474367141723633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,4,2,balanced,0.44985599517822267
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,4,2,power_law_1.01,0.17061759948730468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,4,2,power_law_1.01,0.2919923210144043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,4,2,power_law_1.01,0.02348543882369995
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,4,2,power_law_1.01,0.02422271966934204
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,4,2,power_law_1.01,0.028010239601135255
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,4,2,power_law_1.01,0.0328985595703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,4,2,power_law_1.01,0.03285504102706909
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,4,2,power_law_1.01,0.03505408048629761
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,4,2,power_law_1.01,0.03542655944824219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,4,2,power_law_1.01,0.03568639993667603
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,4,2,power_law_1.01,3.5307943725585935
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,4,2,power_law_1.01,0.035854079723358155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,4,2,power_law_1.01,0.40141056060791014
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,4,2,power_law_1.01,0.036792318820953365
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,4,2,power_law_1.01,0.0366425609588623
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,4,2,power_law_1.01,0.039211521148681636
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,4,2,power_law_1.01,0.7585433959960938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,4,2,power_law_1.01,0.5250828933715821
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,4,2,power_law_1.01,1.0011020660400392
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,4,2,power_law_1.01,4.765068664550781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,4,2,power_law_1.01,0.04115456104278564
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,4,2,power_law_1.01,0.04640255928039551
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,4,2,power_law_1.01,0.0427839994430542
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,4,2,power_law_1.01,1.3787110900878905
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,4,2,power_law_1.01,0.10387711524963379
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,4,2,power_law_1.01,0.04647039890289307
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,4,2,power_law_1.01,0.05136256217956543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,4,2,power_law_1.01,0.05648128032684326
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,4,2,power_law_1.01,0.06861824035644531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,4,2,power_law_1.2,0.06257152080535888
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,4,2,power_law_1.2,0.07917183876037598
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,4,2,power_law_1.01,0.19439359664916994
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,4,2,power_law_1.2,0.13506560325622557
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,4,2,power_law_1.2,0.1372377586364746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,4,2,power_law_1.2,0.13812992095947266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,4,2,power_law_1.01,0.09081215858459472
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,4,2,power_law_1.2,0.13987071990966798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,4,2,power_law_1.2,0.1416972827911377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,4,2,power_law_1.2,0.14417023658752443
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,4,2,power_law_1.2,0.14463232040405274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,4,2,power_law_1.2,0.14727935791015626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,4,2,power_law_1.01,0.14971136093139648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,4,2,power_law_1.2,0.1523532772064209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,4,2,power_law_1.2,0.15963520050048827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,4,2,power_law_1.2,0.22886783599853516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,4,2,power_law_1.2,0.1918681526184082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,4,2,power_law_1.2,0.21278463363647462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,4,2,power_law_1.2,0.25460607528686524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,4,2,power_law_1.2,0.34901119232177735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,4,2,power_law_1.2,0.30948863983154296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,4,2,power_law_1.2,0.0811520004272461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,4,2,power_law_1.2,0.5946419143676758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,4,2,power_law_1.2,0.5164697647094727
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,4,2,power_law_1.01,0.33794944763183593
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,4,2,power_law_1.01,0.28343423843383786
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,4,2,power_law_1.2,0.9002713775634765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,4,2,power_law_1.2,0.04358016014099121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,4,2,power_law_1.2,0.050919680595397955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,4,2,power_law_1.2,0.05591423988342285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,4,2,power_law_1.2,0.061711359024047854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,4,2,power_law_1.2,1.2663654327392577
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,4,2,power_law_1.2,0.06133120059967041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,4,2,power_law_1.2,0.06504960060119629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,4,2,power_law_1.2,0.06371327877044677
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,4,2,power_law_1.2,0.06624767780303956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,4,2,power_law_1.2,0.065032958984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,4,2,power_law_1.2,0.06683519840240479
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,4,2,power_law_1.2,0.06812928199768067
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,4,2,power_law_1.2,0.07033088207244872
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,4,2,power_law_1.2,0.07253759860992431
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,4,2,power_law_1.2,0.07927167892456055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,4,2,power_law_1.2,0.08846976280212401
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,4,2,power_law_1.2,0.09372672080993652
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,4,2,power_law_1.2,0.10707327842712402
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,4,2,power_law_1.2,0.09445631980895997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,4,2,power_law_1.2,0.13483903884887696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,4,2,power_law_1.2,0.2177561569213867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,4,2,power_law_1.2,0.1721625518798828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,4,2,power_law_1.2,0.30763519287109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,4,2,power_law_1.2,0.3933055877685547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,4,2,power_law_1.2,1.7595135498046877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,4,2,power_law_1.2,0.5566899108886718
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,4,2,power_law_1.01,0.4720947265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,4,2,power_law_1.2,2.722181091308594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,4,2,power_law_1.2,0.7416281890869141
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,4,2,power_law_1.2,0.029949440956115726
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,4,2,power_law_1.2,0.03330431938171387
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,4,2,power_law_1.2,0.03187711954116822
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,4,2,power_law_1.2,0.035025920867919925
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,4,2,power_law_1.2,0.03503488063812256
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,4,2,power_law_1.2,0.03532671928405762
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,4,2,power_law_1.2,0.035423998832702634
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,4,2,power_law_1.2,0.03584127902984619
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,4,2,power_law_1.2,0.03583872079849243
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,4,2,power_law_1.2,0.040085759162902836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,4,2,power_law_1.2,0.04111231803894043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,4,2,power_law_1.2,0.04265984058380127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,4,2,power_law_1.2,4.352574768066406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,4,2,power_law_1.2,0.04918911933898926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,4,2,power_law_1.2,1.071615982055664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,4,2,power_law_1.2,0.060208640098571777
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,4,2,power_law_1.2,3.5170816040039066
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,4,2,power_law_1.2,0.021980159282684326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,4,2,power_law_1.2,1.290081329345703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,4,2,power_law_1.2,0.026263039112091064
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,4,2,power_law_1.2,0.2073036766052246
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,4,2,power_law_1.2,0.047238402366638184
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,4,2,power_law_1.2,0.05033728122711182
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,4,2,power_law_1.2,0.07537024021148682
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,4,2,power_law_1.01,0.8908185577392578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,8,1,balanced,0.057761921882629394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,8,1,balanced,0.07287231922149659
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,8,1,balanced,0.08992575645446778
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,8,1,balanced,0.08961471557617187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,8,1,balanced,0.12056768417358399
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,8,1,balanced,0.1213107204437256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,8,1,balanced,0.12171775817871093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,8,1,balanced,0.12273792266845703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,8,1,balanced,0.1232800006866455
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,4,2,power_law_1.2,0.0982528018951416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,8,1,balanced,0.12365568161010743
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,8,1,balanced,0.12523776054382324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,8,1,balanced,0.1266489601135254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,8,1,balanced,0.1273401641845703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,8,1,balanced,0.1301459217071533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,8,1,balanced,0.1436780834197998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,8,1,balanced,0.14706111907958985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,8,1,balanced,0.15314175605773925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,8,1,balanced,0.22233983993530276
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,8,1,balanced,0.24649215698242188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,8,1,balanced,0.3447539138793945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,8,1,balanced,0.44140094757080084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,8,1,balanced,0.6130899047851562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,8,1,balanced,0.8264409637451171
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,8,1,balanced,1.2166585540771484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,8,1,balanced,1.6626803588867187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,8,1,balanced,2.518211212158203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,8,1,balanced,0.04211711883544922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,8,1,balanced,0.04754176139831543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,8,1,balanced,0.06156864166259766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,8,1,balanced,0.0611187219619751
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,4,2,power_law_1.2,0.1281113624572754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,8,1,balanced,0.06407487869262696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,8,1,balanced,0.06434815883636474
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,8,1,balanced,0.06482944011688233
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,8,1,balanced,0.06594560146331788
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,8,1,balanced,0.06655168056488037
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,8,1,balanced,0.06771967887878419
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,8,1,balanced,0.06845632076263428
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,8,1,balanced,0.06963391780853272
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,8,1,balanced,0.07025407791137696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,8,1,balanced,0.07349503993988037
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,8,1,balanced,3.339189758300781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,8,1,balanced,0.07688896179199219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,8,1,balanced,0.07822207927703857
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,8,1,balanced,0.0795417594909668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,8,1,balanced,0.10201984405517579
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,8,1,balanced,0.10902400016784668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,8,1,balanced,0.14931967735290527
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,8,1,balanced,0.17922687530517578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,8,1,balanced,0.24281856536865237
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,8,1,balanced,0.3066099166870117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,8,1,balanced,0.44445438385009767
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,8,1,balanced,0.574455680847168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,8,1,balanced,0.8500109100341797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,4,2,power_law_1.2,0.1623027229309082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,8,1,balanced,1.109881591796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,4,2,power_law_1.2,0.2947583961486816
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,4,2,power_law_1.2,0.3283417510986328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,4,2,power_law_1.2,0.6308505630493164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,8,1,balanced,0.019435520172119143
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,8,1,balanced,0.03802495956420898
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,8,1,balanced,0.021013119220733643
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,8,1,balanced,0.03824959993362427
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,8,1,balanced,0.038614399433135986
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,8,1,balanced,0.03795903921127319
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,8,1,balanced,0.036803839206695554
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,8,1,balanced,0.03858943939208984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,8,1,balanced,0.03855936050415039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,8,1,balanced,0.03888767957687378
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,8,1,balanced,0.04062975883483887
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,8,1,balanced,0.03747136116027832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,8,1,balanced,0.04060351848602295
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,8,1,balanced,0.040770559310913085
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,8,1,balanced,0.0379967999458313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,8,1,balanced,0.03773952007293701
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,8,1,balanced,0.03817728042602539
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,4,2,power_law_1.2,0.4463974380493164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,8,1,balanced,0.05356480121612549
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,8,1,balanced,0.05693312168121338
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,8,1,balanced,0.07212416172027589
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,8,1,power_law_1.01,0.05932032108306885
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,8,1,balanced,0.0839187240600586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,8,1,power_law_1.01,0.06939199924468994
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,8,1,power_law_1.01,0.07324992179870606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,8,1,power_law_1.01,0.08957375526428223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,8,1,power_law_1.01,0.1209004783630371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,8,1,power_law_1.01,0.12131391525268556
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,8,1,power_law_1.01,0.12314816474914551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,8,1,power_law_1.01,0.12228927612304688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,8,1,power_law_1.01,0.1244268798828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,8,1,power_law_1.01,0.12459263801574708
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,8,1,power_law_1.01,0.13011199951171876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,8,1,power_law_1.01,0.13063743591308594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,8,1,power_law_1.01,0.13750335693359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,8,1,power_law_1.01,0.14155839920043944
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,8,1,power_law_1.01,0.14782527923583982
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,8,1,power_law_1.01,0.17740863800048828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,8,1,power_law_1.01,0.21206272125244138
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,8,1,power_law_1.01,0.24596223831176758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,8,1,power_law_1.01,0.28996992111206055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,8,1,power_law_1.01,0.3784371185302734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,8,1,power_law_1.01,0.47026111602783205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,8,1,power_law_1.01,0.654439697265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,8,1,balanced,0.11553471565246583
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,8,1,balanced,0.14434240341186524
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,8,1,balanced,0.20682239532470703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,8,1,power_law_1.01,0.8488979339599609
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,8,1,balanced,0.27172672271728515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,8,1,power_law_1.01,0.04139776229858398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,8,1,power_law_1.01,0.045898241996765135
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,8,1,power_law_1.01,0.04963200092315674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,8,1,power_law_1.01,0.058936958312988286
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,8,1,power_law_1.01,0.06320831775665284
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,8,1,power_law_1.01,0.06379839897155762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,8,1,power_law_1.01,0.06471744060516357
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,8,1,power_law_1.01,0.06485439777374268
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,8,1,power_law_1.01,0.06582848072052003
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,8,1,power_law_1.01,0.06646207809448243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,8,1,power_law_1.01,0.06705152034759522
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,8,1,power_law_1.01,0.06774911880493165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,8,1,power_law_1.01,0.070830078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,8,1,power_law_1.01,1.2504390716552733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,8,1,power_law_1.01,0.07761792182922364
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,8,1,power_law_1.01,0.08406975746154785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,8,1,power_law_1.01,0.09669695854187012
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,8,1,power_law_1.01,0.09470720291137695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,8,1,power_law_1.01,0.1022208023071289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,8,1,power_law_1.01,0.12373248100280762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,8,1,power_law_1.01,0.1526137638092041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,8,1,power_law_1.01,0.18705087661743164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,8,1,power_law_1.01,1.6261555480957033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,8,1,power_law_1.01,0.2520102310180664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,8,1,balanced,0.40411518096923826
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,8,1,balanced,0.5324019241333008
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,8,1,power_law_1.01,0.02354111909866333
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,8,1,power_law_1.01,0.02512320041656494
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,8,1,power_law_1.01,0.026837759017944336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,8,1,power_law_1.01,0.035959041118621825
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,8,1,power_law_1.01,0.0376473593711853
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,8,1,power_law_1.01,0.320060806274414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,8,1,power_law_1.01,0.03786240100860595
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,8,1,power_law_1.01,0.03854527950286865
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,8,1,power_law_1.01,0.03759360074996948
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,8,1,power_law_1.01,0.037683839797973635
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,8,1,power_law_1.01,0.03821696043014526
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,8,1,power_law_1.01,0.03895936012268066
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,8,1,power_law_1.01,0.038360960483551025
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,8,1,power_law_1.01,0.040770559310913085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,8,1,power_law_1.01,0.4508544158935547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,8,1,power_law_1.01,2.4824319458007813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,8,1,power_law_1.01,0.5880928039550781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,8,1,power_law_1.01,0.058490881919860835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,8,1,power_law_1.01,0.8539520263671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,8,1,power_law_1.01,3.235537414550781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,8,1,power_law_1.01,1.130426254272461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,8,1,power_law_1.01,0.039622399806976315
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,8,1,power_law_1.01,0.042847361564636234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,8,1,power_law_1.01,0.04596159934997558
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,8,1,power_law_1.01,0.13999103546142577
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,8,1,power_law_1.01,0.04939263820648193
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,8,1,power_law_1.01,0.06707839965820313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,8,1,power_law_1.2,0.058697600364685056
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,8,1,power_law_1.2,0.06855999946594238
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,8,1,power_law_1.2,0.07053823947906494
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,8,1,power_law_1.2,0.08928256034851074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,8,1,power_law_1.2,0.11836671829223633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,8,1,power_law_1.2,0.1207526397705078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,8,1,power_law_1.2,0.12175616264343261
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,8,1,power_law_1.2,0.12204992294311523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,8,1,power_law_1.2,0.12332096099853515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,8,1,power_law_1.2,0.12459712028503418
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,8,1,power_law_1.01,0.08586175918579102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,8,1,power_law_1.2,0.12900544166564942
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,8,1,power_law_1.2,0.13080639839172364
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,8,1,power_law_1.01,0.10325823783874513
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,8,1,power_law_1.2,0.13631679534912108
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,8,1,power_law_1.2,0.14217408180236818
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,8,1,power_law_1.2,0.1480735969543457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,8,1,power_law_1.2,0.18602304458618163
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,8,1,power_law_1.2,0.2126323127746582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,8,1,power_law_1.2,0.24538303375244142
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,8,1,power_law_1.2,0.29351743698120114
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,8,1,power_law_1.2,0.3829683303833008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,8,1,power_law_1.2,0.4695315170288086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,8,1,power_law_1.2,0.6551289367675781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,8,1,power_law_1.01,0.17752128601074219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,8,1,power_law_1.2,0.8382272338867187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,8,1,power_law_1.01,0.25388736724853517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,8,1,power_law_1.2,0.04274496078491211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,8,1,power_law_1.2,0.04878975868225098
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,8,1,power_law_1.2,0.04940671920776367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,8,1,power_law_1.2,0.06056320190429687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,8,1,power_law_1.2,0.06258495807647704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,8,1,power_law_1.2,0.0643564796447754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,8,1,power_law_1.2,0.06526527881622315
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,8,1,power_law_1.2,0.06514048099517822
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,8,1,power_law_1.2,0.06652544021606446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,8,1,power_law_1.2,0.06750656127929687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,8,1,power_law_1.2,0.0682470417022705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,8,1,power_law_1.2,0.06960447788238525
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,8,1,power_law_1.2,0.07073472023010255
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,8,1,power_law_1.2,0.07845568180084228
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,8,1,power_law_1.01,0.33055934906005857
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,8,1,power_law_1.2,0.08678208351135254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,8,1,power_law_1.2,0.09827839851379394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,8,1,power_law_1.2,0.09587648391723633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,8,1,power_law_1.2,0.10209407806396484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,8,1,power_law_1.2,1.241624298095703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,8,1,power_law_1.2,0.12295743942260742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,8,1,power_law_1.2,0.15081536293029785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,8,1,power_law_1.2,0.1865452766418457
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,8,1,power_law_1.01,0.48519550323486327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,8,1,power_law_1.2,0.2510848045349121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,8,1,power_law_1.2,1.6685548400878907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,8,1,power_law_1.2,0.32060287475585936
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,8,1,power_law_1.2,0.01975167989730835
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,8,1,power_law_1.2,0.024113919734954834
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,8,1,power_law_1.2,0.02470720052719116
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,8,1,power_law_1.2,0.036871039867401124
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,8,1,power_law_1.2,0.03606591939926147
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,8,1,power_law_1.01,0.6391020965576172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,8,1,power_law_1.2,0.03823487997055054
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,8,1,power_law_1.2,0.03947263956069946
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,8,1,power_law_1.2,0.037405440807342524
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,8,1,power_law_1.2,0.038514559268951414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,8,1,power_law_1.2,0.03859391927719116
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,8,1,power_law_1.2,0.03925823926925659
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,8,1,power_law_1.2,0.45164863586425785
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,8,1,power_law_1.2,0.040814080238342286
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,8,1,power_law_1.2,0.04357952117919922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,8,1,power_law_1.2,0.047031679153442384
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,8,1,power_law_1.2,0.5875443267822266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,8,1,power_law_1.2,2.457941131591797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,8,1,power_law_1.2,0.03945535898208618
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,8,1,power_law_1.2,0.039047040939331056
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,8,1,power_law_1.2,0.8534003448486327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,8,1,power_law_1.2,3.2307897949218747
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,8,1,power_law_1.2,0.049685120582580566
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,8,1,power_law_1.2,0.059250559806823734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,8,1,power_law_1.2,0.06826752185821533
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,8,1,power_law_1.2,0.2624064064025879
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,8,1,power_law_1.2,0.08698495864868164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,8,1,power_law_1.2,1.1246649932861328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,1,8,balanced,0.23705984115600587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,1,8,balanced,0.23805183410644531
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,8,1,power_law_1.2,0.10490943908691405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,1,8,balanced,0.3613862228393555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,1,8,balanced,0.24049280166625975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,1,8,balanced,0.3576243209838867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,1,8,balanced,0.35610881805419925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,1,8,balanced,0.35543678283691404
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,1,8,balanced,0.35688831329345705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,1,8,balanced,0.35958526611328123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,1,8,balanced,0.3569599914550781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,1,8,balanced,0.3727142333984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,1,8,balanced,0.3869670486450195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,1,8,balanced,0.395733757019043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,1,8,balanced,0.43740287780761716
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,1,8,balanced,0.4415667343139648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,1,8,balanced,0.4568819046020508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,1,8,balanced,0.4862835311889649
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,1,8,balanced,0.5313600158691406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,1,8,balanced,0.35550846099853517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,1,8,balanced,0.7441395568847656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,1,8,balanced,0.7301964569091797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,1,8,balanced,1.1787213134765626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,1,8,balanced,1.4586073303222657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,1,8,balanced,2.169100799560547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,8,1,power_law_1.2,0.1453286361694336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,1,8,balanced,2.900695190429688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,1,8,balanced,0.1106175994873047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,1,8,balanced,0.11603072166442871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,1,8,balanced,0.11044863700866699
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,1,8,balanced,0.1263475227355957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,1,8,balanced,0.12220543861389159
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,1,8,balanced,0.124966402053833
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,1,8,balanced,0.12437248229980469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,1,8,balanced,0.1248908805847168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,1,8,balanced,0.12470656394958497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,1,8,balanced,0.12549504280090332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,1,8,balanced,0.12552703857421874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,1,8,balanced,0.1271027183532715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,1,8,balanced,0.12728704452514647
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,1,8,balanced,0.12927103996276856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,1,8,balanced,0.13280256271362306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,1,8,balanced,0.13420031547546388
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,1,8,balanced,4.747032470703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,1,8,balanced,0.13671936035156249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,1,8,balanced,0.16680320739746093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,1,8,balanced,0.18963071823120117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,1,8,balanced,0.20871423721313476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,1,8,balanced,0.23052671432495114
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,1,8,balanced,0.30635135650634765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,1,8,balanced,0.39653247833251953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,1,8,balanced,0.5221478271484374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,1,8,balanced,0.6980159759521485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,1,8,balanced,1.0941632080078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,8,1,power_law_1.2,0.1835641670227051
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,1,8,balanced,6.224782104492188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,1,8,balanced,1.4395123291015626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,8,1,power_law_1.2,0.34100990295410155
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,8,1,power_law_1.2,0.4991814422607422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,1,8,balanced,0.055438079833984376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,1,8,balanced,0.05591807842254639
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,1,8,balanced,0.055532798767089844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,1,8,balanced,0.05638400077819824
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,1,8,balanced,0.055822081565856926
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,1,8,balanced,0.05580031871795654
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,1,8,balanced,0.05656064033508301
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,1,8,balanced,0.05688191890716553
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,1,8,balanced,0.055925760269165035
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,1,8,balanced,0.05540095806121827
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,1,8,balanced,0.05638527870178223
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,1,8,balanced,0.0573145580291748
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,1,8,balanced,0.057377281188964846
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,8,1,power_law_1.2,0.6603750610351563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,1,8,balanced,0.058375678062438964
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,1,8,balanced,0.05761919975280762
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,1,8,balanced,0.055870718955993645
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,1,8,balanced,0.056377601623535153
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,1,8,balanced,0.07630720138549804
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,1,8,balanced,0.07836415767669677
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,1,8,balanced,0.09867136001586914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,1,8,balanced,0.12419967651367188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,1,8,power_law_1.01,0.24227327346801758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,1,8,power_law_1.01,0.3570854568481445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,1,8,power_law_1.01,0.35324031829833985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,1,8,power_law_1.01,0.3548479843139648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,1,8,power_law_1.01,0.3535615921020508
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,1,8,balanced,0.16772991180419922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,1,8,power_law_1.01,0.3558310317993164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,1,8,power_law_1.01,0.3634483337402344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,1,8,power_law_1.01,0.37657215118408205
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,1,8,balanced,0.21980031967163086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,1,8,power_law_1.01,0.41220737457275386
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,1,8,power_law_1.01,0.4240959930419922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,1,8,power_law_1.01,0.4416665649414062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,1,8,power_law_1.01,0.4711731338500977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,1,8,power_law_1.01,0.4986726379394531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,1,8,power_law_1.01,0.5840691375732422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,1,8,power_law_1.01,0.6768959808349609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,1,8,power_law_1.01,0.7125721740722656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,1,8,power_law_1.01,1.090767364501953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,1,8,power_law_1.01,1.4191821289062498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,1,8,power_law_1.01,2.0930546569824218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,1,8,power_law_1.01,0.45586303710937504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,1,8,power_law_1.01,2.8588467407226563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,1,8,power_law_1.01,4.390939025878906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,1,8,power_law_1.01,6.305510253906251
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,1,8,balanced,0.31909248352050784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,1,8,power_law_1.01,0.10963968276977538
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,1,8,power_law_1.01,0.12078335762023926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,1,8,power_law_1.01,0.12294400215148926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,1,8,power_law_1.01,0.12352000236511232
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,1,8,power_law_1.01,0.12418687820434571
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,1,8,power_law_1.01,9.379349975585937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,1,8,power_law_1.01,0.12534015655517577
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,1,8,power_law_1.01,0.1296345615386963
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,1,8,power_law_1.01,0.12760704040527343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,1,8,power_law_1.01,0.1319155216217041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,1,8,power_law_1.01,0.13114239692687987
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,1,8,power_law_1.01,0.13548928260803222
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,1,8,power_law_1.01,0.15146368026733398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,1,8,power_law_1.01,0.15445759773254394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,1,8,power_law_1.01,0.16973567962646485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,1,8,power_law_1.01,0.18816896438598635
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,1,8,power_law_1.01,0.1994175910949707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,1,8,power_law_1.01,0.22579456329345704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,1,8,power_law_1.01,0.28316287994384765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,1,8,power_law_1.01,0.39237632751464846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,1,8,power_law_1.01,0.5210854339599609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,1,8,power_law_1.01,14.375144042968751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,1,8,power_law_1.01,0.6769062042236328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,1,8,power_law_1.01,1.4604786682128907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,1,8,power_law_1.01,1.07359619140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,1,8,balanced,0.4118080139160156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,1,8,power_law_1.01,2.162670135498047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,1,8,power_law_1.01,0.06147583961486817
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,1,8,power_law_1.01,0.057451519966125485
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,1,8,power_law_1.01,0.057543678283691405
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,1,8,power_law_1.01,0.05646080017089844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,1,8,power_law_1.01,0.056541438102722164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,1,8,power_law_1.01,0.05625216007232666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,1,8,power_law_1.01,0.056961278915405276
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,1,8,power_law_1.01,0.05679232120513916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,1,8,power_law_1.01,0.05734272003173828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,1,8,power_law_1.01,0.05786240100860596
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,1,8,power_law_1.01,0.058620162010192864
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,1,8,power_law_1.01,2.914547119140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,1,8,power_law_1.01,5.527077026367188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,1,8,power_law_1.01,32.37994140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,1,8,balanced,0.5934592056274414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,1,8,balanced,0.7791718292236328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,1,8,power_law_1.01,23.53064697265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,1,8,power_law_1.01,4.345689697265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,1,8,power_law_1.01,0.07266304016113281
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,1,8,power_law_1.01,0.06951168060302734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,1,8,power_law_1.01,0.07661439895629882
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,1,8,power_law_1.01,0.0894591999053955
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,1,8,power_law_1.01,0.09301119804382324
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,1,8,power_law_1.01,0.12305664062499999
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,1,8,power_law_1.01,0.1589299201965332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,1,8,power_law_1.01,0.22127872467041015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,1,8,power_law_1.01,0.3042611122131348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,1,8,power_law_1.2,0.2354150390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,1,8,power_law_1.2,0.3593510437011719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,1,8,power_law_1.2,0.3548774337768555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,1,8,power_law_1.2,0.3567500686645508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,1,8,power_law_1.2,0.35587841033935547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,1,8,power_law_1.2,0.3578291320800781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,1,8,power_law_1.2,0.3679692840576172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,1,8,power_law_1.2,0.3863923263549805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,1,8,power_law_1.2,0.4210022354125977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,1,8,power_law_1.2,0.42580734252929686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,1,8,power_law_1.2,0.44500991821289065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,1,8,power_law_1.2,0.46048126220703123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,1,8,power_law_1.2,0.47383808135986333
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,1,8,power_law_1.01,0.40285568237304686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,1,8,power_law_1.2,0.505027847290039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,1,8,power_law_1.2,0.6364198303222656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,1,8,power_law_1.2,0.6667826843261719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,1,8,power_law_1.2,0.6836825561523437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,1,8,power_law_1.2,1.0662220764160157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,1,8,power_law_1.2,1.426044158935547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,1,8,power_law_1.2,2.113095703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,1,8,power_law_1.01,0.6927782440185547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,1,8,power_law_1.01,0.5749772644042969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,1,8,power_law_1.2,2.874750671386719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,1,8,power_law_1.2,4.412578430175781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,1,8,power_law_1.01,1.2076544189453124
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,1,8,power_law_1.01,1.59425537109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,1,8,power_law_1.2,6.218968505859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,1,8,power_law_1.2,0.1298585605621338
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,1,8,power_law_1.2,0.13117312431335448
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,1,8,power_law_1.2,0.150577917098999
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,1,8,power_law_1.2,0.1259225559234619
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,1,8,power_law_1.2,0.1495795154571533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,1,8,power_law_1.2,0.1278771209716797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,1,8,power_law_1.2,0.12652799606323242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,1,8,power_law_1.2,0.12863871574401856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,1,8,power_law_1.2,0.1300236797332764
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,1,8,power_law_1.2,0.13182463645935058
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,1,8,power_law_1.2,0.13552255630493165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,1,8,power_law_1.2,0.15174528121948244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,1,8,power_law_1.2,0.15541248321533202
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,1,8,power_law_1.2,0.16946304321289063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,1,8,power_law_1.2,0.18920064926147462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,1,8,power_law_1.2,0.19982847213745117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,1,8,power_law_1.2,0.22100095748901366
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,1,8,power_law_1.2,0.2860428810119629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,1,8,power_law_1.2,0.48266750335693354
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,1,8,power_law_1.2,0.3961548614501953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,1,8,power_law_1.2,9.771889648437499
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,1,8,power_law_1.01,3.0678924560546874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,1,8,power_law_1.2,0.6711462402343751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,1,8,power_law_1.2,14.811788330078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,1,8,power_law_1.2,1.0675775909423828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,1,8,power_law_1.2,1.4253683471679688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,1,8,power_law_1.01,2.476660461425781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,1,8,power_law_1.2,0.05955455780029297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,1,8,power_law_1.2,0.057423357963562016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,1,8,power_law_1.2,0.056619520187377925
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,1,8,power_law_1.2,0.05601920127868652
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,1,8,power_law_1.2,0.056112642288208
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,1,8,power_law_1.2,0.05618048191070557
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,1,8,power_law_1.2,0.05669248104095459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,1,8,power_law_1.2,1.9661439514160157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,1,8,power_law_1.2,0.05718016147613525
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,1,8,power_law_1.2,0.06135807991027832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,1,8,power_law_1.2,0.05824512004852295
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,1,8,power_law_1.2,0.07406720161437988
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,1,8,power_law_1.2,0.07800576210021973
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,1,8,power_law_1.2,22.76577392578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,1,8,power_law_1.2,4.224909973144531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,1,8,power_law_1.2,2.8699545288085937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,1,8,power_law_1.2,5.90874755859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,1,8,power_law_1.2,32.69433837890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,1,8,power_law_1.2,0.055696640014648434
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,1,8,power_law_1.2,0.09239295959472657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,1,8,power_law_1.2,0.07223040103912354
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,1,8,power_law_1.2,0.5627699279785157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,1,8,power_law_1.2,0.09370623588562012
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,1,8,power_law_1.2,0.11374079704284668
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,1,8,power_law_1.2,0.16255231857299807
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,1,8,power_law_1.2,0.21929088592529297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,2,4,balanced,0.12270463943481444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,2,4,balanced,0.12261887550354005
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,2,4,balanced,0.15328384399414063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,2,4,balanced,0.23647232055664064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,2,4,balanced,0.2372812843322754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,2,4,balanced,0.2374655914306641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,2,4,balanced,0.23855743408203126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,2,4,balanced,0.23880319595336913
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,2,4,balanced,0.24197887420654296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,2,4,balanced,0.24318208694458007
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,2,4,balanced,0.2446579170227051
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,2,4,balanced,0.2559424018859863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,2,4,balanced,0.26234367370605466
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,2,4,balanced,0.29369855880737306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,2,4,balanced,0.3319462585449219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,1,8,power_law_1.2,0.3084582328796387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,2,4,balanced,0.33650688171386717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,2,4,balanced,0.34789760589599605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,2,4,balanced,0.4316428756713867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,2,4,balanced,0.4019417572021484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,2,4,balanced,0.6502310180664063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,2,4,balanced,0.742892837524414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,2,4,balanced,1.116826858520508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,2,4,balanced,1.4404301452636719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,2,4,balanced,2.1759283447265623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,2,4,balanced,2.9382540893554685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,2,4,balanced,4.411124572753907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,2,4,balanced,0.07792768001556397
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,2,4,balanced,0.07463168144226075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,2,4,balanced,0.09013504028320313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,2,4,balanced,0.09497471809387206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,2,4,balanced,0.09454976081848145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,2,4,balanced,0.09509632110595703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,2,4,balanced,0.0963584041595459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,2,4,balanced,0.09668352127075194
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,2,4,balanced,0.0964902400970459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,2,4,balanced,0.09727359771728515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,2,4,balanced,0.09890175819396972
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,2,4,balanced,0.10143232345581055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,2,4,balanced,6.294891357421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,2,4,balanced,0.10290047645568848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,2,4,balanced,0.10905599594116211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,2,4,balanced,0.10801919937133789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,2,4,balanced,0.11389951705932617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,2,4,balanced,0.11951744079589843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,2,4,balanced,0.15109888076782227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,2,4,balanced,0.14840576171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,2,4,balanced,0.1868275260925293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,2,4,balanced,0.2367385673522949
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,2,4,balanced,0.3114956855773926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,2,4,balanced,0.39087615966796874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,2,4,balanced,0.5835353469848632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,2,4,balanced,0.7618022155761719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,2,4,balanced,1.153987808227539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,2,4,balanced,1.5319552612304688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,1,8,power_law_1.2,0.41169921875000004
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,1,8,power_law_1.2,0.7274419403076171
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,1,8,power_law_1.2,1.0842406463623047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,1,8,power_law_1.2,1.5837887573242189
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,2,4,balanced,0.02971519947052002
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,2,4,balanced,0.04739200115203858
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,2,4,balanced,0.047202558517456056
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,2,4,balanced,0.028849918842315674
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,2,4,balanced,0.04747519969940185
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,2,4,balanced,0.04799615859985352
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,2,4,balanced,0.047802882194519045
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,2,4,balanced,0.04854144096374512
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,2,4,balanced,0.047815680503845215
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,2,4,balanced,0.04861951828002929
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,1,8,power_law_1.2,2.2740953063964846
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,2,4,balanced,0.04901375770568848
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,2,4,balanced,0.049363198280334475
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,2,4,balanced,0.04665472030639649
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,2,4,balanced,0.04607999801635742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,1,8,power_law_1.2,3.2046810913085935
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,2,4,balanced,0.04677504062652588
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,2,4,balanced,0.04691840171813964
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,2,4,balanced,0.06893824100494385
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,2,4,balanced,0.05011072158813477
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,2,4,balanced,0.07157504081726074
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,2,4,balanced,0.0879270362854004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,2,4,power_law_1.01,0.12759296417236327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,2,4,power_law_1.01,0.13593600273132325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,2,4,power_law_1.01,0.22462848663330076
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,2,4,power_law_1.01,0.23709184646606446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,2,4,power_law_1.01,0.23449344635009767
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,2,4,power_law_1.01,0.23851648330688477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,2,4,power_law_1.01,0.23842687606811525
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,2,4,power_law_1.01,0.24550144195556642
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,2,4,power_law_1.01,0.2727590370178222
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,2,4,balanced,0.10809087753295898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,2,4,power_law_1.01,0.2786393547058105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,2,4,power_law_1.01,0.2920281600952149
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,2,4,power_law_1.01,0.3260889434814453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,2,4,power_law_1.01,0.3329625701904297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,2,4,power_law_1.01,0.35383422851562496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,2,4,power_law_1.01,0.40217601776123046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,2,4,power_law_1.01,0.43722366333007817
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,2,4,power_law_1.01,0.48607616424560546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,2,4,power_law_1.01,0.6785779571533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,2,4,power_law_1.01,0.8930111694335938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,2,4,power_law_1.01,1.074466552734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,2,4,balanced,0.14740096092224123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,2,4,power_law_1.01,1.5321536254882813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,2,4,power_law_1.01,2.377091827392578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,2,4,balanced,0.19376384735107421
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,2,4,power_law_1.01,3.210928649902344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,2,4,power_law_1.01,4.648398132324219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,2,4,power_law_1.01,0.07942783832550049
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,2,4,power_law_1.01,0.07667840003967286
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,2,4,power_law_1.01,0.09200256347656251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,2,4,power_law_1.01,0.09423359870910644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,2,4,power_law_1.01,0.09435775756835937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,2,4,balanced,0.27716096878051755
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,2,4,power_law_1.01,0.09568384170532226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,2,4,power_law_1.01,0.09710463523864746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,2,4,power_law_1.01,0.09780863761901856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,2,4,power_law_1.01,0.10004991531372069
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,2,4,power_law_1.01,0.09998720169067384
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,2,4,power_law_1.01,0.10390527725219727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,2,4,power_law_1.01,0.11416576385498048
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,2,4,power_law_1.01,0.11736703872680665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,2,4,power_law_1.01,0.12848896026611328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,2,4,power_law_1.01,0.13661439895629884
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,2,4,power_law_1.01,0.14190208435058593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,2,4,power_law_1.01,0.16319360733032226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,2,4,power_law_1.01,6.696446533203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,2,4,power_law_1.01,0.20887680053710939
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,2,4,power_law_1.01,0.25155967712402344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,2,4,power_law_1.01,0.3372518539428711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,2,4,power_law_1.01,0.4343142318725586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,2,4,power_law_1.01,0.8138944244384765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,2,4,power_law_1.01,0.690292510986328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,2,4,balanced,0.3666764831542969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,2,4,balanced,0.7100697326660156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,2,4,balanced,0.5355750274658203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,2,4,power_law_1.01,0.03792511940002442
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,2,4,power_law_1.01,0.04266111850738526
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,2,4,power_law_1.01,0.04248447895050049
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,2,4,power_law_1.01,0.04574592113494873
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,2,4,power_law_1.01,0.04711679935455322
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,2,4,power_law_1.01,0.04730239868164062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,2,4,power_law_1.01,0.04716032028198242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,2,4,power_law_1.01,0.0473638391494751
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,2,4,power_law_1.01,0.04748672008514404
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,2,4,power_law_1.01,0.04804992198944092
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,2,4,power_law_1.01,0.04773375988006592
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,2,4,power_law_1.01,0.05099775791168213
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,2,4,power_law_1.01,0.05784448146820068
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,2,4,power_law_1.01,14.799589843749999
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,2,4,power_law_1.01,1.3654130554199218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,2,4,power_law_1.01,11.913475341796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,2,4,power_law_1.01,1.6386227416992187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,2,4,power_law_1.01,3.4521319580078127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,2,4,power_law_1.01,2.751170654296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,2,4,power_law_1.01,0.05074687957763672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,2,4,power_law_1.01,0.0683673620223999
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,2,4,power_law_1.01,0.0697702407836914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,2,4,power_law_1.01,0.07998720169067383
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,2,4,power_law_1.01,0.32119422912597656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,2,4,power_law_1.01,0.40481407165527344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,2,4,power_law_1.01,0.1045145606994629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,2,4,power_law_1.2,0.12174719810485839
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,2,4,power_law_1.2,0.13652992248535156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,2,4,power_law_1.2,0.22722816467285156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,2,4,power_law_1.2,0.23363584518432617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,2,4,power_law_1.2,0.23367551803588865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,2,4,power_law_1.2,0.2380735969543457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,2,4,power_law_1.2,0.24293504714965822
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,2,4,power_law_1.2,0.24605823516845704
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,2,4,power_law_1.01,0.12391679763793946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,2,4,power_law_1.2,0.27978368759155275
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,2,4,power_law_1.2,0.28016384124755855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,2,4,power_law_1.2,0.2988211250305176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,2,4,power_law_1.2,0.3271923065185547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,2,4,power_law_1.2,0.33501056671142576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,2,4,power_law_1.2,0.3538739013671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,2,4,power_law_1.2,0.38201087951660156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,2,4,power_law_1.2,0.43968639373779295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,2,4,power_law_1.2,0.49460735321044924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,2,4,power_law_1.2,0.6762560272216798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,2,4,power_law_1.2,0.8850252532958984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,2,4,power_law_1.2,1.1247769927978515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,2,4,power_law_1.2,1.5360153198242188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,2,4,power_law_1.2,2.300184326171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,2,4,power_law_1.01,0.17634559631347657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,2,4,power_law_1.01,0.2293350410461426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,2,4,power_law_1.2,3.077368469238281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,2,4,power_law_1.2,5.7935766601562495
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,2,4,power_law_1.2,0.08338560104370117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,2,4,power_law_1.2,0.07746816158294678
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,2,4,power_law_1.2,0.09170304298400879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,2,4,power_law_1.2,0.09504768371582031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,2,4,power_law_1.2,0.09535743713378905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,2,4,power_law_1.2,0.09658623695373535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,2,4,power_law_1.2,0.09795455932617188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,2,4,power_law_1.2,0.09881343841552734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,2,4,power_law_1.2,0.10117376327514649
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,2,4,power_law_1.2,0.1014425563812256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,2,4,power_law_1.2,0.10448639869689942
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,2,4,power_law_1.2,0.1149721622467041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,2,4,power_law_1.2,0.11742464065551758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,2,4,power_law_1.2,0.12924160003662108
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,2,4,power_law_1.2,7.58
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,2,4,power_law_1.2,0.13734272003173828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,2,4,power_law_1.2,0.14599424362182617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,2,4,power_law_1.2,0.15895423889160157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,2,4,power_law_1.2,0.1976473617553711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,2,4,power_law_1.01,1.2512985229492188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,2,4,power_law_1.2,0.2425984001159668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,2,4,power_law_1.2,0.33429119110107425
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,2,4,power_law_1.01,0.6811456298828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,2,4,power_law_1.2,0.410145263671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,2,4,power_law_1.2,0.8237145233154297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,2,4,power_law_1.01,0.7824614715576172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,2,4,power_law_1.2,0.6118745422363281
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,2,4,power_law_1.2,0.03285120010375976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,2,4,power_law_1.2,11.97279541015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,2,4,power_law_1.2,0.04461184024810791
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,2,4,power_law_1.2,0.03836031913757324
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,2,4,power_law_1.2,0.04692351818084717
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,2,4,power_law_1.2,0.04766848087310791
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,2,4,power_law_1.2,0.04701183795928955
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,2,4,power_law_1.2,0.04724607944488525
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,2,4,power_law_1.2,0.0471065616607666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,2,4,power_law_1.2,0.04735743999481201
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,2,4,power_law_1.2,0.047726078033447264
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,2,4,power_law_1.2,0.04812543869018555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,2,4,power_law_1.2,15.68714599609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,2,4,power_law_1.2,1.2830450439453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,2,4,power_law_1.2,0.06909056186676026
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,2,4,power_law_1.2,0.07066495895385742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,2,4,power_law_1.2,3.5485375976562503
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,2,4,power_law_1.01,1.83897216796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,2,4,power_law_1.2,1.5335769653320312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,2,4,power_law_1.2,2.8077133178710936
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,2,4,power_law_1.2,0.21799039840698242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,2,4,power_law_1.2,0.05113855838775635
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,2,4,power_law_1.2,0.05508607864379883
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,2,4,power_law_1.2,0.0580620813369751
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,2,4,power_law_1.2,0.07948800086975097
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,2,4,power_law_1.2,0.09739520072937011
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,2,4,power_law_1.2,0.12290176391601562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,4,2,balanced,0.08320384025573731
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,4,2,balanced,0.09775103569030762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,4,2,balanced,0.13045632362365722
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,4,2,balanced,0.18449920654296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,4,2,balanced,0.18545663833618165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,4,2,balanced,0.18576383590698242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,4,2,balanced,0.18944768905639647
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,4,2,balanced,0.1907916831970215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,4,2,balanced,0.19224960327148438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,4,2,balanced,0.19394559860229493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,4,2,balanced,0.19484159469604492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,4,2,balanced,0.1974220848083496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,4,2,balanced,0.20168447494506836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,4,2,balanced,0.21505151748657228
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,4,2,balanced,0.3048294448852539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,4,2,balanced,0.31294464111328124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,4,2,balanced,0.2618099212646484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,4,2,balanced,0.37034496307373044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,4,2,balanced,0.414911994934082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,4,2,balanced,0.6019007873535156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,4,2,balanced,0.7429094696044922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,2,4,power_law_1.2,0.17205759048461916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,4,2,balanced,1.1323661041259765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,4,2,balanced,1.4570777893066407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,4,2,balanced,2.243777313232422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,4,2,balanced,2.9251083374023437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,4,2,balanced,4.427933349609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,4,2,balanced,0.04488063812255859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,4,2,balanced,0.06185215950012207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,4,2,balanced,0.07981567859649659
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,4,2,balanced,0.08327679634094239
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,4,2,balanced,0.08328191757202148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,4,2,balanced,0.0841100788116455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,4,2,balanced,0.08539263725280762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,4,2,balanced,0.08697728157043458
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,4,2,balanced,0.08672256469726562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,4,2,balanced,0.08879615783691405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,4,2,balanced,0.0898739242553711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,4,2,balanced,0.09258239746093751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,4,2,balanced,0.0952575969696045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,4,2,balanced,6.3142810058593755
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,4,2,balanced,0.10047231674194337
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,4,2,balanced,0.1093017578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,4,2,balanced,0.1105228805541992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,4,2,balanced,0.10828672409057619
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,4,2,balanced,0.14515583992004394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,4,2,balanced,0.15842687606811523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,4,2,balanced,0.20425727844238278
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,4,2,balanced,0.2529395294189453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,4,2,balanced,0.35655166625976564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,4,2,balanced,0.45707134246826175
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,4,2,balanced,0.6619251251220704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,4,2,balanced,0.9078886413574219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,4,2,balanced,1.320015411376953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,2,4,power_law_1.2,0.33323390960693355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,4,2,balanced,1.7393778991699218
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,2,4,power_law_1.2,0.44966911315917973
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,2,4,power_law_1.2,0.7208191680908203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,2,4,power_law_1.2,1.1643968200683594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,4,2,balanced,0.02252415895462036
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,4,2,balanced,0.02641151905059814
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,4,2,balanced,0.05042943954467773
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,4,2,balanced,0.05099008083343506
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,4,2,balanced,0.05107840061187744
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,4,2,balanced,0.05164415836334228
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,4,2,balanced,0.05081471920013427
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,4,2,balanced,0.051560959815979
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,4,2,balanced,0.05147264003753662
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,4,2,balanced,0.05167744159698486
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,4,2,balanced,0.052468481063842776
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,4,2,balanced,0.05302271842956543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,4,2,balanced,0.0497651195526123
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,4,2,balanced,0.05445759773254395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,4,2,balanced,0.04927743911743164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,2,4,power_law_1.2,1.1938111877441406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,4,2,balanced,0.04988800048828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,2,4,power_law_1.2,1.6661477661132813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,4,2,balanced,0.050173439979553223
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,4,2,balanced,0.06943359851837158
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,4,2,balanced,0.08902144432067871
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,4,2,balanced,0.0713804817199707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,4,2,power_law_1.01,0.09082367897033691
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,4,2,power_law_1.01,0.0950592041015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,4,2,power_law_1.01,0.10484352111816406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,4,2,power_law_1.01,0.18421760559082032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,4,2,power_law_1.01,0.18379903793334962
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,4,2,power_law_1.01,0.1861030387878418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,4,2,power_law_1.01,0.18756479263305664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,4,2,power_law_1.01,0.19123327255249023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,4,2,power_law_1.01,0.19448575973510743
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,4,2,power_law_1.01,0.19610111236572264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,4,2,power_law_1.01,0.20043008804321288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,4,2,power_law_1.01,0.2218931198120117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,4,2,power_law_1.01,0.24296064376831056
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,4,2,power_law_1.01,0.27003391265869137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,4,2,power_law_1.01,0.33201663970947265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,4,2,power_law_1.01,0.3468377685546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,4,2,power_law_1.01,0.3816896057128906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,4,2,power_law_1.01,0.48088832855224606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,4,2,power_law_1.01,0.5817817687988281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,4,2,power_law_1.01,0.8219213104248047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,4,2,balanced,0.11097087860107421
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,4,2,power_law_1.01,1.4086912536621095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,4,2,power_law_1.01,0.8926630401611328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,4,2,power_law_1.01,2.9107699584960938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,4,2,power_law_1.01,1.954613800048828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,4,2,balanced,0.15064831733703615
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,4,2,balanced,0.19560447692871094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,4,2,power_law_1.01,0.05524223804473877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,4,2,power_law_1.01,0.06336128234863281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,4,2,power_law_1.01,0.07187839984893798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,4,2,power_law_1.01,0.08184191703796387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,4,2,power_law_1.01,0.08291584014892578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,4,2,power_law_1.01,0.08283519744873047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,4,2,power_law_1.01,0.08461567878723145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,4,2,power_law_1.01,0.08502911567687989
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,4,2,power_law_1.01,0.08669952392578126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,4,2,power_law_1.01,0.0869222354888916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,4,2,power_law_1.01,0.08950528144836425
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,4,2,power_law_1.01,0.10195455551147461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,4,2,power_law_1.01,0.10381695747375488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,4,2,power_law_1.01,0.11156096458435058
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,4,2,power_law_1.01,0.12057472229003907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,4,2,power_law_1.01,0.12713855743408203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,4,2,power_law_1.01,0.1310758399963379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,4,2,power_law_1.01,4.128569641113281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,4,2,power_law_1.01,0.16309759140014649
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,4,2,balanced,0.2804377555847168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,4,2,power_law_1.01,0.19198335647583006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,4,2,power_law_1.01,0.30035200119018557
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,4,2,power_law_1.01,0.24735488891601562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,4,2,balanced,0.36584705352783203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,4,2,balanced,0.545157127380371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,4,2,power_law_1.01,0.5515584182739258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,4,2,balanced,0.7426764678955078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,4,2,power_law_1.01,9.099987182617188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,4,2,power_law_1.01,0.03250047922134399
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,4,2,power_law_1.01,0.03847424030303955
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,4,2,power_law_1.01,0.05084671974182129
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,4,2,power_law_1.01,0.048951039314270015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,4,2,power_law_1.01,0.05049344062805176
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,4,2,power_law_1.01,0.047863039970397946
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,4,2,power_law_1.01,0.049616641998291015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,4,2,power_law_1.01,0.05100800037384033
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,4,2,power_law_1.01,0.04938496112823486
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,4,2,power_law_1.01,0.04990464210510254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,4,2,power_law_1.01,0.43773056030273433
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,4,2,power_law_1.01,0.05508992195129394
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,4,2,power_law_1.01,0.057795839309692384
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,4,2,power_law_1.01,0.8443801879882813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,4,2,power_law_1.01,1.0286489868164064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,4,2,power_law_1.01,5.99788818359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,4,2,power_law_1.01,2.465800933837891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,4,2,power_law_1.01,1.7721075439453124
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,4,2,power_law_1.01,0.030300159454345704
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,4,2,power_law_1.01,0.054228482246398924
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,4,2,power_law_1.01,0.06007936000823975
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,4,2,power_law_1.01,0.061789441108703616
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,4,2,power_law_1.01,0.06794496059417725
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,4,2,power_law_1.01,0.08772607803344726
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,4,2,power_law_1.01,0.101397762298584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,4,2,power_law_1.01,0.13042176246643067
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,4,2,power_law_1.2,0.08643072128295899
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,4,2,power_law_1.2,0.09801471710205079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,4,2,power_law_1.2,0.11159296035766601
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,4,2,power_law_1.2,0.18287744522094726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,4,2,power_law_1.2,0.18322816848754883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,4,2,power_law_1.2,0.18457727432250975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,4,2,power_law_1.2,0.18895488739013672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,4,2,power_law_1.2,0.19133567810058594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,4,2,power_law_1.2,0.19487360000610351
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,4,2,power_law_1.2,0.19622400283813476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,4,2,power_law_1.2,0.20257919311523437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,4,2,power_law_1.2,0.21644159317016604
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,4,2,power_law_1.2,0.2368601608276367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,4,2,power_law_1.2,0.2775846481323242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,4,2,power_law_1.2,0.324005126953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,4,2,power_law_1.2,0.3453337478637695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,4,2,power_law_1.2,0.38426239013671876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,4,2,power_law_1.01,0.1506982421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,4,2,power_law_1.2,0.47753215789794917
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,4,2,power_law_1.2,0.5243609619140626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,4,2,power_law_1.01,0.21599872589111327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,4,2,power_law_1.2,0.8216588592529297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,4,2,power_law_1.2,0.8962598419189453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,4,2,power_law_1.01,0.2913203239440918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,4,2,power_law_1.01,0.41223808288574215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,4,2,power_law_1.2,1.6904051208496091
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,4,2,power_law_1.2,2.060832061767578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,4,2,power_law_1.2,0.056929278373718264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,4,2,power_law_1.2,0.06424960136413574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,4,2,power_law_1.2,3.978645629882813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,4,2,power_law_1.2,0.07145984172821045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,4,2,power_law_1.2,0.08437503814697266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,4,2,power_law_1.2,0.08245247840881348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,4,2,power_law_1.2,0.08368895530700685
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,4,2,power_law_1.01,0.5593305587768554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,4,2,power_law_1.2,0.08518527984619141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,4,2,power_law_1.2,0.0850432014465332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,4,2,power_law_1.2,0.08897791862487793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,4,2,power_law_1.2,0.0874022388458252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,4,2,power_law_1.2,0.09005056381225586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,4,2,power_law_1.2,0.1022822380065918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,4,2,power_law_1.2,0.1034329605102539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,4,2,power_law_1.2,0.11355135917663575
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,4,2,power_law_1.2,0.12355199813842774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,4,2,power_law_1.2,0.12764415740966797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,4,2,power_law_1.2,0.13026304244995118
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,4,2,power_law_1.2,0.16069631576538085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,4,2,power_law_1.2,2.7853927612304688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,4,2,power_law_1.2,0.18929023742675782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,4,2,power_law_1.2,0.23757440567016602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,4,2,power_law_1.2,0.3019071960449219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,4,2,power_law_1.2,0.42926464080810545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,4,2,power_law_1.2,0.54853759765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,4,2,power_law_1.01,0.8312230682373046
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,4,2,power_law_1.2,0.03267071962356567
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,4,2,power_law_1.2,0.0344320011138916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,4,2,power_law_1.2,0.0366540789604187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,4,2,power_law_1.2,0.046673917770385744
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,4,2,power_law_1.2,0.04611839771270752
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,4,2,power_law_1.2,0.048183040618896486
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,4,2,power_law_1.2,0.049388799667358395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,4,2,power_law_1.2,0.04967167854309082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,4,2,power_law_1.2,0.04967167854309082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,4,2,power_law_1.2,0.05009280204772949
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,4,2,power_law_1.2,0.04915328025817871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,4,2,power_law_1.2,0.8343193817138672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,4,2,power_law_1.2,0.055773439407348635
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,4,2,power_law_1.2,0.058695678710937504
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,4,2,power_law_1.2,0.059619841575622556
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,4,2,power_law_1.2,0.063439359664917
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,4,2,power_law_1.2,8.440459594726562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,4,2,power_law_1.01,1.1303616333007813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,4,2,power_law_1.2,0.10320256233215333
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,4,2,power_law_1.2,0.13542271614074708
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,4,2,power_law_1.2,1.076500473022461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,4,2,power_law_1.2,5.93067138671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,4,2,power_law_1.2,1.7754074096679688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,4,2,power_law_1.2,2.3868275451660157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,4,2,power_law_1.2,0.055930881500244146
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,4,2,power_law_1.2,0.06888448238372803
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,4,2,power_law_1.2,0.08629887580871583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,8,1,balanced,0.07731776237487793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,8,1,balanced,0.09244864463806153
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,8,1,balanced,0.12657535552978516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,8,1,balanced,0.13094847679138183
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,8,1,balanced,0.16455808639526368
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,8,1,balanced,0.16537151336669922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,8,1,balanced,0.1666752052307129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,8,1,balanced,0.16764991760253906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,8,1,balanced,0.16816959381103516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,8,1,balanced,0.16810688018798828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,8,1,balanced,0.16941247940063478
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,8,1,balanced,0.17099136352539063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,8,1,balanced,0.17361408233642578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,8,1,balanced,0.17726848602294923
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,8,1,balanced,0.22132799148559573
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,8,1,balanced,0.24482496261596678
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,8,1,balanced,0.24816192626953124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,8,1,balanced,0.3312108612060547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,8,1,balanced,0.3866329574584961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,8,1,balanced,0.5662054443359376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,8,1,balanced,0.7049945831298828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,4,2,power_law_1.2,0.14661120414733886
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,8,1,balanced,1.0942829132080079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,8,1,balanced,1.3669606018066407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,8,1,balanced,2.1426963806152344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,8,1,balanced,2.7701620483398437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,8,1,balanced,4.186820983886719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,8,1,balanced,0.04649792194366455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,8,1,balanced,0.06246975898742676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,8,1,balanced,0.08650752067565917
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,4,2,power_law_1.2,0.2280499267578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,8,1,balanced,0.09007231712341308
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,8,1,balanced,0.09153216361999512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,8,1,balanced,0.09305279731750489
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,8,1,balanced,0.09351424217224122
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,4,2,power_law_1.2,0.2957235145568847
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,8,1,balanced,0.09388671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,8,1,balanced,0.09471039772033692
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,8,1,balanced,0.09589759826660156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,8,1,balanced,0.09650943756103517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,8,1,balanced,0.0991744041442871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,8,1,balanced,5.5502777099609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,8,1,balanced,0.10154239654541015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,8,1,balanced,0.10501184463500976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,8,1,balanced,0.10773951530456542
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,8,1,balanced,0.1074681568145752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,8,1,balanced,0.14184767723083497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,8,1,balanced,0.1501375961303711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,8,1,balanced,0.19840831756591798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,8,1,balanced,0.2520249557495117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,8,1,balanced,0.3531167984008789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,8,1,balanced,0.4388051223754883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,8,1,balanced,0.6496927642822266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,8,1,balanced,0.862449951171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,8,1,balanced,0.08553536415100098
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,8,1,balanced,1.2870002746582032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,8,1,balanced,1.7162156677246094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,4,2,power_law_1.2,0.42265472412109373
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,4,2,power_law_1.2,0.8169113922119141
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,4,2,power_law_1.2,0.5411635208129882
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,4,2,power_law_1.2,0.9817830657958984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,8,1,balanced,0.02177983999252319
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,8,1,balanced,0.0491980791091919
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,8,1,balanced,0.04933504104614258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,8,1,balanced,0.02939840078353882
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,8,1,balanced,0.049428482055664066
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,8,1,balanced,0.04947072029113769
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,8,1,balanced,0.04940032005310059
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,8,1,balanced,0.051543679237365726
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,8,1,balanced,0.05212927818298339
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,8,1,balanced,0.05227200031280518
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,8,1,balanced,0.048009600639343265
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,8,1,balanced,0.053951997756958005
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,8,1,balanced,0.04810880184173584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,8,1,balanced,0.04789247989654541
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,8,1,balanced,0.048534398078918455
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,8,1,balanced,0.049015679359436036
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,8,1,balanced,0.04889920234680176
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,8,1,balanced,0.10025919914245604
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,8,1,balanced,0.07550464153289796
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,8,1,balanced,0.07843967914581298
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,8,1,power_law_1.01,0.07646527767181396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,8,1,power_law_1.01,0.0909171199798584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,8,1,balanced,0.1209164810180664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,8,1,power_law_1.01,0.10007807731628418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,8,1,power_law_1.01,0.1645894432067871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,8,1,power_law_1.01,0.16555456161499021
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,8,1,power_law_1.01,0.16668415069580078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,8,1,power_law_1.01,0.16617088317871093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,8,1,power_law_1.01,0.1673574447631836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,8,1,power_law_1.01,0.17024831771850585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,8,1,power_law_1.01,0.18466880798339844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,8,1,power_law_1.01,0.19257535934448242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,8,1,power_law_1.01,0.12966591835021973
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,8,1,power_law_1.01,0.20812095642089842
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,8,1,power_law_1.01,0.21915199279785158
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,8,1,balanced,0.16845567703247072
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,8,1,power_law_1.01,0.22783424377441408
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,8,1,power_law_1.01,0.28359487533569333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,8,1,power_law_1.01,0.3126137542724609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,8,1,power_law_1.01,0.381921272277832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,8,1,power_law_1.01,0.4628121566772461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,8,1,power_law_1.01,0.6054022216796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,8,1,power_law_1.01,0.7575341033935546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,8,1,power_law_1.01,1.105761260986328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,8,1,power_law_1.01,1.442515869140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,8,1,balanced,0.21926591873168944
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,8,1,balanced,0.3172313690185547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,8,1,power_law_1.01,0.05288447856903076
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,8,1,power_law_1.01,0.060012798309326175
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,8,1,power_law_1.01,0.06417791843414307
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,8,1,power_law_1.01,0.08392448425292968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,8,1,power_law_1.01,2.158849334716797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,8,1,power_law_1.01,0.09055616378784179
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,8,1,power_law_1.01,0.0931116771697998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,8,1,power_law_1.01,0.09230591773986815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,8,1,power_law_1.01,0.0933619213104248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,8,1,power_law_1.01,0.09288767814636231
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,8,1,power_law_1.01,0.09460288047790527
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,8,1,power_law_1.01,0.09536895751953126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,8,1,power_law_1.01,0.0961849594116211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,8,1,power_law_1.01,0.09940095901489257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,8,1,power_law_1.01,0.11220288276672363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,8,1,power_law_1.01,0.11754624366760254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,8,1,power_law_1.01,0.11920448303222657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,8,1,power_law_1.01,0.1280019187927246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,8,1,power_law_1.01,0.14204992294311522
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,8,1,power_law_1.01,0.17849536895751955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,8,1,power_law_1.01,0.2260211181640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,8,1,power_law_1.01,2.8353778076171876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,8,1,power_law_1.01,0.2658387184143066
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,8,1,balanced,0.41714942932128907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,8,1,power_law_1.01,0.3623052978515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,8,1,balanced,0.6429177856445312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,8,1,balanced,0.9232179260253905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,8,1,power_law_1.01,0.46367553710937504
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,8,1,power_law_1.01,0.02530816078186035
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,8,1,power_law_1.01,0.03166143894195557
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,8,1,power_law_1.01,0.039059200286865235
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,8,1,power_law_1.01,0.04788544178009033
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,8,1,power_law_1.01,0.04814144134521484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,8,1,power_law_1.01,0.04873407840728759
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,8,1,power_law_1.01,0.04858367919921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,8,1,power_law_1.01,0.04850111961364746
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,8,1,power_law_1.01,0.04821760177612305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,8,1,power_law_1.01,0.049108481407165526
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,8,1,power_law_1.01,0.04967872142791748
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,8,1,power_law_1.01,0.05214144229888916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,8,1,power_law_1.01,0.6668544006347656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,8,1,power_law_1.01,0.05727424144744873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,8,1,power_law_1.01,4.15593994140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,8,1,power_law_1.01,0.8754195404052734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,8,1,power_law_1.01,5.656679077148437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,8,1,power_law_1.01,1.3108819580078124
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,8,1,power_law_1.01,0.09528127670288086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,8,1,power_law_1.01,1.7741722106933593
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,8,1,power_law_1.01,0.20613056182861328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,8,1,power_law_1.01,0.05151936054229737
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,8,1,power_law_1.01,0.05947072029113769
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,8,1,power_law_1.01,0.06058432102203369
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,8,1,power_law_1.01,0.06672319889068604
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,8,1,power_law_1.01,0.08054656028747559
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,8,1,power_law_1.2,0.07762623786926269
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,8,1,power_law_1.2,0.08988863945007323
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,8,1,power_law_1.2,0.09600704193115235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,8,1,power_law_1.2,0.13055808067321778
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,8,1,power_law_1.2,0.1623955154418945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,8,1,power_law_1.2,0.16551807403564453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,8,1,power_law_1.2,0.16666368484497068
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,8,1,power_law_1.2,0.16671871185302733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,8,1,power_law_1.2,0.16912256240844728
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,8,1,power_law_1.2,0.17002752304077148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,8,1,power_law_1.2,0.18792831420898437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,8,1,power_law_1.2,0.19700607299804687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,8,1,power_law_1.2,0.20635391235351563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,8,1,power_law_1.2,0.2208332824707031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,8,1,power_law_1.2,0.23000959396362303
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,8,1,power_law_1.2,0.2954521560668945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,8,1,power_law_1.2,0.3122054481506348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,8,1,power_law_1.2,0.38408321380615235
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,8,1,power_law_1.01,0.12323519706726074
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,8,1,power_law_1.01,0.14952768325805663
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,8,1,power_law_1.2,0.46487552642822266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,8,1,power_law_1.2,0.611009292602539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,8,1,power_law_1.2,0.7601177978515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,8,1,power_law_1.2,1.095807342529297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,8,1,power_law_1.01,0.2624627113342285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,8,1,power_law_1.2,1.439173126220703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,8,1,power_law_1.01,0.3768646240234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,8,1,power_law_1.2,0.053659520149230956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,8,1,power_law_1.2,0.06050687789916992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,8,1,power_law_1.2,0.06210752010345459
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,8,1,power_law_1.01,0.4945062255859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,8,1,power_law_1.2,0.09240768432617188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,8,1,power_law_1.2,0.09226240158081055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,8,1,power_law_1.2,0.09276479721069336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,8,1,power_law_1.2,0.09290047645568847
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,8,1,power_law_1.2,2.1569932556152347
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,8,1,power_law_1.2,0.09390975952148437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,8,1,power_law_1.2,0.09701184272766114
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,8,1,power_law_1.2,0.09554495811462402
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,8,1,power_law_1.2,0.09811391830444335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,8,1,power_law_1.2,0.09861760139465332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,8,1,power_law_1.2,0.11340160369873047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,8,1,power_law_1.2,0.12160320281982422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,8,1,power_law_1.2,0.12410943984985352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,8,1,power_law_1.2,0.1316806411743164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,8,1,power_law_1.2,0.1419968032836914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,8,1,power_law_1.2,0.08348928451538086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,8,1,power_law_1.2,0.1778335952758789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,8,1,power_law_1.01,0.7409062194824219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,8,1,power_law_1.2,0.22074943542480469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,8,1,power_law_1.2,0.26688959121704103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,8,1,power_law_1.2,0.36279041290283204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,8,1,power_law_1.2,2.8221343994140624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,8,1,power_law_1.01,0.990439682006836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,8,1,power_law_1.2,0.02594239950180054
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,8,1,power_law_1.2,0.03118720054626465
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,8,1,power_law_1.2,0.03403840065002441
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,8,1,power_law_1.2,0.4640838241577148
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,8,1,power_law_1.2,0.047585282325744625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,8,1,power_law_1.2,0.048056321144104006
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,8,1,power_law_1.2,0.048253440856933595
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,8,1,power_law_1.2,0.04863232135772705
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,8,1,power_law_1.2,0.04837567806243896
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,8,1,power_law_1.2,0.049110398292541504
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,8,1,power_law_1.2,0.04919487953186035
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,8,1,power_law_1.2,0.049467520713806154
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,8,1,power_law_1.2,0.052100481986999514
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,8,1,power_law_1.2,0.057623682022094724
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,8,1,power_law_1.2,0.06100800037384033
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,8,1,power_law_1.2,0.0628704023361206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,8,1,power_law_1.2,0.6614860534667969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,8,1,power_law_1.2,0.8882022094726562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,8,1,power_law_1.2,4.183225708007813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,8,1,power_law_1.2,0.05165696144104004
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,8,1,power_law_1.2,0.06734015941619872
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,8,1,power_law_1.2,1.3132614135742187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,8,1,power_law_1.2,5.601234741210938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,8,1,power_law_1.2,0.08163007736206054
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,8,1,power_law_1.2,0.09571007728576661
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,8,1,power_law_1.2,0.12405952453613281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,8,1,power_law_1.2,1.7642259216308596
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,8,balanced,0.07444479942321777
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,8,balanced,0.08130687713623047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,8,balanced,0.09863807678222655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,8,balanced,0.14634880065917968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,8,balanced,0.3006707191467285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,8,balanced,0.528276481628418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,8,balanced,0.6127078247070312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,8,balanced,0.6116505432128906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,8,balanced,0.6214720153808594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,8,balanced,0.6307238388061523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,8,balanced,0.6306611251831055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,8,balanced,0.6359436798095703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,8,balanced,0.6279577636718751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,8,balanced,0.6578316497802734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,8,balanced,0.6696243286132812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,8,balanced,0.6741632080078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,8,balanced,0.6900761413574219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,8,1,power_law_1.2,0.1523366355895996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,8,balanced,0.7485632324218751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,8,balanced,0.7207936096191406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,8,balanced,0.8089049530029296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,8,balanced,0.8927168273925782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,8,balanced,1.278375701904297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,8,balanced,1.4443519592285157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,8,balanced,2.0721638488769534
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,8,1,power_law_1.2,0.21259199142456056
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,8,balanced,2.7146420288085937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,8,balanced,0.039950079917907715
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,8,1,power_law_1.2,0.26865983963012696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,8,balanced,4.4348928833007815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,8,balanced,0.08939007759094239
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,8,balanced,0.04697984218597412
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,8,balanced,0.1526425552368164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,8,balanced,0.29407615661621095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,8,balanced,0.29831039428710937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,8,balanced,0.3003264045715332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,8,balanced,0.06546815872192382
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,8,balanced,0.2754470443725586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,8,balanced,0.3031667137145996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,8,balanced,0.29255680084228514
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,8,balanced,0.2964595222473144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,8,balanced,0.2975308799743652
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,8,balanced,0.30081407546997074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,8,balanced,6.584090576171874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,8,balanced,0.3031091117858887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,8,balanced,0.3115328025817871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,8,balanced,0.3330316925048828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,8,balanced,0.36404991149902344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,8,balanced,0.38017921447753905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,8,balanced,0.4317798233032226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,8,balanced,0.4815488052368164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,8,balanced,0.587185935974121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,8,balanced,0.70906494140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,8,balanced,0.9654195404052734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,8,balanced,1.1672358703613281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,8,balanced,1.6973158264160155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,8,balanced,2.1938099670410156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,8,1,power_law_1.2,0.3893747329711914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,8,1,power_law_1.2,0.5062879943847657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,1,8,balanced,0.024817919731140135
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,1,8,balanced,0.15632639884948732
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,1,8,balanced,0.025768959522247316
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,1,8,balanced,0.15647871971130373
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,1,8,balanced,0.15598591804504394
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,1,8,balanced,0.1561676788330078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,1,8,balanced,0.15647359848022463
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,1,8,balanced,0.15713024139404297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,1,8,balanced,0.03621376037597656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,1,8,balanced,0.15704447746276856
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,1,8,balanced,0.1573209571838379
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,1,8,balanced,0.15804415702819824
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,1,8,balanced,0.15889151573181154
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,1,8,balanced,0.05683584213256836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,1,8,balanced,0.1599283218383789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,1,8,balanced,0.1617625617980957
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,1,8,balanced,0.16332544326782225
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,1,8,balanced,0.17029760360717774
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,1,8,balanced,0.1728985595703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,1,8,balanced,0.18263168334960939
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,1,8,balanced,0.19110143661499024
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,8,1,power_law_1.2,0.7449209594726562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,8,power_law_1.01,0.14713600158691406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,8,power_law_1.01,0.2997619247436524
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,1,8,balanced,0.0920691204071045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,8,power_law_1.01,0.34208255767822265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,8,power_law_1.01,0.3803993606567383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,8,power_law_1.01,0.4252787017822266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,8,power_law_1.01,0.487248649597168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,8,power_law_1.01,0.5587366485595704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,8,power_law_1.01,0.5524927902221679
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,8,power_law_1.01,0.5710195159912109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,8,power_law_1.01,0.5582310485839843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,8,power_law_1.01,0.5971724700927734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,8,power_law_1.01,0.5751359939575196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,8,power_law_1.01,0.5846963119506836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,8,power_law_1.01,0.5861248016357422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,8,power_law_1.01,0.6331942367553711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,8,power_law_1.01,0.6495308685302734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,1,8,balanced,0.15680768013000487
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,8,power_law_1.01,0.6692851257324219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,8,power_law_1.01,0.7132006072998047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,8,power_law_1.01,0.7855757141113282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,8,power_law_1.01,0.946698226928711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,8,power_law_1.01,1.125722885131836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,8,power_law_1.01,1.4237081909179687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,8,power_law_1.01,1.7749427795410155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,8,power_law_1.01,2.7349774169921877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,8,power_law_1.01,3.306199035644531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,8,power_law_1.01,5.018563842773437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,8,power_law_1.01,0.08623104095458985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,8,power_law_1.01,0.14901375770568848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,8,power_law_1.01,0.19625600814819336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,8,power_law_1.01,0.2332915115356445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,8,power_law_1.01,0.17771135330200197
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,8,power_law_1.01,0.23979776382446288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,8,power_law_1.01,6.074633178710938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,8,power_law_1.01,0.2614912033081055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,8,power_law_1.01,0.255098876953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,8,power_law_1.01,0.2776383972167969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,8,1,power_law_1.2,1.0149081420898436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,8,power_law_1.01,0.2652249526977539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,8,power_law_1.01,0.2666227149963379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,8,power_law_1.01,0.27890047073364255
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,8,power_law_1.01,0.2733888053894043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,8,power_law_1.01,0.292043514251709
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,8,power_law_1.01,0.296810245513916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,8,power_law_1.01,0.302295036315918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,8,power_law_1.01,0.25063552856445315
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,8,power_law_1.01,0.33859710693359374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,8,power_law_1.01,0.3610943984985352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,8,power_law_1.01,0.4261119842529297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,8,power_law_1.01,0.4799603271484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,8,power_law_1.01,0.5720703887939453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,8,power_law_1.01,0.7024409484863281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,8,power_law_1.01,1.0198604583740234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,8,power_law_1.01,1.3616525268554687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,1,8,power_law_1.01,0.09286911964416504
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,1,8,power_law_1.01,0.11098752021789551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,8,power_law_1.01,1.921644744873047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,1,8,power_law_1.01,0.11369983673095703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,1,8,power_law_1.01,0.1348620796203613
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,1,8,power_law_1.01,0.132291841506958
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,8,power_law_1.01,2.440021820068359
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,1,8,power_law_1.01,0.14180480003356935
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,1,8,power_law_1.01,0.14694527626037596
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,1,8,power_law_1.01,0.146494722366333
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,1,8,power_law_1.01,0.13811967849731446
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,1,8,power_law_1.01,0.14317312240600585
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,1,8,power_law_1.01,0.14561663627624513
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,1,8,power_law_1.01,0.14569984436035158
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,1,8,power_law_1.01,0.15093888282775877
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,1,8,power_law_1.01,0.1541644763946533
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,1,8,power_law_1.01,0.1556991958618164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,1,8,balanced,0.5245222473144532
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,1,8,balanced,0.29519744873046877
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,1,8,balanced,0.4082854461669922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,1,8,balanced,0.27232383728027343
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,1,8,power_law_1.01,0.057657599449157715
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,1,8,power_law_1.01,0.12904704093933106
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,1,8,power_law_1.01,0.168907527923584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,1,8,power_law_1.01,0.17869823455810546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,1,8,power_law_1.01,0.20452224731445315
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,1,8,power_law_1.01,0.30232959747314453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,8,power_law_1.2,0.30384767532348633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,8,power_law_1.2,0.14368895530700682
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,8,power_law_1.2,0.2626688003540039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,8,power_law_1.2,0.4604044723510742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,8,power_law_1.2,0.3895923233032227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,8,power_law_1.2,0.39616512298583983
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,1,8,power_law_1.01,0.33673728942871095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,8,power_law_1.2,0.5429516983032227
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,1,8,power_law_1.01,0.22822784423828124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,8,power_law_1.2,0.5100979232788087
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,8,power_law_1.2,0.5838297653198242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,8,power_law_1.2,0.563801612854004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,8,power_law_1.2,0.5220492935180664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,8,power_law_1.2,0.5456192016601562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,8,power_law_1.2,0.5911296081542969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,8,power_law_1.2,0.6100799942016601
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,8,power_law_1.2,0.6450624084472656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,8,power_law_1.2,0.6526143646240234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,8,power_law_1.2,0.6703065490722656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,8,power_law_1.2,0.7284992218017579
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,8,power_law_1.2,0.8606310272216797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,8,power_law_1.2,1.1094502258300782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,8,power_law_1.2,1.2040345764160156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,8,power_law_1.2,1.5470693969726563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,8,power_law_1.2,2.144061431884766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,8,power_law_1.2,2.9779367065429687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,8,power_law_1.2,0.08784640312194825
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,8,power_law_1.2,0.14029312133789062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,8,power_law_1.2,0.16498943328857424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,8,power_law_1.2,4.3854373168945315
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,8,power_law_1.2,0.20994176864624023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,8,power_law_1.2,0.23139455795288083
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,8,power_law_1.2,0.2418880081176758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,8,power_law_1.2,0.26173696517944334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,8,power_law_1.2,5.71720703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,8,power_law_1.2,0.2383500862121582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,8,power_law_1.2,0.2752422332763672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,8,power_law_1.2,0.2574323272705078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,8,power_law_1.2,0.2488921546936035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,8,power_law_1.2,0.27571199417114256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,8,power_law_1.2,0.27382015228271483
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,8,power_law_1.2,0.2857382392883301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,8,power_law_1.2,0.2973721694946289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,8,power_law_1.2,0.29181568145751957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,8,power_law_1.2,0.313698558807373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,8,power_law_1.2,0.3495091247558594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,8,power_law_1.2,0.36918529510498044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,8,power_law_1.2,7.578543090820313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,8,power_law_1.2,0.4652544021606445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,8,power_law_1.2,0.5109145736694336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,8,power_law_1.2,0.6538188934326172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,8,power_law_1.2,0.8346546936035157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,8,power_law_1.2,1.1209292602539063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,1,8,power_law_1.2,0.057625598907470696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,8,power_law_1.2,1.3558694458007812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,1,8,power_law_1.2,0.09047039985656738
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,1,8,power_law_1.2,0.09969280242919923
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,1,8,power_law_1.2,0.10372351646423339
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,1,8,power_law_1.2,0.14113408088684082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,1,8,power_law_1.2,0.12828160285949708
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,8,power_law_1.2,2.0292390441894534
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,1,8,power_law_1.2,0.13736191749572754
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,1,8,power_law_1.2,0.14445952415466307
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,1,8,power_law_1.2,0.13687680244445802
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,1,8,power_law_1.2,0.14044544219970703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,1,8,power_law_1.2,0.14466560363769532
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,1,8,power_law_1.2,0.13924351692199707
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,1,8,power_law_1.2,0.14460927963256837
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,1,8,power_law_1.2,0.14350208282470703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,1,8,power_law_1.2,0.15458815574645995
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,1,8,power_law_1.2,0.15850239753723144
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,1,8,power_law_1.2,0.1842086410522461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,1,8,power_law_1.2,0.20231935501098636
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,8,power_law_1.2,2.8289215087890627
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,1,8,power_law_1.01,0.49711872100830073
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,1,8,power_law_1.01,0.570087661743164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,1,8,power_law_1.01,0.9033164978027344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,1,8,power_law_1.01,1.128257293701172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,1,8,power_law_1.2,0.16254335403442383
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,1,8,power_law_1.2,0.22855680465698242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,1,8,power_law_1.2,0.2608614349365234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,4,balanced,0.06361087799072265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,4,balanced,0.07605631828308106
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,4,balanced,0.09731840133666993
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,4,balanced,0.1478604793548584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,1,8,power_law_1.2,0.3691289520263672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,4,balanced,0.2934515190124512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,4,balanced,0.5401497650146484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,4,balanced,0.5192204666137695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,4,balanced,0.5959884643554687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,4,balanced,0.5939519882202149
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,4,balanced,0.6082982254028321
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,4,balanced,0.5256665420532227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,4,balanced,0.6095910263061524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,4,balanced,0.6067148971557617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,4,balanced,0.6240703964233398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,4,balanced,0.6392319869995118
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,1,8,power_law_1.2,0.45377025604248045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,4,balanced,0.6498278045654298
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,4,balanced,0.6665792083740234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,4,balanced,0.6899609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,4,balanced,0.697286376953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,4,balanced,0.7607231903076171
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,4,balanced,0.8397337341308594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,4,balanced,1.1991436767578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,1,8,power_law_1.2,0.6468045043945312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,4,balanced,1.3597811889648437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,4,balanced,0.03772799968719483
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,4,balanced,1.9183193969726564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,4,balanced,0.04774271965026856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,4,balanced,0.06709760189056396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,4,balanced,0.10266112327575683
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,4,balanced,0.167457275390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,4,balanced,0.28520576477050785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,4,balanced,2.429251861572266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,4,balanced,0.27897344589233397
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,4,balanced,0.28238592147827146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,4,balanced,0.304268798828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,4,balanced,0.3043046379089355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,4,balanced,0.3069708824157715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,4,balanced,0.31041664123535156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,1,8,power_law_1.2,0.859389419555664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,4,balanced,0.31255680084228515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,4,balanced,0.3163046455383301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,4,balanced,0.317073917388916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,4,balanced,3.580332946777344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,4,balanced,0.32380416870117185
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,1,8,power_law_1.2,1.1704959869384766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,4,balanced,0.33741054534912107
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,4,balanced,0.3703692626953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,4,balanced,0.3844351959228516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,4,balanced,0.4335871887207031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,4,balanced,0.48324737548828123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,4,balanced,0.5510835266113281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,4,balanced,0.6629388427734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,4,balanced,0.9346495819091796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,4,balanced,5.230869750976562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,4,balanced,1.1083916473388673
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,4,balanced,1.6241407775878904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,4,balanced,2.0832281494140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,1,8,power_law_1.2,1.7832896423339843
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,2,4,balanced,0.02930943965911865
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,2,4,balanced,0.032592639923095704
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,2,4,balanced,0.16613887786865233
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,2,4,balanced,0.1668172836303711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,2,4,balanced,0.06034175872802734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,2,4,balanced,0.16660863876342774
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,2,4,balanced,0.041507840156555176
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,2,4,balanced,0.16739839553833008
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,2,4,balanced,0.1673459243774414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,2,4,balanced,0.1701145553588867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,2,4,balanced,0.16994176864624025
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,2,4,balanced,0.17061759948730468
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,2,4,balanced,0.17239551544189452
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,2,4,balanced,0.1791257667541504
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,2,4,balanced,0.1873855972290039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,2,4,balanced,0.09735679626464844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,2,4,balanced,0.19354112625122072
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,2,4,balanced,0.20838911056518555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,2,4,balanced,0.22197248458862306
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,2,4,balanced,0.1659008026123047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,4,power_law_1.01,0.09709823608398438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,2,4,balanced,0.17570816040039064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,4,power_law_1.01,0.14022656440734865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,4,power_law_1.01,0.17619968414306642
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,2,4,balanced,0.16572032928466798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,4,power_law_1.01,0.2943564796447754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,4,power_law_1.01,0.3322547149658203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,4,power_law_1.01,0.3875724792480469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,4,power_law_1.01,0.481187858581543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,4,power_law_1.01,0.46575103759765624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,4,power_law_1.01,0.5553715133666992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,4,power_law_1.01,0.5067776107788087
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,4,power_law_1.01,0.5527475357055665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,4,power_law_1.01,0.538424301147461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,4,power_law_1.01,0.5501529693603515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,4,power_law_1.01,0.5922060775756836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,4,power_law_1.01,0.6073561477661132
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,4,power_law_1.01,0.5979596710205077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,4,power_law_1.01,0.6311999893188476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,4,power_law_1.01,0.676622085571289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,4,power_law_1.01,0.7494617462158203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,4,power_law_1.01,0.8529766082763672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,4,power_law_1.01,1.005722885131836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,4,power_law_1.01,1.3122930908203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,4,power_law_1.01,1.5881765747070313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,4,power_law_1.01,2.2032063293457034
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,4,power_law_1.01,2.9973248291015624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,4,power_law_1.01,0.06208640098571777
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,4,power_law_1.01,0.08966912269592285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,4,power_law_1.01,0.11992192268371582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,4,power_law_1.01,4.162232360839844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,4,power_law_1.01,0.14712575912475584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,4,power_law_1.01,0.18423295974731446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,4,power_law_1.01,0.21469055175781251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,4,power_law_1.01,0.22816640853881837
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,4,power_law_1.01,0.24578815460205078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,4,power_law_1.01,0.25198591232299805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,4,power_law_1.01,0.24166015624999998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,4,power_law_1.01,0.2521881675720215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,4,power_law_1.01,0.2667558479309082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,4,power_law_1.01,0.28953216552734373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,4,power_law_1.01,0.30418432235717774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,4,power_law_1.01,5.993192749023438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,4,power_law_1.01,0.30164863586425783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,4,power_law_1.01,0.3128742408752442
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,4,power_law_1.01,0.3482291030883789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,4,power_law_1.01,0.3734207916259765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,4,power_law_1.01,0.3091801643371582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,4,power_law_1.01,0.44055934906005856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,4,power_law_1.01,0.4919526290893555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,4,power_law_1.01,0.6109324645996094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,4,power_law_1.01,0.7608499145507812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,4,power_law_1.01,0.9996083068847657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,4,power_law_1.01,1.2568755340576172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,2,4,power_law_1.01,0.04509183883666992
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,2,4,balanced,0.16631423950195312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,2,4,power_law_1.01,0.0637824010848999
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,4,power_law_1.01,1.79981689453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,2,4,power_law_1.01,0.07517695903778077
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,2,4,power_law_1.01,0.09161343574523925
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,2,4,power_law_1.01,0.10644864082336425
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,2,4,power_law_1.01,0.12896384239196776
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,2,4,power_law_1.01,0.14976511955261232
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,2,4,power_law_1.01,0.14600319862365724
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,2,4,power_law_1.01,0.14893312454223634
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,2,4,power_law_1.01,0.14804736137390137
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,2,4,power_law_1.01,0.15049599647521972
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,2,4,power_law_1.01,0.1541926383972168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,4,power_law_1.01,2.4480128479003906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,2,4,power_law_1.01,0.14664319992065428
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,2,4,power_law_1.01,0.14968576431274414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,2,4,power_law_1.01,0.17075328826904296
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,2,4,power_law_1.01,0.16261375427246094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,2,4,power_law_1.01,0.17273471832275392
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,2,4,balanced,0.32491264343261717
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,2,4,balanced,0.3557823944091797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,2,4,balanced,0.5034124755859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,2,4,balanced,0.6477439880371094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,2,4,power_law_1.01,0.19056640625000001
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,2,4,power_law_1.01,0.24107007980346679
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,2,4,power_law_1.01,0.20534784317016602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,4,power_law_1.2,0.09655808448791504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,4,power_law_1.2,0.13553536415100098
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,4,power_law_1.2,0.155918083190918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,4,power_law_1.2,0.27009279251098633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,2,4,power_law_1.01,0.27273855209350584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,4,power_law_1.2,0.34149505615234377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,4,power_law_1.2,0.3015884780883789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,4,power_law_1.2,0.4622796630859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,4,power_law_1.2,0.490827522277832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,4,power_law_1.2,0.5167398452758789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,4,power_law_1.2,0.5575270462036133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,4,power_law_1.2,0.5287897491455078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,4,power_law_1.2,0.5199897766113282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,4,power_law_1.2,0.5704627227783203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,4,power_law_1.2,0.594442253112793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,4,power_law_1.2,0.6213593673706055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,4,power_law_1.2,0.6333427047729492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,4,power_law_1.2,0.6487590026855469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,4,power_law_1.2,0.7107059478759765
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,2,4,power_law_1.01,0.4208793640136719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,2,4,power_law_1.01,0.32937984466552733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,4,power_law_1.2,0.7579634857177735
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,2,4,power_law_1.01,0.6237811279296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,2,4,power_law_1.01,0.8296153259277343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,4,power_law_1.2,0.9571686553955079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,4,power_law_1.2,1.0823474884033204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,4,power_law_1.2,1.402960662841797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,4,power_law_1.2,0.06260096073150635
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,4,power_law_1.2,1.8096934509277343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,4,power_law_1.2,0.08456192016601563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,4,power_law_1.2,0.10079360008239746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,4,power_law_1.2,2.2548544311523435
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,4,power_law_1.2,0.1368832015991211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,4,power_law_1.2,0.17251583099365236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,4,power_law_1.2,0.2388403129577637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,4,power_law_1.2,0.21554304122924806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,4,power_law_1.2,3.0169165039062498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,4,power_law_1.2,0.24809984207153318
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,4,power_law_1.2,0.2437196731567383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,4,power_law_1.2,0.2835417556762695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,4,power_law_1.2,0.2660070419311523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,4,power_law_1.2,0.3012339210510254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,4,power_law_1.2,0.30344448089599607
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,4,power_law_1.2,0.3101312065124512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,4,power_law_1.2,0.3166566467285156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,4,power_law_1.2,4.616039733886718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,4,power_law_1.2,0.32064769744873045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,4,power_law_1.2,0.3711872100830078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,4,power_law_1.2,0.3864908981323242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,4,power_law_1.2,0.4638988876342774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,4,power_law_1.2,0.5263449478149413
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,4,power_law_1.2,0.6356288146972656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,4,power_law_1.2,0.7925222778320313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,4,power_law_1.2,5.934801635742188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,4,power_law_1.2,1.0634226989746094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,2,4,power_law_1.2,0.04627583980560303
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,4,power_law_1.2,1.2953765869140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,2,4,power_law_1.2,0.06157055854797363
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,2,4,power_law_1.2,0.06711423873901368
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,2,4,power_law_1.2,0.07804416179656982
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,2,4,power_law_1.2,0.10341119766235352
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,2,4,power_law_1.2,0.12514944076538087
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,2,4,power_law_1.2,0.146496000289917
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,2,4,power_law_1.2,0.14753664016723633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,2,4,power_law_1.2,0.14852095603942872
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,4,power_law_1.2,1.9672525024414065
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,2,4,power_law_1.2,0.14828160285949707
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,2,4,power_law_1.2,0.146496000289917
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,2,4,power_law_1.2,0.14853119850158691
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,2,4,power_law_1.2,0.16516992568969727
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,2,4,power_law_1.2,0.16595712661743164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,2,4,power_law_1.2,0.16871807098388672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,2,4,power_law_1.2,0.16656255722045898
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,2,4,power_law_1.2,0.17771520614624023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,4,power_law_1.2,2.388890838623047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,4,power_law_1.2,0.2025152015686035
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,2,4,power_law_1.01,1.1398719787597655
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,2,4,power_law_1.01,1.3807629394531251
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,2,4,power_law_1.2,0.20899583816528322
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,2,4,power_law_1.2,0.2775551986694336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,2,4,power_law_1.2,0.22743808746337893
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,2,balanced,0.06301439762115478
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,2,balanced,0.07500927925109864
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,2,balanced,0.09454463958740235
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,2,4,power_law_1.2,0.34510208129882813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,2,balanced,0.14049792289733887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,2,balanced,0.3023654365539551
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,2,4,power_law_1.2,0.4401663970947266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,2,balanced,0.5253516769409179
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,2,balanced,0.5260198211669922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,2,balanced,0.5251724624633789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,2,balanced,0.5279500961303711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,2,balanced,0.5249395370483398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,2,balanced,0.5252223968505859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,2,balanced,0.5988313674926757
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,2,balanced,0.6096652984619141
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,2,4,power_law_1.2,0.5771929550170898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,2,balanced,0.6116262435913086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,2,balanced,0.6195827102661132
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,2,balanced,0.6161318588256836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,2,balanced,0.6409740447998047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,2,balanced,0.6780518341064453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,2,balanced,0.6795699310302734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,2,balanced,0.7399935913085938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,2,balanced,0.8481075286865234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,2,balanced,1.1542208099365234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,2,balanced,1.313269805908203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,2,balanced,1.8789939880371094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,2,balanced,0.0377510404586792
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,2,balanced,0.04947328090667725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,2,balanced,2.3163839721679684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,2,balanced,0.06985856056213378
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,2,balanced,0.10502911567687989
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,2,balanced,0.18968320846557618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,2,balanced,0.31665536880493167
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,2,balanced,0.31656703948974607
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,2,balanced,0.3167193603515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,2,balanced,0.31944704055786133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,2,balanced,3.4700146484374996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,2,balanced,0.32173694610595704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,2,balanced,0.32409473419189455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,2,balanced,0.36606334686279296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,2,balanced,0.36360702514648435
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,2,balanced,0.36798206329345706
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,2,balanced,0.3707020950317383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,2,balanced,0.3758924865722656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,2,balanced,0.38369663238525387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,2,balanced,0.40374401092529294
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,2,balanced,4.999676208496093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,2,balanced,0.4330598449707031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,2,balanced,0.46155776977539065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,2,balanced,0.5099673461914063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,2,balanced,0.5857139205932617
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,2,4,power_law_1.2,0.8744614410400391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,2,balanced,0.675098876953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,2,balanced,0.98023681640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,2,balanced,1.1488985443115234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,2,balanced,1.6704934692382813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,2,balanced,2.1423437499999998
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,2,4,power_law_1.2,1.1065229034423827
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,2,4,power_law_1.2,1.8814373779296876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,2,4,power_law_1.2,2.4479603576660156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,4,2,balanced,0.040294399261474614
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,4,2,balanced,0.19798912048339842
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,4,2,balanced,0.04397696018218994
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,4,2,balanced,0.1981248092651367
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,4,2,balanced,0.0543398380279541
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,4,2,balanced,0.19853439331054687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,4,2,balanced,0.20092416763305665
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,4,2,balanced,0.20036735534667968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,4,2,balanced,0.20190847396850584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,4,2,balanced,0.2047871971130371
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,4,2,balanced,0.20557312011718748
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,4,2,balanced,0.20976383209228516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,4,2,balanced,0.07528448104858398
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,4,2,balanced,0.2162291145324707
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,4,2,balanced,0.22239616394042966
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,4,2,balanced,0.23507455825805662
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,4,2,balanced,0.24716672897338868
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,4,2,balanced,0.11408896446228027
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,4,2,balanced,0.27452415466308594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,4,2,balanced,0.30042495727539065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,2,power_law_1.01,0.0752665615081787
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,2,power_law_1.01,0.09447551727294921
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,4,2,balanced,0.19663488388061523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,2,power_law_1.01,0.1174118423461914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,2,power_law_1.01,0.16631935119628907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,2,power_law_1.01,0.247457275390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,2,power_law_1.01,0.34754688262939454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,2,power_law_1.01,0.3985868835449219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,2,power_law_1.01,0.40130046844482425
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,2,power_law_1.01,0.42534912109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,2,power_law_1.01,0.4130291366577149
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,2,power_law_1.01,0.4454067230224609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,2,power_law_1.01,0.49389312744140623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,2,power_law_1.01,0.5001548767089844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,2,power_law_1.01,0.5434124755859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,2,power_law_1.01,0.593946876525879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,2,power_law_1.01,0.6015667343139649
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,2,power_law_1.01,0.6310515213012695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,2,power_law_1.01,0.6711564636230468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,2,power_law_1.01,0.7130278778076172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,2,power_law_1.01,0.8645555114746093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,2,power_law_1.01,0.9911820983886719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,2,power_law_1.01,1.2842930603027345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,2,power_law_1.01,1.5791885375976562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,2,power_law_1.01,2.0841357421874998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,2,power_law_1.01,2.6174591064453123
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,4,2,balanced,0.1964851188659668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,2,power_law_1.01,0.04521215915679931
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,4,2,balanced,0.1974937629699707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,2,power_law_1.01,0.06629504203796387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,2,power_law_1.01,0.08192895889282227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,2,power_law_1.01,3.8092095947265627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,2,power_law_1.01,0.12485759735107423
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,2,power_law_1.01,0.17107328414916992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,2,power_law_1.01,0.21025407791137693
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,2,power_law_1.01,0.24762624740600586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,2,power_law_1.01,0.25679487228393555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,2,power_law_1.01,0.26658304214477535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,2,power_law_1.01,0.2697599983215332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,2,power_law_1.01,0.2812172889709473
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,2,power_law_1.01,0.31408384323120114
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,2,power_law_1.01,0.33195648193359373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,2,power_law_1.01,0.33566463470458985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,2,power_law_1.01,0.3426662445068359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,2,power_law_1.01,4.608582458496094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,2,power_law_1.01,0.3621388626098633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,2,power_law_1.01,0.3879935836791992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,2,power_law_1.01,0.4181222534179687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,2,power_law_1.01,0.46972030639648443
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,2,power_law_1.01,0.35055358886718746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,2,power_law_1.01,0.545557746887207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,2,power_law_1.01,0.6645069122314453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,2,power_law_1.01,0.7743052673339844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,2,power_law_1.01,1.0474329376220703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,4,2,power_law_1.01,0.0439628791809082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,2,power_law_1.01,1.317063751220703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,4,2,power_law_1.01,0.0589734411239624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,4,2,power_law_1.01,0.06959104061126709
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,4,2,power_law_1.01,0.08354304313659669
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,2,power_law_1.01,1.8358131408691407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,4,2,power_law_1.01,0.10516736030578613
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,4,2,power_law_1.01,0.1249459171295166
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,4,2,power_law_1.01,0.1605721664428711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,4,2,power_law_1.01,0.16856960296630857
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,4,2,power_law_1.01,0.1715328025817871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,2,power_law_1.01,2.3240447998046876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,4,2,power_law_1.01,0.17242624282836913
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,4,2,power_law_1.01,0.1763942337036133
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,4,2,power_law_1.01,0.19068288803100586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,4,2,power_law_1.01,0.19568384170532227
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,4,2,power_law_1.01,0.2037913513183594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,4,2,power_law_1.01,0.1509721565246582
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,4,2,power_law_1.01,0.20310400009155272
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,4,2,power_law_1.01,0.2183705520629883
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,4,2,balanced,0.434205436706543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,4,2,balanced,0.7127474975585938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,4,2,balanced,0.491223030090332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,4,2,balanced,0.9342438507080079
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,4,2,power_law_1.01,0.2491059112548828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,4,2,power_law_1.01,0.27471872329711916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,4,2,power_law_1.01,0.32916351318359377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,2,power_law_1.2,0.07526912212371825
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,4,2,power_law_1.01,0.3909081649780274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,2,power_law_1.2,0.09356160163879394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,2,power_law_1.2,0.1076582431793213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,2,power_law_1.2,0.16351360321044922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,2,power_law_1.2,0.23130367279052733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,2,power_law_1.2,0.3154969596862793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,2,power_law_1.2,0.3897740936279297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,2,power_law_1.2,0.4058265686035156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,2,power_law_1.2,0.4058879852294922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,2,power_law_1.2,0.4302822494506836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,2,power_law_1.2,0.4464281463623047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,2,power_law_1.2,0.5049356842041016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,4,2,power_law_1.01,0.5085491180419922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,4,2,power_law_1.01,0.6753523254394531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,2,power_law_1.2,0.5211724853515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,2,power_law_1.2,0.5532326507568359
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,4,2,power_law_1.01,0.8422745513916017
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,2,power_law_1.2,0.561082878112793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,2,power_law_1.2,0.6084390258789062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,2,power_law_1.2,0.6406208038330078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,2,power_law_1.2,0.6815487670898437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,4,2,power_law_1.01,1.1687296295166016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,2,power_law_1.2,0.7285517120361329
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,2,power_law_1.2,0.8780902099609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,2,power_law_1.2,0.992958755493164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,2,power_law_1.2,1.2760806274414063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,2,power_law_1.2,0.046521601676940916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,2,power_law_1.2,1.5659020996093749
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,2,power_law_1.2,0.0642303991317749
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,2,power_law_1.2,0.08007424354553223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,2,power_law_1.2,0.10515071868896483
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,2,power_law_1.2,2.1510092163085934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,2,power_law_1.2,0.16204416275024414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,2,power_law_1.2,0.18216192245483398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,2,power_law_1.2,2.583098754882813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,2,power_law_1.2,0.2558464050292969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,2,power_law_1.2,0.25448959350585937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,2,power_law_1.2,0.26141183853149413
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,2,power_law_1.2,0.27942527770996095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,2,power_law_1.2,0.3212979125976562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,2,power_law_1.2,0.33514240264892575
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,2,power_law_1.2,0.33998081207275394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,2,power_law_1.2,3.7297317504882814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,2,power_law_1.2,0.34873855590820313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,2,power_law_1.2,0.36349441528320314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,2,power_law_1.2,0.37246593475341794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,2,power_law_1.2,0.27700479507446285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,2,power_law_1.2,0.401739501953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,2,power_law_1.2,0.437204475402832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,2,power_law_1.2,0.5059878540039062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,2,power_law_1.2,0.5571494293212891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,2,power_law_1.2,0.6802342224121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,2,power_law_1.2,5.083091125488282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,2,power_law_1.2,0.8060428619384765
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,4,2,power_law_1.2,0.043738880157470704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,2,power_law_1.2,1.0953548431396485
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,4,2,power_law_1.2,0.05812352180480958
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,2,power_law_1.2,1.3608717346191406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,4,2,power_law_1.2,0.06678400039672852
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,4,2,power_law_1.2,0.0772108793258667
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,4,2,power_law_1.2,0.1022489643096924
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,4,2,power_law_1.2,0.10926976203918456
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,4,2,power_law_1.2,0.15141375541687013
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,4,2,power_law_1.2,0.1613363265991211
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,4,2,power_law_1.2,0.1663052749633789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,2,power_law_1.2,1.9062489318847657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,4,2,power_law_1.2,0.1729292869567871
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,4,2,power_law_1.2,0.1804262351989746
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,4,2,power_law_1.2,0.1923072052001953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,4,2,power_law_1.2,0.20089216232299806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,2,power_law_1.2,2.36756103515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,4,2,power_law_1.2,0.210119686126709
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,4,2,power_law_1.2,0.21048959732055664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,4,2,power_law_1.2,0.23206016540527347
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,4,2,power_law_1.2,0.28100095748901366
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,4,2,power_law_1.2,0.17529983520507814
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,4,2,power_law_1.01,1.6937689208984374
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,4,2,power_law_1.01,2.403621063232422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,4,2,power_law_1.2,0.324060173034668
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,4,2,power_law_1.2,0.3885440063476563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,4,2,power_law_1.2,0.47259265899658204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,8,1,balanced,0.0664793586730957
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,4,2,power_law_1.2,0.6872255706787109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,8,1,balanced,0.08736448287963867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,4,2,power_law_1.2,0.9179788970947266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,8,1,balanced,0.1423468780517578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,4,2,power_law_1.2,1.207104034423828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,8,1,balanced,0.4870240020751953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,8,1,balanced,0.4905913543701172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,8,1,balanced,0.4888070297241211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,8,1,balanced,0.05216127872467041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,8,1,balanced,0.2521785545349121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,8,1,balanced,0.4864684677124023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,8,1,balanced,0.45171264648437504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,8,1,balanced,0.47189182281494146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,8,1,balanced,0.45510272979736327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,8,1,balanced,0.48008384704589846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,8,1,balanced,0.458671989440918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,8,1,balanced,0.4634700775146484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,8,1,balanced,0.4710681533813476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,8,1,balanced,0.552347526550293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,8,1,balanced,0.5690617752075195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,8,1,balanced,0.6358015823364258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,8,1,balanced,0.6831597137451172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,4,2,power_law_1.2,1.5368511962890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,4,2,power_law_1.2,2.6456524658203127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,8,1,balanced,0.7786483001708985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,8,1,balanced,0.04838399887084961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,8,1,balanced,0.06142784118652343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,8,1,balanced,0.08474240303039551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,8,1,balanced,0.13034432411193847
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,8,1,balanced,0.2239315223693848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,8,1,balanced,0.41496768951416013
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,8,1,balanced,0.41710849761962887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,8,1,balanced,0.4189452743530273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,8,1,balanced,0.8406649780273436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,8,1,balanced,0.9033081817626953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,8,1,balanced,0.42244991302490237
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,8,1,balanced,0.42512577056884765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,8,1,balanced,0.43024768829345705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,8,1,balanced,1.3654771423339844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,8,1,balanced,0.43477439880371094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,8,1,balanced,0.43773056030273433
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,8,1,balanced,0.4450579071044922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,8,1,balanced,0.4942047882080078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,8,1,balanced,1.4930181884765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,8,1,balanced,0.49938110351562504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,8,1,balanced,0.506949119567871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,8,1,balanced,0.5419327926635742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,8,1,balanced,0.5753311920166015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,8,1,balanced,0.6123500823974609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,8,1,balanced,0.6830278778076172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,8,1,balanced,0.5228633499145507
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,4,2,power_law_1.2,4.427694091796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,8,1,balanced,0.7560825347900391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,8,1,balanced,2.127342071533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,8,1,balanced,1.1540633392333983
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,8,1,balanced,1.3124473571777344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,8,1,balanced,1.8905241394042966
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,8,1,balanced,2.7038861083984376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,8,1,balanced,2.4374847412109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,8,1,balanced,0.06848959922790528
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,8,1,balanced,0.07955904006958008
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,8,1,balanced,0.06429567813873291
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,8,1,balanced,0.10610624313354493
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,8,1,balanced,0.15744832038879394
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,8,1,balanced,0.25405311584472656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,8,1,balanced,0.2544998359680176
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,8,1,balanced,0.25609344482421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,8,1,balanced,0.257708797454834
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,8,1,balanced,0.2518412780761719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,8,1,balanced,0.2519801521301269
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,8,1,balanced,0.2591391944885254
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,8,1,balanced,0.26203264236450197
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,8,1,balanced,0.25341375350952144
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,8,1,balanced,0.26415103912353516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,8,1,balanced,0.26778879165649416
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,8,1,balanced,0.28528703689575197
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,8,1,balanced,0.2755251121520996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,8,1,balanced,0.29834239959716796
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,8,1,balanced,0.3224095916748047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,8,1,balanced,0.3480006408691406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,8,1,balanced,0.3992486572265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,8,1,power_law_1.01,0.0668991994857788
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,8,1,balanced,0.4504121780395508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,8,1,power_law_1.01,0.08824576377868652
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,8,1,power_law_1.01,0.14271936416625977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,8,1,power_law_1.01,0.1946246337890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,8,1,power_law_1.01,0.3021299171447754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,8,1,power_law_1.01,0.35044288635253906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,8,1,power_law_1.01,0.3673721694946289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,8,1,power_law_1.01,0.3735513687133789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,8,1,power_law_1.01,0.38800319671630856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,8,1,power_law_1.01,0.40191486358642575
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,8,1,power_law_1.01,0.4122489547729492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,8,1,power_law_1.01,0.4213280105590821
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,8,1,power_law_1.01,0.049537281990051266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,8,1,power_law_1.01,0.43702529907226567
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,8,1,power_law_1.01,0.4973043060302734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,8,1,power_law_1.01,0.5169670486450195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,8,1,power_law_1.01,0.5344908905029297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,8,1,power_law_1.01,0.5962374496459961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,8,1,power_law_1.01,0.7428083038330078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,8,1,power_law_1.01,0.9021119689941406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,8,1,power_law_1.01,1.0024671936035157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,8,1,power_law_1.01,1.1082694244384765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,8,1,power_law_1.01,1.4243865966796876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,8,1,power_law_1.01,1.7493753051757814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,8,1,power_law_1.01,0.04761407852172851
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,8,1,power_law_1.01,0.06088191986083984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,8,1,power_law_1.01,0.08616959571838378
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,8,1,power_law_1.01,2.3913926696777343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,8,1,power_law_1.01,0.12865663528442384
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,8,1,power_law_1.01,0.6255724716186524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,8,1,power_law_1.01,0.1964467239379883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,8,1,power_law_1.01,0.26669824600219727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,8,1,power_law_1.01,0.3323628616333008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,8,1,power_law_1.01,3.0119219970703126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,8,1,power_law_1.01,0.3451551818847656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,8,1,power_law_1.01,0.35096767425537106
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,8,1,power_law_1.01,0.3551641464233398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,8,1,power_law_1.01,0.3685657501220703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,8,1,power_law_1.01,0.37833728790283205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,8,1,power_law_1.01,0.3925017547607422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,8,1,power_law_1.01,0.3842144012451172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,8,1,power_law_1.01,0.4500147247314453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,8,1,power_law_1.01,0.45122814178466797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,8,1,power_law_1.01,0.47875198364257815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,8,1,power_law_1.01,0.5013926315307617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,8,1,power_law_1.01,0.5317337417602539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,8,1,power_law_1.01,0.5903366470336915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,8,1,power_law_1.01,0.6629503631591798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,8,1,power_law_1.01,0.7865312194824219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,8,1,power_law_1.01,0.9177766418457031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,8,1,power_law_1.01,1.1996953582763672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,8,1,power_law_1.01,1.4908230590820313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,8,1,power_law_1.01,0.06380864143371581
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,8,1,power_law_1.01,2.036420440673828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,8,1,power_law_1.01,0.06988224029541015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,8,1,power_law_1.01,2.5989312744140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,8,1,power_law_1.01,0.106048641204834
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,8,1,power_law_1.01,0.1325964832305908
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,8,1,power_law_1.01,0.08433983802795411
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,8,1,power_law_1.01,0.1577023983001709
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,8,1,power_law_1.01,0.19965824127197265
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,8,1,power_law_1.01,0.2103334426879883
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,8,1,power_law_1.01,0.21592384338378906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,8,1,power_law_1.01,0.21763135910034181
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,8,1,power_law_1.01,0.22765632629394533
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,8,1,power_law_1.01,0.23475135803222655
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,8,1,power_law_1.01,0.2398975944519043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,8,1,power_law_1.01,0.24817792892456056
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,8,1,power_law_1.01,0.2626976013183594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,8,1,power_law_1.01,0.26971328735351563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,8,1,power_law_1.01,0.29349184036254883
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,8,1,balanced,0.6985247802734376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,8,1,balanced,0.8005388641357423
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,8,1,balanced,1.1690969848632813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,8,1,balanced,1.542270050048828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,8,1,power_law_1.01,0.3315065765380859
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,8,1,power_law_1.01,0.37794689178466795
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,8,1,power_law_1.01,0.4709363174438477
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,8,1,power_law_1.01,0.5711155319213868
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,8,1,power_law_1.01,0.761391372680664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,8,1,power_law_1.01,0.975638427734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,8,1,power_law_1.2,0.08903039932250977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,8,1,power_law_1.2,0.134017276763916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,8,1,power_law_1.2,0.18467264175415038
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,8,1,power_law_1.2,0.050762238502502444
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,8,1,power_law_1.01,1.3885395812988282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,8,1,power_law_1.2,0.2804351997375488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,8,1,power_law_1.2,0.3437139129638672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,8,1,power_law_1.2,0.35900543212890623
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,8,1,power_law_1.01,1.8232794189453123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,8,1,power_law_1.2,0.36798015594482425
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,8,1,power_law_1.2,0.38698368072509765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,8,1,power_law_1.2,0.3970809555053711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,8,1,power_law_1.2,0.4099935913085938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,8,1,power_law_1.2,0.06579904079437256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,8,1,power_law_1.2,0.4472851181030274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,8,1,power_law_1.2,0.5053241729736329
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,8,1,power_law_1.2,0.5326515197753906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,8,1,power_law_1.2,0.5459411239624024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,8,1,power_law_1.2,0.6383327865600587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,8,1,power_law_1.2,0.4231129455566406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,8,1,power_law_1.2,0.7603903961181641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,8,1,power_law_1.2,0.9295340728759764
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,8,1,power_law_1.2,1.032047348022461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,8,1,power_law_1.2,1.2422918701171874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,8,1,power_law_1.2,1.4521522521972656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,8,1,power_law_1.2,0.04813568115234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,8,1,power_law_1.2,0.06137216091156006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,8,1,power_law_1.2,0.08506431579589843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,8,1,power_law_1.2,1.7830905151367187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,8,1,power_law_1.2,0.12138943672180176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,8,1,power_law_1.2,0.18679616928100587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,8,1,power_law_1.2,0.2553273582458496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,8,1,power_law_1.2,2.4181056213378906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,8,1,power_law_1.2,0.32937854766845703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,8,1,power_law_1.2,0.3395507049560547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,8,1,power_law_1.2,0.34805057525634764
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,8,1,power_law_1.2,0.6330060958862305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,8,1,power_law_1.2,0.35572673797607424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,8,1,power_law_1.2,0.3652659225463867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,8,1,power_law_1.2,0.3762489700317383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,8,1,power_law_1.2,3.0514337158203126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,8,1,power_law_1.2,0.3851007843017578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,8,1,power_law_1.2,0.4033465576171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,8,1,power_law_1.2,0.45842815399169917
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,8,1,power_law_1.2,0.47207809448242183
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,8,1,power_law_1.2,0.4844012832641601
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,8,1,power_law_1.2,0.51687744140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,8,1,power_law_1.2,0.5470175933837891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,8,1,power_law_1.2,0.6221516799926758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,8,1,power_law_1.2,0.6810784149169922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,8,1,power_law_1.2,0.8225631713867188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,8,1,power_law_1.2,0.9660998535156249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,8,1,power_law_1.2,1.2151232147216797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,8,1,power_law_1.2,1.513360595703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,8,1,power_law_1.2,0.06395648002624513
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,8,1,power_law_1.2,0.0707539176940918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,8,1,power_law_1.2,2.0736985778808594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,8,1,power_law_1.2,0.08220671653747559
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,8,1,power_law_1.2,0.102424955368042
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,8,1,power_law_1.2,0.12573247909545898
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,8,1,power_law_1.2,0.15203583717346192
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,8,1,power_law_1.2,0.1970675277709961
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,8,1,power_law_1.2,0.2081203269958496
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,8,1,power_law_1.2,0.21261247634887695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,8,1,power_law_1.2,0.22046464920043946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,8,1,power_law_1.2,2.633194885253906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,8,1,power_law_1.2,0.22821760177612305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,8,1,power_law_1.2,0.23679168701171874
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,8,1,power_law_1.2,0.25965759277343753
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,8,1,power_law_1.2,0.2751865577697754
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,8,1,power_law_1.2,0.2888051223754883
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,8,1,power_law_1.2,0.24442367553710936
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,8,1,power_law_1.01,2.6856344604492186
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,8,1,power_law_1.01,3.7047308349609374
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,8,1,power_law_1.2,0.32131649017333985
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,8,1,power_law_1.2,0.3815103912353516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,8,1,power_law_1.2,0.45218944549560547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,8,1,power_law_1.2,0.5947776031494141
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,8,1,power_law_1.2,0.7505632019042968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,8,balanced,0.03670016050338745
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,8,1,power_law_1.2,1.0587673950195313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,8,balanced,0.07170944213867188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,8,balanced,0.12431488037109376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,8,balanced,0.2570060729980469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,8,balanced,0.08630911827087402
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,8,balanced,0.4701875305175781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,8,balanced,0.5040166473388672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,8,balanced,0.5217177581787109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,8,balanced,0.5165516662597656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,8,balanced,0.5351168060302735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,8,balanced,0.5249971389770508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,8,balanced,0.5404940795898437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,8,balanced,0.5478438568115235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,8,balanced,0.5647564697265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,8,balanced,0.5688307189941406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,8,balanced,0.5749785614013672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,8,balanced,0.5781580734252929
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,8,balanced,0.6380863952636718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,8,balanced,0.6233433532714844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,8,balanced,0.7030335998535155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,8,balanced,0.7850726318359376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,8,balanced,1.1158386993408205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,8,balanced,1.2807603454589844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,8,balanced,1.8184909057617187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,8,balanced,2.4266316223144533
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,8,1,power_law_1.2,1.3979122924804688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,8,balanced,0.03782144069671631
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,8,balanced,0.05780096054077148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,8,balanced,0.08011136054992676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,8,balanced,0.1328831958770752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,8,balanced,3.8600012207031247
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,8,balanced,0.03880448102951049
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,8,balanced,0.23991680145263672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,8,balanced,0.261527042388916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,8,balanced,0.2672243118286133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,8,balanced,0.2640524864196777
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,8,balanced,0.2685491180419922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,8,balanced,0.2587827110290527
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,8,balanced,0.260392951965332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,8,balanced,0.26153472900390623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,8,balanced,0.26330240249633785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,8,balanced,0.26755327224731446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,8,balanced,5.237241821289063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,8,balanced,0.2712038421630859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,8,balanced,0.293438720703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,8,balanced,0.3212556838989258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,8,balanced,0.33560192108154296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,8,balanced,0.3844364929199219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,8,1,power_law_1.2,2.167902679443359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,8,balanced,0.42975486755371095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,8,balanced,0.5290995025634766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,8,balanced,0.6351513671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,8,balanced,0.8686592102050781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,8,balanced,1.0538713836669922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,8,balanced,1.5299533081054686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,8,balanced,1.9754739379882813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,8,1,power_law_1.2,2.9987237548828127
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,1,8,balanced,0.024216320514678955
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,1,8,balanced,0.1354316806793213
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,1,8,balanced,0.1356595230102539
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,1,8,balanced,0.13586175918579102
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,1,8,balanced,0.025114879608154294
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,1,8,balanced,0.13632896423339844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,1,8,balanced,0.1364889621734619
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,1,8,balanced,0.13692159652709962
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,1,8,balanced,0.050658559799194335
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,1,8,balanced,0.13649791717529297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,1,8,balanced,0.13780991554260252
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,1,8,balanced,0.031943678855895996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,1,8,balanced,0.13737728118896483
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,1,8,balanced,0.1375385570526123
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,1,8,balanced,0.1477683162689209
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,1,8,balanced,0.1390182399749756
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,1,8,balanced,0.14072575569152831
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,1,8,balanced,0.14252415657043457
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,1,8,balanced,0.15066495895385743
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,1,8,balanced,0.15999360084533693
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,1,8,balanced,0.16679296493530274
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,1,8,balanced,0.08133503913879395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,8,power_law_1.01,0.12390144348144531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,8,power_law_1.01,0.25521408081054686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,8,power_law_1.01,0.3026367950439453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,8,power_law_1.01,0.35976318359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,8,power_law_1.01,0.3964031982421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,8,power_law_1.01,0.4099891281127929
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,8,power_law_1.01,0.49136383056640626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,8,power_law_1.01,0.456622085571289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,8,power_law_1.01,0.45216129302978514
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,8,power_law_1.01,0.4682777786254883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,8,power_law_1.01,0.46039039611816407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,8,power_law_1.01,0.48748542785644533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,8,power_law_1.01,0.488353271484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,8,power_law_1.01,0.5191974258422851
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,8,power_law_1.01,0.5456934356689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,1,8,balanced,0.1482585620880127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,8,power_law_1.01,0.5575167846679687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,8,power_law_1.01,0.5641011047363281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,8,power_law_1.01,0.6338304138183594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,8,power_law_1.01,0.6566989135742187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,8,power_law_1.01,0.8461952209472656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,8,power_law_1.01,0.9970060729980469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,8,power_law_1.01,1.2649523162841798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,8,power_law_1.01,1.5926873779296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,8,power_law_1.01,2.142420501708984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,8,power_law_1.01,2.9723519897460937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,8,power_law_1.01,4.081489868164063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,8,power_law_1.01,0.08154496192932129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,8,power_law_1.01,0.130763521194458
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,8,power_law_1.01,0.15635456085205077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,8,power_law_1.01,0.19094655990600587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,8,power_law_1.01,0.20756607055664061
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,1,8,balanced,0.2420915222167969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,8,power_law_1.01,0.237445125579834
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,1,8,balanced,0.4557324981689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,8,power_law_1.01,5.50779541015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,8,power_law_1.01,0.2196339225769043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,8,power_law_1.01,0.24517120361328124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,8,power_law_1.01,0.23641984939575195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,8,power_law_1.01,0.25249151229858396
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,1,8,balanced,0.25807743072509765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,8,power_law_1.01,0.2563520050048828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,8,power_law_1.01,0.25673599243164064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,8,power_law_1.01,0.25196416854858394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,8,power_law_1.01,0.2550284767150879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,8,power_law_1.01,0.2482828712463379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,8,power_law_1.01,0.2506342315673828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,8,power_law_1.01,0.25651840209960936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,8,power_law_1.01,0.29789056777954104
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,8,power_law_1.01,0.328776969909668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,8,power_law_1.01,0.3714368057250977
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,1,8,balanced,0.3577536010742187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,8,power_law_1.01,0.434469108581543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,8,power_law_1.01,0.5628582382202149
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,8,power_law_1.01,0.6693170928955079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,8,power_law_1.01,0.9218764495849608
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,1,8,power_law_1.01,0.0514086389541626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,1,8,power_law_1.01,0.08196224212646484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,1,8,power_law_1.01,0.09389439582824707
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,1,8,power_law_1.01,0.0981657600402832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,8,power_law_1.01,1.1867046356201172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,1,8,power_law_1.01,0.11640576362609863
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,1,8,power_law_1.01,0.11338879585266112
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,8,power_law_1.01,1.6779405212402345
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,1,8,power_law_1.01,0.12822015762329103
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,1,8,power_law_1.01,0.12183039665222169
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,8,power_law_1.01,2.0330099487304687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,1,8,power_law_1.01,0.1185331153869629
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,1,8,power_law_1.01,0.1303116798400879
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,1,8,power_law_1.01,0.12444800376892089
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,1,8,power_law_1.01,0.12635519981384277
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,1,8,power_law_1.01,0.12190591812133789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,1,8,power_law_1.01,0.13459199905395508
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,1,8,power_law_1.01,0.13881983757019042
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,1,8,power_law_1.01,0.1355174446105957
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,1,8,power_law_1.01,0.13804415702819822
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,1,8,power_law_1.01,0.15169919967651366
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,1,8,power_law_1.01,0.18534143447875978
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,1,8,power_law_1.01,0.1578880023956299
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,1,8,power_law_1.01,0.20921472549438475
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,1,8,power_law_1.01,0.2646297645568848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,8,power_law_1.2,0.12121343612670898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,8,power_law_1.2,0.2510668754577637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,8,power_law_1.2,0.2589363288879395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,1,8,power_law_1.01,0.31580032348632814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,8,power_law_1.2,0.3477849578857422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,8,power_law_1.2,0.36463489532470705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,8,power_law_1.2,0.41801216125488283
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,8,power_law_1.2,0.4900352096557617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,8,power_law_1.2,0.4465625762939453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,8,power_law_1.2,0.45974784851074213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,8,power_law_1.2,0.4428121566772461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,8,power_law_1.2,0.46285568237304686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,8,power_law_1.2,0.4617446517944336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,8,power_law_1.2,0.4671244812011719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,8,power_law_1.2,0.5241203308105469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,8,power_law_1.2,0.551148796081543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,8,power_law_1.2,0.563191032409668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,8,power_law_1.2,0.5820288085937501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,8,power_law_1.2,0.6470976257324219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,8,power_law_1.2,0.6995532989501954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,8,power_law_1.2,0.8809382629394531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,8,power_law_1.2,1.0222771453857422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,8,power_law_1.2,1.3582080078124998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,8,power_law_1.2,1.7094989013671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,8,power_law_1.2,2.5474636840820315
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,8,power_law_1.2,3.53525390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,8,power_law_1.2,4.454127502441406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,8,power_law_1.2,0.07924736022949219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,8,power_law_1.2,0.13059071540832518
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,8,power_law_1.2,0.13100671768188477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,8,power_law_1.2,0.18569856643676758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,8,power_law_1.2,0.19330432891845703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,8,power_law_1.2,0.2127270317077637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,8,power_law_1.2,6.492946166992188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,8,power_law_1.2,0.23737728118896487
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,8,power_law_1.2,0.23130752563476564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,8,power_law_1.2,0.25090431213378905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,8,power_law_1.2,0.24558591842651367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,8,power_law_1.2,0.2619392013549805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,8,power_law_1.2,0.24086912155151366
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,8,power_law_1.2,0.24769792556762696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,8,power_law_1.2,0.25436800003051757
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,8,power_law_1.2,0.2479782485961914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,8,power_law_1.2,0.25292928695678707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,8,power_law_1.2,0.34271488189697263
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,8,power_law_1.2,0.3894566345214844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,8,power_law_1.2,0.460682258605957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,8,power_law_1.2,0.5863782501220702
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,8,power_law_1.2,0.6961497497558594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,1,8,power_law_1.01,0.38688385009765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,8,power_law_1.2,1.015788803100586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,1,8,power_law_1.01,0.7819302368164063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,1,8,power_law_1.01,0.5365414428710937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,1,8,power_law_1.2,0.05181568145751954
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,1,8,power_law_1.2,0.078854398727417
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,1,8,power_law_1.2,0.0811302375793457
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,1,8,power_law_1.01,1.0534130859375002
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,1,8,power_law_1.2,0.09534591674804688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,1,8,power_law_1.2,0.10659199714660646
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,1,8,power_law_1.2,0.12157440185546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,8,power_law_1.2,1.8390719604492187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,1,8,power_law_1.2,0.11167743682861328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,8,power_law_1.2,1.3032153320312498
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,1,8,power_law_1.2,0.11926655769348145
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,1,8,power_law_1.2,0.1331110382080078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,1,8,power_law_1.2,0.12576255798339844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,8,power_law_1.2,2.631724853515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,1,8,power_law_1.2,0.12965375900268555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,1,8,power_law_1.2,0.1344115161895752
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,1,8,power_law_1.2,0.1131328010559082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,1,8,power_law_1.2,0.1341273593902588
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,1,8,power_law_1.2,0.14066944122314454
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,1,8,power_law_1.2,0.14402303695678711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,1,8,power_law_1.2,0.1664896011352539
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,1,8,power_law_1.2,0.13066111564636232
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,8,power_law_1.2,0.2933913612365723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,8,power_law_1.2,0.2663039970397949
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,1,8,power_law_1.2,0.22640127182006836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,1,8,power_law_1.2,0.1594585609436035
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,1,8,power_law_1.2,0.3178495979309082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,1,8,power_law_1.2,0.23961984634399416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,2,4,balanced,0.0348198390007019
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,2,4,balanced,0.0666534423828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,2,4,balanced,0.08379648208618165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,2,4,balanced,0.12446847915649414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,1,8,power_law_1.2,0.3922662353515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,2,4,balanced,0.2539583969116211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,2,4,balanced,0.46181377410888674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,2,4,balanced,0.4573708724975586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,2,4,balanced,0.457391357421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,2,4,balanced,0.48893825531005863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,2,4,balanced,0.5187097549438476
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,1,8,power_law_1.2,0.5780044937133789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,2,4,balanced,0.5119667053222656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,2,4,balanced,0.5357478332519532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,2,4,balanced,0.5219660949707031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,2,4,balanced,0.548430061340332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,2,4,balanced,0.5439219284057617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,2,4,balanced,0.563583984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,2,4,balanced,0.5849536132812501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,2,4,balanced,0.5980108642578126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,2,4,balanced,0.6610163116455079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,2,4,balanced,0.735937271118164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,2,4,balanced,0.5562355041503906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,2,4,balanced,1.0356198120117188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,2,4,balanced,1.190456314086914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,2,4,balanced,0.035024640560150144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,2,4,balanced,0.0378982400894165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,2,4,balanced,1.6795750427246094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,2,4,balanced,0.05961728096008301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,2,4,balanced,0.0894655990600586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,2,4,balanced,0.143886079788208
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,1,8,power_law_1.2,0.7919808197021484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,2,4,balanced,2.1226841735839845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,2,4,balanced,0.24431360244750974
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,2,4,balanced,0.24510080337524415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,2,4,balanced,0.2656025505065918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,2,4,balanced,0.2666828727722168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,2,4,balanced,0.243057918548584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,2,4,balanced,0.2675712013244629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,2,4,balanced,0.27120256423950195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,2,4,balanced,0.272808952331543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,2,4,balanced,0.275549430847168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,2,4,balanced,0.2779571151733399
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,1,8,power_law_1.2,1.1161420440673828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,2,4,balanced,3.101953430175781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,2,4,balanced,0.29663360595703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,2,4,balanced,0.3248102569580078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,2,4,balanced,0.3357056045532226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,2,4,balanced,0.42892929077148434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,2,4,balanced,0.48834049224853515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,2,4,balanced,0.3805184173583984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,2,4,balanced,0.5851993560791016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,2,4,balanced,4.115586547851562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,2,4,balanced,0.2813862419128418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,2,4,balanced,0.8308633422851562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,2,4,balanced,0.9867430114746094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,2,4,balanced,1.4248768615722658
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,2,4,balanced,1.8463027954101563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,1,8,power_law_1.2,1.4505856323242188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,2,4,balanced,0.030791680812835693
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,2,4,balanced,0.14468095779418946
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,2,4,balanced,0.14476927757263183
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,2,4,balanced,0.05602176189422607
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,2,4,balanced,0.1450150394439697
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,2,4,balanced,0.14518143653869628
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,2,4,balanced,0.03754240036010742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,2,4,balanced,0.14600576400756834
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,2,4,balanced,0.14720255851745606
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,2,4,balanced,0.14406399726867675
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,2,4,balanced,0.14347392082214355
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,2,4,balanced,0.14820096015930176
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,2,4,balanced,0.14813440322875976
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,2,4,balanced,0.15034111976623535
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,2,4,balanced,0.15363327980041502
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,2,4,balanced,0.15600128173828126
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,2,4,balanced,0.16375808715820311
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,2,4,balanced,0.18181631088256836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,2,4,balanced,0.1931635284423828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,2,4,balanced,0.02856575965881348
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,2,4,balanced,0.08621312141418457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,2,4,power_law_1.01,0.08265215873718261
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,2,4,balanced,0.1694438362121582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,2,4,power_law_1.01,0.12341504096984864
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,2,4,power_law_1.01,0.16738304138183593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,2,4,power_law_1.01,0.23002496719360352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,2,4,power_law_1.01,0.28797824859619137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,2,4,power_law_1.01,0.31477376937866214
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,2,4,power_law_1.01,0.375654411315918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,2,4,power_law_1.01,0.36503681182861325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,2,4,power_law_1.01,0.416124153137207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,2,4,power_law_1.01,0.4298739242553711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,2,4,power_law_1.01,0.4363148880004883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,2,4,power_law_1.01,0.45017471313476565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,2,4,power_law_1.01,0.4254553604125976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,2,4,power_law_1.01,0.5127910232543945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,2,4,power_law_1.01,0.528980484008789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,2,4,power_law_1.01,0.5428505706787109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,2,4,power_law_1.01,0.5621977615356445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,2,4,power_law_1.01,0.5917427062988281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,2,4,power_law_1.01,0.62587646484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,2,4,power_law_1.01,0.7850418853759765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,2,4,power_law_1.01,0.8836544036865235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,2,4,power_law_1.01,1.1969459533691407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,2,4,power_law_1.01,1.4026176452636718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,2,4,power_law_1.01,1.9957849121093751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,2,4,power_law_1.01,2.4120999145507813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,2,4,power_law_1.01,0.05788032054901123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,2,4,power_law_1.01,0.07916672229766845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,2,4,power_law_1.01,0.11028096199035645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,2,4,power_law_1.01,3.470743103027344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,2,4,power_law_1.01,0.1400268840789795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,2,4,power_law_1.01,0.16425088882446287
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,2,4,power_law_1.01,0.18929792404174806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,2,4,power_law_1.01,0.22755199432373047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,2,4,power_law_1.01,0.23024511337280273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,2,4,power_law_1.01,0.2535513687133789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,2,4,power_law_1.01,0.24951040267944338
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,2,4,power_law_1.01,0.2563532829284668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,2,4,power_law_1.01,0.2508211135864258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,2,4,power_law_1.01,0.2523404884338379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,2,4,power_law_1.01,4.892886962890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,2,4,power_law_1.01,0.26823680877685546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,2,4,power_law_1.01,0.2693683242797852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,2,4,power_law_1.01,0.27938304901123046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,2,4,power_law_1.01,0.31893760681152344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,2,4,power_law_1.01,0.33557376861572263
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,2,4,power_law_1.01,0.3917337417602539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,2,4,power_law_1.01,0.43503360748291015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,2,4,power_law_1.01,0.27397247314453127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,2,4,power_law_1.01,0.5333055877685547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,2,4,power_law_1.01,0.6737702178955078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,2,4,power_law_1.01,0.8913471984863282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,2,4,power_law_1.01,1.1017740631103516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,2,4,power_law_1.01,0.040332798957824705
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,2,4,power_law_1.01,0.057715201377868654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,2,4,power_law_1.01,1.5336102294921876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,2,4,power_law_1.01,0.07100031852722169
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,2,4,power_law_1.01,0.08693632125854492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,2,4,power_law_1.01,1.9552883911132812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,2,4,power_law_1.01,0.096746244430542
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,2,4,power_law_1.01,0.11036800384521483
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,2,4,power_law_1.01,0.1262476825714111
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,2,4,power_law_1.01,0.13454719543457033
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,2,4,power_law_1.01,0.12850048065185546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,2,4,balanced,0.1445631980895996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,2,4,power_law_1.01,0.1303091239929199
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,2,4,power_law_1.01,0.13478400230407714
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,2,4,power_law_1.01,0.13357184410095216
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,2,4,power_law_1.01,0.13760895729064943
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,2,4,power_law_1.01,0.13785216331481934
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,2,4,power_law_1.01,0.13979392051696776
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,2,4,power_law_1.01,0.14187647819519042
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,2,4,balanced,0.4366630554199219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,2,4,balanced,0.28213632583618165
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,2,4,balanced,0.31007871627807615
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,2,4,balanced,0.5686502456665039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,2,4,power_law_1.01,0.15022080421447753
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,2,4,power_law_1.01,0.16404224395751954
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,2,4,power_law_1.01,0.20380672454833984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,2,4,power_law_1.01,0.1740774345397949
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,2,4,power_law_1.2,0.08198143959045409
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,2,4,power_law_1.2,0.12127360343933105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,2,4,power_law_1.2,0.1378816032409668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,2,4,power_law_1.2,0.20624767303466798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,2,4,power_law_1.2,0.255786247253418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,2,4,power_law_1.2,0.3002649688720703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,2,4,power_law_1.2,0.38255359649658205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,2,4,power_law_1.2,0.40500350952148434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,2,4,power_law_1.2,0.4419494247436523
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,2,4,power_law_1.01,0.252677116394043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,2,4,power_law_1.2,0.4400831985473633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,2,4,power_law_1.2,0.4469696044921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,2,4,power_law_1.01,0.3038566398620605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,2,4,power_law_1.2,0.4338751983642578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,2,4,power_law_1.01,0.5071974563598632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,2,4,power_law_1.2,0.4528704071044922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,2,4,power_law_1.01,0.37890430450439455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,2,4,power_law_1.2,0.47348480224609374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,2,4,power_law_1.2,0.5273535919189454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,2,4,power_law_1.2,0.546220817565918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,2,4,power_law_1.2,0.5610726547241212
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,2,4,power_law_1.2,0.6162969589233398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,2,4,power_law_1.2,0.670645751953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,2,4,power_law_1.2,0.7604684448242187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,2,4,power_law_1.2,0.9403008270263673
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,2,4,power_law_1.2,1.181644821166992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,2,4,power_law_1.2,1.4367372131347655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,2,4,power_law_1.2,0.05616000175476075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,2,4,power_law_1.2,0.07894911766052246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,2,4,power_law_1.2,2.0932313537597658
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,2,4,power_law_1.2,0.09727231979370117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,2,4,power_law_1.2,0.14755071640014647
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,2,4,power_law_1.2,2.715978088378906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,2,4,power_law_1.2,0.18357759475708008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,2,4,power_law_1.2,0.22060800552368165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,2,4,power_law_1.2,0.2281920051574707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,2,4,power_law_1.2,0.2517913627624512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,2,4,power_law_1.2,0.2456537628173828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,2,4,power_law_1.2,3.746085205078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,2,4,power_law_1.2,0.24587520599365234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,2,4,power_law_1.2,0.24166143417358396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,2,4,power_law_1.2,0.24221824645996093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,2,4,power_law_1.2,0.27184511184692384
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,2,4,power_law_1.2,0.2668902397155762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,2,4,power_law_1.2,0.2742412757873535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,2,4,power_law_1.2,0.28176767349243165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,2,4,power_law_1.2,0.3223283386230469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,2,4,power_law_1.2,0.3434521484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,2,4,power_law_1.2,0.40531967163085936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,2,4,power_law_1.2,0.4467993545532226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,2,4,power_law_1.2,4.9315621948242185
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,2,4,power_law_1.2,0.5525222396850585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,2,4,power_law_1.2,0.691572494506836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,2,4,power_law_1.2,0.9458393859863282
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,2,4,power_law_1.2,0.039146239757537844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,2,4,power_law_1.2,0.05805823802947998
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,2,4,power_law_1.2,0.06269696235656738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,2,4,power_law_1.2,1.6238233947753904
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,2,4,power_law_1.2,0.07674880027770996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,2,4,power_law_1.2,0.09060992240905762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,2,4,power_law_1.2,2.171735076904297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,2,4,power_law_1.2,0.10052096366882324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,2,4,power_law_1.2,1.1614617919921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,2,4,power_law_1.2,0.12429311752319336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,2,4,power_law_1.2,0.13636223793029784
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,2,4,power_law_1.2,0.13536128044128418
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,2,4,power_law_1.2,0.1292031955718994
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,2,4,power_law_1.2,0.1322060775756836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,2,4,power_law_1.2,0.13657216072082518
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,2,4,power_law_1.2,0.13674240112304686
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,2,4,power_law_1.2,0.13871359825134277
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,2,4,power_law_1.2,0.14467840194702147
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,2,4,power_law_1.2,0.14571776390075683
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,2,4,power_law_1.2,0.1606489562988281
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,2,4,power_law_1.2,0.17683584213256837
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,2,4,power_law_1.2,0.1328435230255127
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,2,4,power_law_1.01,0.6770585632324219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,2,4,power_law_1.01,1.1986278533935546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,2,4,power_law_1.01,0.9738854217529298
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,2,4,power_law_1.2,0.19117824554443358
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,2,4,power_law_1.2,0.23538688659667967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,4,2,balanced,0.05377920150756836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,2,4,power_law_1.2,0.32264190673828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,4,2,balanced,0.0653875207901001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,4,2,balanced,0.08353152275085449
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,4,2,balanced,0.12378368377685547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,2,4,power_law_1.2,0.40034305572509765
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,2,4,power_law_1.2,0.47145729064941405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,4,2,balanced,0.25749759674072265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,4,2,balanced,0.45938816070556643
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,4,2,balanced,0.45772159576416016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,4,2,balanced,0.45846527099609374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,4,2,balanced,0.45832702636718753
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,4,2,balanced,0.4592243194580078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,4,2,balanced,0.4622246551513672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,4,2,balanced,0.5188531112670899
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,4,2,balanced,0.5104102325439454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,4,2,balanced,0.5232729721069336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,4,2,balanced,0.5348044967651366
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,4,2,balanced,0.5393280029296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,4,2,balanced,0.5323955154418945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,4,2,balanced,0.5618304061889648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,4,2,balanced,0.5790604782104493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,4,2,balanced,0.6442316436767579
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,4,2,balanced,0.7131161499023437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,4,2,balanced,1.009890594482422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,4,2,balanced,1.1498675537109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,4,2,balanced,1.6429862976074219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,4,2,balanced,0.03872895956039429
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,4,2,balanced,2.04102783203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,4,2,balanced,0.04292479991912842
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,4,2,balanced,0.0934886360168457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,4,2,balanced,0.16471168518066406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,4,2,balanced,0.2724812889099121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,4,2,balanced,0.06288127899169922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,4,2,balanced,2.9651083374023437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,2,4,power_law_1.2,0.7444915008544922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,4,2,balanced,0.27536767959594727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,4,2,balanced,0.27940479278564456
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,4,2,balanced,0.27874687194824216
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,4,2,balanced,0.2787980842590332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,4,2,balanced,0.2823564720153809
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,4,2,balanced,0.3149299240112305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,4,2,balanced,0.3165465545654297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,4,2,balanced,3.9442764282226563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,4,2,balanced,0.320843505859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,4,2,balanced,0.32301185607910154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,4,2,balanced,0.3256639862060547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,2,4,power_law_1.2,1.1103974151611329
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,4,2,balanced,0.3387161636352539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,4,2,balanced,0.35113216400146485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,4,2,balanced,0.3760755157470703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,4,2,balanced,0.40396671295166015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,4,2,balanced,0.4506175994873047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,4,2,balanced,0.5187891387939454
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,2,4,power_law_1.2,1.3857037353515627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,4,2,balanced,0.593098258972168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,4,2,balanced,0.8623782348632811
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,4,2,balanced,1.005770263671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,4,2,balanced,1.4486604309082032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,4,2,balanced,1.8789389038085935
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,2,4,power_law_1.2,2.094115905761719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,4,2,balanced,0.049966077804565426
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,4,2,balanced,0.06835328102111816
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,4,2,balanced,0.1705356788635254
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,4,2,balanced,0.17082624435424804
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,4,2,balanced,0.1733760070800781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,4,2,balanced,0.17339263916015626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,4,2,balanced,0.1730393600463867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,4,2,balanced,0.10281344413757325
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,4,2,balanced,0.17608448028564455
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,4,2,balanced,0.17085311889648438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,4,2,balanced,0.17144704818725587
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,4,2,balanced,0.17532672882080078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,4,2,balanced,0.04129663944244385
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,4,2,balanced,0.17782400131225587
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,4,2,balanced,0.18766080856323242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,4,2,balanced,0.19385343551635742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,4,2,balanced,0.203057918548584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,4,2,balanced,0.2142681694030762
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,4,2,balanced,0.23669248580932617
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,4,2,balanced,0.2592972755432129
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,4,2,balanced,0.039470078945159914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,4,2,balanced,0.18272512435913085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,4,2,power_law_1.01,0.06480512142181397
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,4,2,power_law_1.01,0.08270591735839844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,4,2,power_law_1.01,0.1050649642944336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,4,2,power_law_1.01,0.15008895874023437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,4,2,power_law_1.01,0.21863040924072266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,4,2,power_law_1.01,0.2810611152648926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,4,2,power_law_1.01,0.3324544143676758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,4,2,power_law_1.01,0.3407820892333985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,4,2,power_law_1.01,0.342476806640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,4,2,power_law_1.01,0.35340415954589843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,4,2,power_law_1.01,0.36901248931884767
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,4,2,power_law_1.01,0.4198771286010742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,4,2,power_law_1.01,0.42100864410400385
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,4,2,power_law_1.01,0.4606937789916993
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,4,2,power_law_1.01,0.489483528137207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,4,2,power_law_1.01,0.5155174255371093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,4,2,power_law_1.01,0.5447103881835937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,4,2,power_law_1.01,0.5773452758789063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,4,2,power_law_1.01,0.6161395263671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,4,2,power_law_1.01,0.7302937316894531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,4,2,power_law_1.01,0.8478451538085938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,4,2,power_law_1.01,1.0489561462402344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,4,2,power_law_1.01,1.3261080932617186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,4,2,power_law_1.01,1.790330810546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,4,2,power_law_1.01,2.226588134765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,4,2,power_law_1.01,0.040293121337890626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,4,2,power_law_1.01,0.060823040008544924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,4,2,power_law_1.01,0.0772160005569458
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,4,2,power_law_1.01,3.2085131835937504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,4,2,power_law_1.01,0.12208383560180665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,4,2,power_law_1.01,0.1585318374633789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,4,2,power_law_1.01,0.19265535354614258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,4,2,power_law_1.01,0.22590591430664064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,4,2,power_law_1.01,0.2307571220397949
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,4,2,power_law_1.01,0.2427622413635254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,4,2,power_law_1.01,0.24418815612792968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,4,2,power_law_1.01,4.287061767578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,4,2,power_law_1.01,0.2531404876708984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,4,2,power_law_1.01,0.2725913619995117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,4,2,power_law_1.01,0.2720191955566406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,4,2,power_law_1.01,0.29507328033447267
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,4,2,power_law_1.01,0.2998182487487793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,4,2,power_law_1.01,0.30273792266845706
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,4,2,power_law_1.01,0.31265279769897464
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,4,2,power_law_1.01,0.3418560028076172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,4,2,power_law_1.01,0.36878719329833987
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,4,2,power_law_1.01,0.42323841094970704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,4,2,power_law_1.01,0.474073600769043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,4,2,power_law_1.01,0.581187858581543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,4,2,power_law_1.01,0.6909913635253906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,4,2,balanced,0.17175296783447264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,4,2,power_law_1.01,0.9098662567138671
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,4,2,power_law_1.01,1.1373286437988281
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,4,2,power_law_1.01,0.041149439811706545
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,4,2,power_law_1.01,0.054239997863769526
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,4,2,power_law_1.01,1.5748838806152343
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,4,2,power_law_1.01,0.06457087993621827
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,4,2,power_law_1.01,0.07900288105010986
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,4,2,power_law_1.01,0.0960870361328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,4,2,power_law_1.01,0.11437055587768555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,4,2,power_law_1.01,0.14378111839294433
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,4,2,power_law_1.01,2.0495616149902345
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,4,2,power_law_1.01,0.1458086395263672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,4,2,power_law_1.01,0.14821120262145998
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,4,2,power_law_1.01,0.1533017635345459
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,4,2,power_law_1.01,0.1569587230682373
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,4,2,power_law_1.01,0.1596019172668457
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,4,2,power_law_1.01,0.1755366325378418
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,4,2,power_law_1.01,0.17797119140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,4,2,power_law_1.01,0.188144645690918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,4,2,power_law_1.01,0.19010303497314454
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,4,2,power_law_1.01,0.16639360427856445
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,4,2,balanced,0.3764825439453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,4,2,balanced,0.42754688262939455
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,4,2,balanced,0.6089471817016602
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,4,2,balanced,0.8023104095458985
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,4,2,power_law_1.01,0.2162995147705078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,4,2,power_law_1.01,0.24552064895629883
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,4,2,power_law_1.01,0.30438272476196293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,4,2,power_law_1.01,0.33778430938720705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,4,2,power_law_1.2,0.06462080001831054
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,4,2,power_law_1.2,0.09682304382324218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,4,2,power_law_1.2,0.1377241611480713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,4,2,power_law_1.2,0.17626623153686521
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,4,2,power_law_1.2,0.2562649536132812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,4,2,power_law_1.2,0.32986495971679686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,4,2,power_law_1.2,0.0825932788848877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,4,2,power_law_1.2,0.3373222351074219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,4,2,power_law_1.2,0.3587827301025391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,4,2,power_law_1.2,0.354284782409668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,4,2,power_law_1.2,0.38517887115478516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,4,2,power_law_1.2,0.435148811340332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,4,2,power_law_1.2,0.4327961730957031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,4,2,power_law_1.2,0.4844787216186523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,4,2,power_law_1.2,0.5045196914672851
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,4,2,power_law_1.2,0.5243532943725586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,4,2,power_law_1.2,0.5483609771728515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,4,2,power_law_1.2,0.583681297302246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,4,2,power_law_1.2,0.6247308731079102
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,4,2,power_law_1.01,0.4246144104003906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,4,2,power_law_1.2,0.7409510040283204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,4,2,power_law_1.2,0.8636748504638672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,4,2,power_law_1.01,0.5497203063964844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,4,2,power_law_1.01,0.7538009643554687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,4,2,power_law_1.2,1.0867520141601563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,4,2,power_law_1.2,1.3588018798828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,4,2,power_law_1.2,0.03904128074645996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,4,2,power_law_1.01,1.0126937866210937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,4,2,power_law_1.2,0.06002048015594482
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,4,2,power_law_1.2,1.8551423645019534
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,4,2,power_law_1.2,0.07435647964477539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,4,2,power_law_1.2,0.107957763671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,4,2,power_law_1.2,2.213376007080078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,4,2,power_law_1.2,0.16788095474243164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,4,2,power_law_1.2,0.22312320709228514
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,4,2,power_law_1.2,0.2261734390258789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,4,2,power_law_1.2,0.23452415466308593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,4,2,power_law_1.2,0.24022911071777342
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,4,2,power_law_1.2,0.24415103912353517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,4,2,power_law_1.2,0.2710848045349121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,4,2,power_law_1.2,0.29154176712036134
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,4,2,power_law_1.2,3.278776245117187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,4,2,power_law_1.2,0.30223743438720707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,4,2,power_law_1.2,0.30474367141723635
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,4,2,power_law_1.2,0.3085312080383301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,4,2,power_law_1.2,0.3235315322875977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,4,2,power_law_1.2,0.3460019302368164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,4,2,power_law_1.2,0.3804390335083008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,4,2,power_law_1.2,0.43394817352294923
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,4,2,power_law_1.2,0.4885823822021484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,4,2,power_law_1.2,4.384968872070313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,4,2,power_law_1.2,0.5887603378295898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,4,2,power_law_1.2,0.7094182586669922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,4,2,power_law_1.2,0.04199679851531982
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,4,2,power_law_1.2,0.9464653015136719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,4,2,power_law_1.2,0.053489918708801265
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,4,2,power_law_1.2,0.06189184188842774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,4,2,power_law_1.2,1.154535675048828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,4,2,power_law_1.2,0.07290112018585206
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,4,2,power_law_1.2,0.09199359893798828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,4,2,power_law_1.2,0.10510208129882812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,4,2,power_law_1.2,1.6384115600585936
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,4,2,power_law_1.2,0.13950592041015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,4,2,power_law_1.2,0.1438169574737549
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,4,2,power_law_1.2,0.1479308795928955
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,4,2,power_law_1.2,0.15146752357482912
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,4,2,power_law_1.2,0.16116863250732422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,4,2,power_law_1.2,2.0882330322265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,4,2,power_law_1.2,0.16943231582641602
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,4,2,power_law_1.2,0.17512319564819337
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,4,2,power_law_1.2,0.1872870445251465
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,4,2,power_law_1.2,0.1912499237060547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,4,2,power_law_1.2,0.28478336334228516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,4,2,power_law_1.2,0.15831168174743654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,4,2,power_law_1.2,0.14609024047851563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,4,2,power_law_1.01,1.4569754028320312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,4,2,power_law_1.01,2.027438049316406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,4,2,power_law_1.2,0.20499839782714843
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,4,2,power_law_1.2,0.24397823333740237
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,4,2,power_law_1.2,0.35821311950683593
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,4,2,power_law_1.2,0.4404134368896484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,8,1,balanced,0.04635519981384277
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,8,1,balanced,0.060698242187499994
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,8,1,balanced,0.07866047859191895
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,4,2,power_law_1.2,0.5520486450195312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,8,1,balanced,0.12643712043762206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,8,1,balanced,0.21405887603759766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,8,1,balanced,0.42006782531738285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,8,1,balanced,0.4195660781860352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,8,1,balanced,0.4207302474975586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,8,1,balanced,0.4239168167114258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,8,1,balanced,0.3923814392089844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,4,2,power_law_1.2,0.8064256286621093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,8,1,balanced,0.3973049545288086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,8,1,balanced,0.39982463836669924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,8,1,balanced,0.4071596908569336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,8,1,balanced,0.4088505554199219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,8,1,balanced,0.4169446563720703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,8,1,balanced,0.49082496643066403
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,8,1,balanced,0.4927590560913086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,8,1,balanced,0.5474515151977539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,8,1,balanced,0.5862310409545899
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,8,1,balanced,0.6738899230957032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,8,1,balanced,0.730789794921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,8,1,balanced,0.7947833251953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,8,1,balanced,0.39131137847900394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,8,1,balanced,1.1889254760742187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,8,1,balanced,0.045537919998168946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,8,1,balanced,1.3068109130859376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,8,1,balanced,0.05597887992858887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,8,1,balanced,0.07709248065948486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,8,1,balanced,0.11627391815185546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,8,1,balanced,0.19795520782470705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,8,1,balanced,1.8681382751464846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,8,1,balanced,0.3661593627929688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,8,1,balanced,0.3646937561035156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,8,1,balanced,0.3672313690185547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,8,1,balanced,0.3684473419189453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,8,1,balanced,0.3709408187866211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,8,1,balanced,0.37469886779785155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,8,1,balanced,0.3775040054321289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,8,1,balanced,0.3818975830078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,8,1,balanced,0.38786815643310546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,8,1,balanced,0.4313241577148437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,8,1,balanced,0.4336467361450195
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,4,2,power_law_1.2,1.0497203063964844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,8,1,balanced,0.4420345687866211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,8,1,balanced,0.45604991912841797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,8,1,balanced,0.4705132675170899
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,8,1,balanced,0.5021599960327149
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,8,1,balanced,0.5349856185913086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,8,1,balanced,0.5974387359619141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,8,1,balanced,0.6644525146484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,8,1,balanced,1.0061004638671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,4,2,power_law_1.2,1.5713113403320311
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,8,1,balanced,1.138946533203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,8,1,balanced,1.6570790100097657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,8,1,balanced,2.129843902587891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,8,1,balanced,2.351631317138672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,4,2,power_law_1.2,2.291577606201172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,4,2,power_law_1.2,2.9421810913085937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,8,1,balanced,0.06232704162597656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,8,1,balanced,0.07536831855773926
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,8,1,balanced,0.06572351932525636
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,8,1,balanced,0.2235673522949219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,8,1,balanced,0.09766079902648925
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,8,1,balanced,0.224389762878418
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,8,1,balanced,0.22499967575073243
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,8,1,balanced,0.14125184059143064
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,8,1,balanced,0.22816192626953127
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,8,1,balanced,0.2301158332824707
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,8,1,balanced,0.2327078437805176
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,8,1,balanced,0.22689088821411132
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,8,1,balanced,0.23562559127807617
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,8,1,balanced,0.2407084846496582
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,8,1,balanced,0.2507251167297363
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,8,1,balanced,0.26072704315185546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,8,1,balanced,0.2824448013305664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,8,1,balanced,0.3024972724914551
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,8,1,balanced,0.3471507263183594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,8,1,balanced,0.3915846252441406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,8,1,power_law_1.01,0.04053760051727295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,8,1,power_law_1.01,0.059411840438842775
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,8,1,power_law_1.01,0.07776639938354493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,8,1,power_law_1.01,0.12454463958740233
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,8,1,balanced,0.22099840164184573
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,8,1,power_law_1.01,0.16766016006469725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,8,1,power_law_1.01,0.25755712509155276
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,8,1,power_law_1.01,0.30254783630371096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,8,1,power_law_1.01,0.3183558464050293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,8,1,power_law_1.01,0.32454399108886717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,8,1,power_law_1.01,0.33586753845214845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,8,1,power_law_1.01,0.3483481597900391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,8,1,power_law_1.01,0.3578623962402344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,8,1,power_law_1.01,0.363746566772461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,8,1,power_law_1.01,0.37759487152099613
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,8,1,power_law_1.01,0.4306406402587891
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,8,1,balanced,0.22208639144897463
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,8,1,power_law_1.01,0.44461952209472655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,8,1,power_law_1.01,0.5072467041015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,8,1,balanced,0.22283071517944336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,8,1,power_law_1.01,0.5376351928710937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,8,1,power_law_1.01,0.6359052658081055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,8,1,power_law_1.01,0.7818681335449218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,8,1,power_law_1.01,0.8698028564453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,8,1,power_law_1.01,0.9616153717041016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,8,1,power_law_1.01,1.2349651336669922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,8,1,power_law_1.01,1.5205331420898438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,8,1,power_law_1.01,0.04433728218078613
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,8,1,power_law_1.01,0.46601089477539065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,8,1,power_law_1.01,0.05079616069793701
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,8,1,power_law_1.01,0.07864831924438477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,8,1,power_law_1.01,2.0714405822753905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,8,1,power_law_1.01,0.11441856384277345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,8,1,power_law_1.01,0.1724684715270996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,8,1,power_law_1.01,0.23048896789550782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,8,1,power_law_1.01,0.2899423980712891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,8,1,power_law_1.01,0.30180032730102535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,8,1,power_law_1.01,0.3072543907165527
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,8,1,power_law_1.01,0.3097913551330566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,8,1,power_law_1.01,2.6188531494140626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,8,1,power_law_1.01,0.3213100814819336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,8,1,power_law_1.01,0.33047550201416015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,8,1,power_law_1.01,0.3353420639038086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,8,1,power_law_1.01,0.3424595260620117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,8,1,power_law_1.01,0.3933184051513672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,8,1,power_law_1.01,0.3923219299316406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,8,1,power_law_1.01,0.41797824859619137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,8,1,power_law_1.01,0.4364531326293945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,8,1,power_law_1.01,0.4636627197265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,8,1,power_law_1.01,0.5144070434570313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,8,1,power_law_1.01,0.5782086563110351
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,8,1,power_law_1.01,0.6872652435302735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,8,1,power_law_1.01,0.8047596740722657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,8,1,power_law_1.01,1.0433952331542968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,8,1,power_law_1.01,1.3003187561035157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,8,1,power_law_1.01,1.7777017211914063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,8,1,power_law_1.01,0.06188928127288819
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,8,1,power_law_1.01,0.06665535926818847
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,8,1,power_law_1.01,2.283646697998047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,8,1,power_law_1.01,0.08007488250732422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,8,1,power_law_1.01,0.09868288040161133
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,8,1,power_law_1.01,0.1209273624420166
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,8,1,power_law_1.01,0.14206720352172852
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,8,1,power_law_1.01,0.178604793548584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,8,1,power_law_1.01,0.1861849594116211
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,8,1,power_law_1.01,0.1907321548461914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,8,1,power_law_1.01,0.19377023696899415
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,8,1,balanced,0.6938214111328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,8,1,balanced,0.610579833984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,8,1,power_law_1.01,0.2018060874938965
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,8,1,power_law_1.01,0.21352575302124022
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,8,1,power_law_1.01,0.22049407958984374
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,8,1,balanced,1.0126566314697265
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,8,1,power_law_1.01,0.20821439743041995
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,8,1,power_law_1.01,0.23196928024291993
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,8,1,power_law_1.01,0.2582758331298828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,8,1,balanced,1.3303794860839844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,8,1,power_law_1.01,0.2379136085510254
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,8,1,power_law_1.01,0.3316620635986328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,8,1,power_law_1.01,0.2943827247619629
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,8,1,power_law_1.01,0.6719071960449219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,8,1,power_law_1.01,0.4135436630249023
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,8,1,power_law_1.01,0.48896896362304687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,8,1,power_law_1.2,0.06047999858856201
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,8,1,power_law_1.2,0.07952703952789306
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,8,1,power_law_1.01,0.8385849761962891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,8,1,power_law_1.2,0.11735360145568847
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,8,1,power_law_1.2,0.1576211166381836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,8,1,power_law_1.01,1.2255372619628906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,8,1,power_law_1.2,0.29782976150512697
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,8,1,power_law_1.2,0.31195327758789065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,8,1,power_law_1.2,0.3350252914428711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,8,1,power_law_1.2,0.3194579124450684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,8,1,power_law_1.2,0.3442911911010742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,8,1,power_law_1.2,0.3548697662353516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,8,1,power_law_1.2,0.3657548904418945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,8,1,power_law_1.2,0.3865388870239258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,8,1,power_law_1.2,0.03991039991378784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,8,1,power_law_1.2,0.43464321136474615
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,8,1,power_law_1.2,0.4582457733154297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,8,1,power_law_1.01,1.5571359252929686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,8,1,power_law_1.2,0.4738431930541992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,8,1,power_law_1.2,0.5406265640258789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,8,1,power_law_1.2,0.5481062316894532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,8,1,power_law_1.2,0.6519449615478516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,8,1,power_law_1.2,0.8039373016357422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,8,1,power_law_1.2,0.2399776077270508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,8,1,power_law_1.2,0.8925113677978516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,8,1,power_law_1.2,0.04878208160400391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,8,1,power_law_1.2,1.0765087890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,8,1,power_law_1.2,0.05219264030456543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,8,1,power_law_1.2,0.07928127765655517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,8,1,power_law_1.2,1.5446733093261718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,8,1,power_law_1.2,0.10889727592468261
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,8,1,power_law_1.2,1.2612844848632814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,8,1,power_law_1.2,2.100552978515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,8,1,power_law_1.2,0.22255231857299806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,8,1,power_law_1.2,0.28753984451293946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,8,1,power_law_1.2,0.29672512054443356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,8,1,power_law_1.2,2.6501708984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,8,1,power_law_1.2,0.16722688674926758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,8,1,power_law_1.2,0.30334848403930664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,8,1,power_law_1.2,0.31111423492431645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,8,1,power_law_1.2,0.3193247985839844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,8,1,power_law_1.2,0.3285561752319336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,8,1,power_law_1.2,0.335731201171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,8,1,power_law_1.2,0.35158592224121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,8,1,power_law_1.2,0.4026790237426757
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,8,1,power_law_1.2,0.4089177703857422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,8,1,power_law_1.2,0.42139457702636723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,8,1,power_law_1.2,0.44975872039794923
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,8,1,power_law_1.2,0.47501697540283205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,8,1,power_law_1.2,0.544951057434082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,8,1,power_law_1.2,0.5944876861572266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,8,1,power_law_1.2,0.7184941101074218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,8,1,power_law_1.2,0.8450387573242187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,8,1,power_law_1.2,1.0600908660888673
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,8,1,power_law_1.2,0.0619488000869751
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,8,1,power_law_1.2,0.06681215763092041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,8,1,power_law_1.2,1.8072869873046877
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,8,1,power_law_1.2,0.07811456203460694
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,8,1,power_law_1.2,1.3171206665039061
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,8,1,power_law_1.2,0.09535743713378905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,8,1,power_law_1.2,2.2926220703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,8,1,power_law_1.2,0.11499520301818848
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,8,1,power_law_1.2,0.13720128059387207
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,8,1,power_law_1.2,0.17586496353149414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,8,1,power_law_1.2,0.18493375778198243
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,8,1,power_law_1.2,0.18974847793579103
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,8,1,power_law_1.2,0.19474752426147462
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,8,1,power_law_1.2,0.20266048431396486
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,8,1,power_law_1.2,0.20960960388183594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,8,1,power_law_1.2,0.2161452865600586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,8,1,power_law_1.2,0.22926399230957034
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,8,1,power_law_1.2,0.24400447845458983
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,8,1,power_law_1.2,0.2549587249755859
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,8,1,power_law_1.2,0.2825791931152344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,8,1,power_law_1.01,2.298014678955078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,8,1,power_law_1.01,3.0304113769531247
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,8,1,power_law_1.2,0.33622207641601565
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,8,1,power_law_1.2,0.39090240478515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,8,1,power_law_1.2,0.5163366317749023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,8,balanced,0.021016321182250976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,8,balanced,0.021091840267181396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,8,balanced,0.023086080551147463
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,8,1,power_law_1.2,0.6425209808349609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,8,balanced,0.03133055925369262
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,8,1,power_law_1.2,0.9185830688476562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,8,balanced,0.046456317901611324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,8,balanced,0.06883967876434326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,8,balanced,0.07374207973480225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,8,balanced,0.07562496185302733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,8,balanced,0.07158016204833985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,8,balanced,0.08054271697998047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,8,balanced,0.07089151859283446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,8,balanced,0.07329023838043212
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,8,balanced,0.07511424064636231
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,8,balanced,0.07827328205108643
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,8,1,power_law_1.2,1.2234822082519532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,8,balanced,0.08185215950012206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,8,balanced,0.08546815872192383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,8,balanced,0.09076095581054687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,8,balanced,0.11815423965454101
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,8,balanced,0.10541184425354004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,8,balanced,0.15744128227233886
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,8,balanced,0.18905088424682617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,8,balanced,0.27012096405029296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,8,balanced,0.33683582305908205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,8,balanced,0.4791500854492187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,8,balanced,0.6163967895507813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,8,balanced,0.024381439685821533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,8,balanced,0.025547521114349363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,8,balanced,0.026023681163787844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,8,balanced,0.026414079666137697
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,8,balanced,0.9044480133056642
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,8,balanced,0.03469183921813965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,8,balanced,0.03667968034744263
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,8,balanced,0.038717439174652105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,8,balanced,0.041834239959716794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,8,balanced,1.190439682006836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,8,balanced,0.045585918426513675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,8,balanced,0.03663232088088989
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,8,balanced,0.04069375991821289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,8,balanced,0.04360576152801514
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,8,balanced,0.04690432071685791
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,8,balanced,0.048244481086730955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,8,balanced,0.05100160121917725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,8,balanced,0.05623424053192139
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,8,balanced,0.06655871868133545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,8,balanced,0.07636096000671386
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,8,balanced,0.09908479690551758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,8,balanced,0.11552767753601074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,8,balanced,0.17271936416625977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,8,balanced,0.21287296295166014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,8,balanced,0.29657472610473634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,8,balanced,0.39300865173339844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,8,balanced,0.5673100662231445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,8,balanced,0.03216383934020996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,8,balanced,0.7429888153076172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,8,1,power_law_1.2,1.85453369140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,8,1,power_law_1.2,2.5921356201171877
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,1,8,balanced,0.018983680009841918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,1,8,balanced,0.017420799732208253
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,1,8,balanced,0.023498239517211913
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,1,8,balanced,0.023555839061737062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,1,8,balanced,0.017775360345840454
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,1,8,balanced,0.02388607978820801
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,1,8,balanced,0.02347520112991333
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,1,8,balanced,0.0240012788772583
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,1,8,balanced,0.023962879180908205
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,1,8,balanced,0.023912959098815918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,1,8,balanced,0.024216320514678955
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,1,8,balanced,0.024046080112457274
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,1,8,balanced,0.024721920490264893
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,1,8,balanced,0.02020224094390869
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,1,8,balanced,0.02502784013748169
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,1,8,balanced,0.027229440212249757
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,1,8,balanced,0.023924479484558107
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,1,8,balanced,0.026851840019226074
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,1,8,balanced,0.023907840251922607
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,1,8,balanced,0.023491840362548828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,1,8,balanced,0.02348927974700928
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,1,8,balanced,0.06238080024719238
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,1,8,balanced,0.03874176025390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,1,8,balanced,0.04186880111694336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,1,8,balanced,0.05288447856903076
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,8,power_law_1.01,0.03683327913284302
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,8,power_law_1.01,0.03919615983963013
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,8,power_law_1.01,0.04128255844116211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,8,power_law_1.01,0.041196799278259276
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,8,power_law_1.01,0.06500480175018311
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,8,power_law_1.01,0.06855552196502686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,8,power_law_1.01,0.07083648204803467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,8,power_law_1.01,0.07241983890533447
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,8,power_law_1.01,0.07477248191833497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,8,power_law_1.01,0.02636415958404541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,8,power_law_1.01,0.08014847755432128
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,8,power_law_1.01,0.0728704023361206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,8,power_law_1.01,0.07860095977783203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,8,power_law_1.01,0.07541888236999512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,8,power_law_1.01,0.08083583831787108
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,8,power_law_1.01,0.08792192459106445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,8,power_law_1.01,0.1025113582611084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,8,power_law_1.01,0.1283942413330078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,8,power_law_1.01,0.16624256134033205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,8,power_law_1.01,0.2012723159790039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,8,power_law_1.01,0.28312063217163086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,8,power_law_1.01,0.07225599765777588
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,8,power_law_1.01,0.35811710357666016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,8,power_law_1.01,0.5399871826171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,8,power_law_1.01,0.025937919616699216
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,8,power_law_1.01,0.6496486663818359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,8,power_law_1.01,1.0000179290771485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,8,power_law_1.01,0.030183680057525634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,8,power_law_1.01,0.03005055904388428
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,8,power_law_1.01,0.030266880989074707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,8,power_law_1.01,1.3753318786621094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,8,power_law_1.01,0.03578111886978149
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,8,power_law_1.01,0.033780479431152345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,8,power_law_1.01,0.03837696075439453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,8,power_law_1.01,0.03837824106216431
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,8,power_law_1.01,0.04227712154388428
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,1,8,balanced,0.08286208152770996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,8,power_law_1.01,0.0505459213256836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,8,power_law_1.01,0.039946238994598385
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,8,power_law_1.01,0.04790016174316406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,8,power_law_1.01,0.0428223991394043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,8,power_law_1.01,0.05048960208892822
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,8,power_law_1.01,0.0564031982421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,8,power_law_1.01,0.06608767986297608
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,8,power_law_1.01,0.08045439720153809
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,8,power_law_1.01,0.10191360473632813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,8,power_law_1.01,0.1220582389831543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,8,power_law_1.01,0.16864511489868164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,8,power_law_1.01,0.028239359855651857
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,8,power_law_1.01,0.206243839263916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,8,power_law_1.01,0.30567935943603514
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,8,power_law_1.01,0.0420684814453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,1,8,power_law_1.01,0.021424639225006103
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,1,8,power_law_1.01,0.019409919977188113
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,1,8,balanced,0.10299136161804198
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,1,8,power_law_1.01,0.022658560276031494
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,8,power_law_1.01,0.3721535873413086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,1,8,power_law_1.01,0.0202239990234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,1,8,power_law_1.01,0.023155200481414794
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,1,8,power_law_1.01,0.023089919090270996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,8,power_law_1.01,0.561157112121582
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,1,8,power_law_1.01,0.02364543914794922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,1,8,power_law_1.01,0.02350719928741455
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,1,8,power_law_1.01,0.023500800132751465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,8,power_law_1.01,0.7452019500732422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,1,8,power_law_1.01,0.0234879994392395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,1,8,power_law_1.01,0.02388351917266846
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,1,8,power_law_1.01,0.023508479595184324
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,1,8,power_law_1.01,0.024477438926696776
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,1,8,power_law_1.01,0.024463360309600828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,1,8,power_law_1.01,0.026492159366607666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,1,8,power_law_1.01,0.03472512006759644
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,1,8,power_law_1.01,0.02148736000061035
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,1,8,power_law_1.01,0.030780160427093507
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,1,8,power_law_1.01,0.05113344192504883
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,1,8,power_law_1.01,0.06156288146972656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,1,8,power_law_1.01,0.02790015935897827
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,1,8,power_law_1.01,0.04041855812072754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,8,power_law_1.2,0.026785280704498295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,8,power_law_1.2,0.035846400260925296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,8,power_law_1.2,0.038558719158172605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,8,power_law_1.2,0.04049536228179932
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,8,power_law_1.2,0.03991679906845093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,8,power_law_1.2,0.06464767932891846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,8,power_law_1.2,0.06834943771362305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,8,power_law_1.2,0.07257472038269043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,8,power_law_1.2,0.07376128196716308
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,1,8,power_law_1.01,0.06952447891235351
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,8,power_law_1.2,0.07768447875976563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,8,power_law_1.2,0.0789247989654541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,8,power_law_1.2,0.07655168056488038
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,8,power_law_1.2,0.07263360023498536
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,8,power_law_1.2,0.07556096076965332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,8,power_law_1.2,0.08017791748046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,8,power_law_1.2,0.08368000030517578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,8,power_law_1.2,0.0919705581665039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,8,power_law_1.2,0.11062015533447264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,8,power_law_1.2,0.13667072296142577
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,8,power_law_1.2,0.17377023696899413
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,8,power_law_1.2,0.22612352371215821
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,8,power_law_1.2,0.30812416076660154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,8,power_law_1.2,0.4042227172851563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,8,power_law_1.2,0.5883622360229492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,8,power_law_1.2,0.7915302276611328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,8,power_law_1.2,1.5140415954589845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,8,power_law_1.2,1.1270540618896485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,8,power_law_1.2,0.02551039934158325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,8,power_law_1.2,0.027656960487365722
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,8,power_law_1.2,0.030108160972595215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,8,power_law_1.2,0.030937600135803222
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,8,power_law_1.2,0.0296230411529541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,8,power_law_1.2,0.03501568078994751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,8,power_law_1.2,0.03501568078994751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,8,power_law_1.2,0.03878400087356567
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,8,power_law_1.2,0.03789567947387695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,8,power_law_1.2,0.04034048080444336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,8,power_law_1.2,0.04335616111755371
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,1,8,power_law_1.01,0.10003071784973146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,8,power_law_1.2,0.039883520603179935
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,8,power_law_1.2,0.04283904075622559
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,8,power_law_1.2,0.04624512195587158
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,8,power_law_1.2,0.049779200553894044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,8,power_law_1.2,0.051904001235961915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,8,power_law_1.2,0.0564518404006958
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,1,8,power_law_1.01,0.13569536209106445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,8,power_law_1.2,0.0710963201522827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,8,power_law_1.2,0.08376192092895508
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,1,8,power_law_1.01,0.1835852813720703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,8,power_law_1.2,0.10599424362182616
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,8,power_law_1.2,0.13226495742797853
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,1,8,power_law_1.01,0.24548736572265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,8,power_law_1.2,0.22349184036254882
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,8,power_law_1.2,0.18609024047851563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,1,8,power_law_1.2,0.019456000328063966
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,1,8,power_law_1.2,0.019368959665298464
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,1,8,power_law_1.2,0.020980479717254637
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,1,8,power_law_1.2,0.021413118839263917
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,1,8,power_law_1.2,0.0207039999961853
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,1,8,power_law_1.2,0.023033599853515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,1,8,power_law_1.2,0.023290879726409912
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,1,8,power_law_1.2,0.023463680744171142
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,1,8,power_law_1.2,0.02348927974700928
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,1,8,power_law_1.2,0.02347775936126709
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,1,8,power_law_1.2,0.02348288059234619
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,1,8,power_law_1.2,0.023868160247802733
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,1,8,power_law_1.2,0.023880960941314696
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,1,8,power_law_1.2,0.026260480880737302
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,1,8,power_law_1.2,0.02723072052001953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,8,power_law_1.2,0.42297214508056646
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,8,power_law_1.2,0.3328307342529297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,8,power_law_1.2,0.6371315383911134
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,8,power_law_1.2,0.8388416290283203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,1,8,power_law_1.2,0.025300478935241698
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,1,8,power_law_1.2,0.02930176019668579
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,1,8,power_law_1.2,0.03333888053894043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,1,8,power_law_1.2,0.038693120479583745
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,1,8,power_law_1.2,0.04710271835327148
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,1,8,power_law_1.2,0.06279808044433594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,1,8,power_law_1.2,0.07245312213897706
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,4,balanced,0.0202675199508667
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,4,balanced,0.021493759155273438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,4,balanced,0.02313472032546997
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,1,8,power_law_1.2,0.14369407653808594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,4,balanced,0.04830463886260986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,4,balanced,0.05094783782958985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,4,balanced,0.06928768157958984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,4,balanced,0.07199488162994386
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,4,balanced,0.07311488151550292
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,4,balanced,0.07802495956420899
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,1,8,power_law_1.2,0.10595456123352051
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,4,balanced,0.06869760036468506
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,4,balanced,0.0714841604232788
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,4,balanced,0.0732147216796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,4,balanced,0.07573631763458252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,4,balanced,0.07802112102508545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,4,balanced,0.08340479850769042
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,4,balanced,0.09438847541809083
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,4,balanced,0.1050534439086914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,4,balanced,0.13765376091003417
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,4,balanced,0.16075519561767576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,4,balanced,0.0667302417755127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,4,balanced,0.23190143585205075
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,1,8,power_law_1.2,0.1632806396484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,4,balanced,0.2925503921508789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,4,balanced,0.4201254272460938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,4,balanced,0.5413068771362305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,4,balanced,0.024336640834808347
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,4,balanced,0.025519359111785888
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,4,balanced,0.025922560691833497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,4,balanced,0.7945843505859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,4,balanced,0.028609280586242673
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,4,balanced,0.0353600001335144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,4,balanced,0.04089856147766113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,4,balanced,0.04166016101837158
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,4,balanced,1.0437017822265626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,4,balanced,0.04332159996032715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,4,balanced,0.04517759799957276
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,4,balanced,0.04028800010681152
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,4,balanced,0.04288383960723877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,4,balanced,0.04575232028961181
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,4,balanced,0.052080640792846675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,4,balanced,0.047467517852783206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,4,balanced,0.040085759162902836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,4,balanced,0.057121281623840336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,4,balanced,0.06667520046234131
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,4,balanced,0.07297791957855224
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,4,balanced,0.04345600128173828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,4,balanced,0.10662783622741698
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,4,balanced,0.16065664291381837
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,4,balanced,0.1938662338256836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,4,balanced,0.2732057571411133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,4,balanced,0.35576576232910156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,4,balanced,0.09240575790405273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,4,balanced,0.5144384002685547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,4,balanced,0.6790937805175782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,4,balanced,0.03168895959854126
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,2,4,balanced,0.021388800144195558
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,2,4,balanced,0.0281279993057251
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,2,4,balanced,0.023654398918151857
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,2,4,balanced,0.028302080631256103
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,2,4,balanced,0.02811903953552246
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,2,4,balanced,0.02841088056564331
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,2,4,balanced,0.0276364803314209
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,2,4,balanced,0.022305281162261964
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,2,4,balanced,0.028947200775146488
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,2,4,balanced,0.028913919925689695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,2,4,balanced,0.028190720081329345
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,2,4,balanced,0.02905087947845459
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,2,4,balanced,0.028458240032196047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,2,4,balanced,0.03020031929016113
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,2,4,balanced,0.029920001029968262
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,2,4,balanced,0.032460799217224126
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,2,4,balanced,0.03208575963973999
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,2,4,balanced,0.03517312049865723
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,2,4,balanced,0.037056000232696534
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,1,8,power_law_1.2,0.24749183654785156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,2,4,balanced,0.019831039905548097
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,2,4,balanced,0.027253758907318116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,4,power_law_1.01,0.02187648057937622
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,4,power_law_1.01,0.02986112117767334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,4,power_law_1.01,0.03248127937316895
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,2,4,balanced,0.0710694408416748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,4,power_law_1.01,0.03777920007705689
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,2,4,balanced,0.056026878356933596
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,2,4,balanced,0.08694399833679199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,4,power_law_1.01,0.038696959018707275
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,4,power_law_1.01,0.051870718002319335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,4,power_law_1.01,0.07409152030944824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,4,power_law_1.01,0.06902400016784668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,4,power_law_1.01,0.0743116807937622
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,4,power_law_1.01,0.07801087856292724
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,4,power_law_1.01,0.06794879913330079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,4,power_law_1.01,0.06965375900268554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,4,power_law_1.01,0.07662847995758057
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,2,4,balanced,0.11774975776672363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,4,power_law_1.01,0.07923967838287353
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,4,power_law_1.01,0.08555264472961426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,4,power_law_1.01,0.09859456062316894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,4,power_law_1.01,0.11438719749450683
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,2,4,balanced,0.05257343769073486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,4,power_law_1.01,0.19024896621704102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,4,power_law_1.01,0.06738687992095947
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,4,power_law_1.01,0.25657344818115235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,4,power_law_1.01,0.305166072845459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,4,power_law_1.01,0.07257472038269043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,4,power_law_1.01,0.46238464355468756
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,4,power_law_1.01,0.586901741027832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,4,power_law_1.01,0.15269248008728026
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,4,power_law_1.01,0.845617904663086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,4,power_law_1.01,0.026768639087677005
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,4,power_law_1.01,0.027560958862304686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,4,power_law_1.01,0.031726078987121584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,4,power_law_1.01,0.031674880981445316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,4,power_law_1.01,0.03725568056106567
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,4,power_law_1.01,0.03953408002853394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,4,power_law_1.01,1.1339622497558595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,4,power_law_1.01,0.04493696212768555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,4,power_law_1.01,0.04073984146118164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,4,power_law_1.01,0.0448422384262085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,4,power_law_1.01,0.05231488227844239
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,4,power_law_1.01,0.025131518840789797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,4,power_law_1.01,0.041753602027893064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,4,power_law_1.01,0.0424179220199585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,4,power_law_1.01,0.04401919841766357
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,1,8,power_law_1.2,0.37071231842041014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,4,power_law_1.01,0.04882688045501709
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,4,power_law_1.01,0.05043712139129639
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,4,power_law_1.01,0.05798399925231934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,4,power_law_1.01,0.06925183773040772
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,4,power_law_1.01,0.07761919975280762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,4,power_law_1.01,0.10217984199523926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,4,power_law_1.01,0.12305536270141601
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,4,power_law_1.01,0.1673779106140137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,4,power_law_1.01,0.20779775619506835
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,2,4,power_law_1.01,0.021350400447845457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,4,power_law_1.01,0.2956723213195801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,4,power_law_1.01,0.38610816955566407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,2,4,power_law_1.01,0.02389631986618042
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,2,4,power_law_1.01,0.02469504117965698
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,2,4,power_law_1.01,0.0240883207321167
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,4,power_law_1.01,0.5494707107543946
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,2,4,power_law_1.01,0.027228159904479982
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,2,4,power_law_1.01,0.02846208095550537
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,2,4,power_law_1.01,0.02879744052886963
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,2,4,power_law_1.01,0.028801279067993162
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,2,4,power_law_1.01,0.028494079113006592
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,4,power_law_1.01,0.7253926086425782
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,2,4,power_law_1.01,0.029797120094299318
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,2,4,power_law_1.01,0.03004928112030029
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,2,4,power_law_1.01,0.030529279708862305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,2,4,power_law_1.01,0.03072128057479858
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,2,4,power_law_1.01,0.027411201000213624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,2,4,balanced,0.1493824005126953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,2,4,power_law_1.01,0.022452480792999267
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,2,4,power_law_1.01,0.03529344081878662
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,2,4,power_law_1.01,0.04010623931884766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,2,4,power_law_1.01,0.044994559288024906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,2,4,power_law_1.01,0.05395455837249756
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,2,4,power_law_1.01,0.03302783966064453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,4,power_law_1.2,0.021840639114379883
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,2,4,power_law_1.01,0.06253183841705322
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,4,power_law_1.2,0.027655680179595948
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,4,power_law_1.2,0.03214207887649536
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,4,power_law_1.2,0.03580415964126587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,4,power_law_1.2,0.038932480812072755
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,4,power_law_1.2,0.04777728080749512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,4,power_law_1.2,0.0674393606185913
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,4,power_law_1.2,0.07172607898712158
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,4,power_law_1.2,0.07247360229492188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,4,power_law_1.2,0.0743616008758545
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,2,4,power_law_1.01,0.0773145580291748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,4,power_law_1.2,0.06855040073394776
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,4,power_law_1.2,0.06833920001983643
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,4,power_law_1.2,0.07359488010406494
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,4,power_law_1.2,0.077642240524292
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,4,power_law_1.2,0.08006272315979004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,4,power_law_1.2,0.08643712043762207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,4,power_law_1.2,0.10021375656127929
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,4,power_law_1.2,0.11560704231262206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,4,power_law_1.2,0.1627187156677246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,4,power_law_1.2,0.26163711547851565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,4,power_law_1.2,0.19371519088745118
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,4,power_law_1.2,0.33464191436767576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,4,power_law_1.2,0.48323200225830076
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,4,power_law_1.2,0.07091839790344238
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,4,power_law_1.2,0.6324940872192383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,4,power_law_1.2,0.02522111892700195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,4,power_law_1.2,0.02617599964141846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,4,power_law_1.2,0.027644159793853762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,4,power_law_1.2,0.03054336071014404
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,4,power_law_1.2,0.02959872007369995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,4,power_law_1.2,0.03655296087265015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,4,power_law_1.2,0.03953408002853394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,4,power_law_1.2,0.040755200386047366
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,4,power_law_1.2,0.04281472206115723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,4,power_law_1.2,1.1846451568603515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,4,power_law_1.2,0.044440321922302246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,4,power_law_1.2,0.040737280845642085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,4,power_law_1.2,0.04093440055847168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,4,power_law_1.2,0.04255743980407715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,4,power_law_1.2,0.04576767921447754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,4,power_law_1.2,0.04848896026611328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,4,power_law_1.2,0.05223552227020264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,4,power_law_1.2,0.05976831912994385
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,4,power_law_1.2,0.0701862382888794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,4,power_law_1.2,0.07844351768493653
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,4,power_law_1.2,0.10419967651367187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,4,power_law_1.2,0.9072895812988282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,4,power_law_1.2,0.12682623863220216
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,4,power_law_1.2,0.17069952011108397
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,4,power_law_1.2,0.2129497528076172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,4,power_law_1.2,0.2999488067626953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,2,4,power_law_1.01,0.10456192016601562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,2,4,power_law_1.2,0.021056001186370847
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,2,4,power_law_1.2,0.021856000423431395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,2,4,power_law_1.2,0.023349759578704835
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,2,4,power_law_1.01,0.14142208099365233
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,2,4,power_law_1.2,0.025464320182800294
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,2,4,power_law_1.2,0.027601919174194335
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,2,4,power_law_1.2,0.025108480453491212
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,2,4,power_law_1.2,0.028454399108886717
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,2,4,power_law_1.2,0.029012479782104493
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,2,4,power_law_1.01,0.1754803276062012
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,2,4,power_law_1.01,0.2578892707824707
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,2,4,power_law_1.2,0.02878976106643677
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,2,4,power_law_1.2,0.02923392057418823
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,2,4,power_law_1.2,0.029791998863220214
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,2,4,power_law_1.2,0.030081279277801513
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,2,4,power_law_1.2,0.03240447998046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,2,4,power_law_1.2,0.03294719934463501
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,2,4,power_law_1.01,0.3511731338500977
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,2,4,power_law_1.2,0.03755520105361938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,2,4,power_law_1.2,0.028796160221099855
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,2,4,power_law_1.2,0.03416320085525513
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,4,power_law_1.2,0.39270015716552736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,4,power_law_1.2,0.5655846405029297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,2,4,power_law_1.2,0.04386943817138672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,2,4,power_law_1.2,0.04662655830383301
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,2,4,power_law_1.2,0.06312960147857666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,2,4,power_law_1.2,0.10412927627563477
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,2,4,power_law_1.2,0.07445888042449951
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,4,2,balanced,0.020234239101409913
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,2,balanced,0.022328319549560546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,2,4,power_law_1.2,0.1405235195159912
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,2,balanced,0.03167743921279907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,2,balanced,0.04774144172668457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,2,balanced,0.04790783882141113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,2,balanced,0.048229122161865236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,2,balanced,0.06822144031524659
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,2,balanced,0.06830336093902588
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,2,balanced,0.06892288208007813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,2,balanced,0.07173120021820069
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,2,balanced,0.07264639854431151
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,4,2,balanced,0.02030719995498657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,2,balanced,0.07388671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,2,balanced,0.07588863849639893
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,2,balanced,0.07806079864501954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,2,balanced,0.08378111839294433
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,2,balanced,0.09495424270629883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,2,balanced,0.1058675193786621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,2,balanced,0.13258111953735352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,2,balanced,0.15369088172912598
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,2,balanced,0.22379392623901367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,2,balanced,0.27616128921508787
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,2,balanced,0.3936511993408203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,2,balanced,0.5076863861083984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,2,balanced,0.7474598693847656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,2,balanced,0.9795916748046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,2,balanced,0.02390399932861328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,2,balanced,0.0239846396446228
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,2,balanced,0.03250175952911377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,2,balanced,0.04332799911499023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,2,balanced,0.04592895984649658
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,2,4,power_law_1.2,0.20804224014282227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,2,balanced,0.04239999771118164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,2,balanced,0.04813439846038818
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,2,balanced,0.04276351928710938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,2,balanced,0.04888192176818847
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,2,balanced,0.04769408226013184
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,2,balanced,0.04985599994659424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,2,balanced,0.05222655773162842
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,2,balanced,0.05648767948150635
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,2,balanced,0.053217282295227056
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,2,balanced,0.057715201377868654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,2,balanced,0.059697918891906745
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,2,balanced,0.0649292802810669
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,2,balanced,0.07165567874908448
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,2,balanced,0.07858560085296631
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,2,balanced,0.09247103691101075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,2,balanced,0.10762880325317384
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,2,balanced,0.16350976943969725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,2,balanced,0.194388484954834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,2,balanced,0.27949823379516603
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,2,balanced,0.36378494262695316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,2,balanced,0.5321177673339844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,2,balanced,0.6933554840087891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,2,balanced,0.026408960819244386
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,4,power_law_1.2,0.7520089721679687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,2,4,power_law_1.2,0.2769753646850586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,4,2,balanced,0.028257279396057128
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,4,2,balanced,0.027587840557098387
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,4,2,balanced,0.039237120151519776
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,4,2,balanced,0.03893376111984253
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,4,2,balanced,0.038877439498901364
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,4,2,balanced,0.0288755202293396
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,4,2,balanced,0.03936000108718872
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,4,2,balanced,0.03960063934326172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,4,2,balanced,0.0395251202583313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,4,2,balanced,0.040249600410461425
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,4,2,balanced,0.04034048080444336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,4,2,balanced,0.041626877784729004
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,4,2,balanced,0.04231423854827881
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,4,2,balanced,0.042626562118530276
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,4,2,balanced,0.04531455993652343
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,4,2,balanced,0.0477235221862793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,4,2,balanced,0.05175680160522461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,4,2,balanced,0.05555200099945068
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,4,2,balanced,0.03113343954086304
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,4,2,balanced,0.03865087985992431
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,4,2,balanced,0.039197440147399905
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,2,4,power_law_1.2,0.39692928314208986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,4,2,power_law_1.01,0.021111040115356444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,4,2,power_law_1.01,0.023155200481414794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,2,power_law_1.01,0.03069823980331421
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,2,power_law_1.01,0.035084800720214845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,2,power_law_1.01,0.0393612790107727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,2,power_law_1.01,0.04988671779632568
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,2,power_law_1.01,0.04649343967437744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,2,power_law_1.01,0.047298560142517085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,2,power_law_1.01,0.0661836814880371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,2,power_law_1.01,0.06660863876342774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,2,power_law_1.01,0.06762239933013917
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,2,power_law_1.01,0.07013120174407958
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,2,power_law_1.01,0.0713753604888916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,2,power_law_1.01,0.07364863872528075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,2,power_law_1.01,0.07721471786499023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,2,power_law_1.01,0.07933055877685546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,2,power_law_1.01,0.08627584457397461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,2,power_law_1.01,0.09797120094299316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,2,power_law_1.01,0.11050111770629882
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,2,power_law_1.01,0.1390118408203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,2,power_law_1.01,0.1760166358947754
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,2,4,power_law_1.2,0.46813312530517576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,2,power_law_1.01,0.2356928062438965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,2,power_law_1.01,0.2932032012939453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,2,power_law_1.01,0.42358657836914065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,2,power_law_1.01,0.5449075317382812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,2,power_law_1.01,0.7715277099609376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,2,power_law_1.01,1.040279006958008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,2,power_law_1.01,0.02562175989151001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,2,power_law_1.01,0.028880639076232912
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,2,power_law_1.01,0.0325491189956665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,2,power_law_1.01,0.03452415943145752
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,4,2,balanced,0.08336640357971191
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,2,power_law_1.01,0.04025343894958496
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,4,2,balanced,0.07548287868499756
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,2,power_law_1.01,0.04193024158477783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,2,power_law_1.01,0.04255616188049317
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,2,power_law_1.01,0.04868095874786377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,2,power_law_1.01,0.05020031929016113
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,4,2,balanced,0.10708479881286621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,2,power_law_1.01,0.053360638618469236
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,4,2,balanced,0.13607423782348632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,2,power_law_1.01,0.058617601394653315
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,2,power_law_1.01,0.060789761543273924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,2,power_law_1.01,0.06729599952697754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,2,power_law_1.01,0.08025856018066406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,2,power_law_1.01,0.09307135581970215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,2,power_law_1.01,0.04856319904327393
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,2,power_law_1.01,0.13273728370666504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,2,power_law_1.01,0.025208320617675785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,2,power_law_1.01,0.18177536010742187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,4,2,balanced,0.18747392654418946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,2,power_law_1.01,0.2206950378417969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,2,power_law_1.01,0.3951052856445313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,4,2,power_law_1.01,0.02845567941665649
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,4,2,power_law_1.01,0.0288102388381958
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,4,2,power_law_1.01,0.030275840759277344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,4,2,power_law_1.01,0.03224319934844971
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,4,2,power_law_1.01,0.03271039962768555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,2,power_law_1.01,0.5667059326171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,4,2,power_law_1.01,0.03829119920730591
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,4,2,power_law_1.01,0.03895296096801758
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,4,2,power_law_1.01,0.039133439064025874
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,4,2,power_law_1.01,0.04001152038574219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,4,2,power_law_1.01,0.040862717628479005
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,4,2,power_law_1.01,0.03786623954772949
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,4,2,power_law_1.01,0.04150015830993652
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,2,power_law_1.01,0.05596288204193115
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,2,power_law_1.01,0.7520819091796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,4,2,power_law_1.01,0.041776638031005855
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,4,2,power_law_1.01,0.043048958778381344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,4,2,power_law_1.01,0.044290561676025395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,4,2,power_law_1.01,0.05238656044006348
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,4,2,balanced,0.23812223434448243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,2,power_law_1.01,0.11266048431396485
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,4,2,power_law_1.01,0.09288448333740235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,2,power_law_1.01,0.31283327102661135
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,2,power_law_1.01,0.05252480030059814
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,4,2,power_law_1.01,0.04844031810760498
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,4,2,power_law_1.01,0.06726399898529054
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,4,2,power_law_1.01,0.07817344188690185
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,4,2,power_law_1.01,0.12862463951110842
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,4,2,power_law_1.01,0.06023808002471924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,4,2,power_law_1.2,0.02759552001953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,4,2,power_law_1.01,0.14919936180114746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,2,power_law_1.2,0.030115840435028074
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,4,2,power_law_1.01,0.23287296295166016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,2,power_law_1.2,0.03314304113388061
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,2,power_law_1.2,0.04868864059448243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,2,power_law_1.2,0.045666561126708985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,2,power_law_1.2,0.04715007781982422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,2,power_law_1.2,0.06639488220214844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,2,power_law_1.2,0.0670911979675293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,2,power_law_1.2,0.06802815914154052
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,2,power_law_1.2,0.07031551837921143
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,2,power_law_1.2,0.0709222412109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,4,2,power_law_1.01,0.2937830352783203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,2,power_law_1.2,0.0748198413848877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,2,power_law_1.2,0.03666687965393066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,2,power_law_1.2,0.07801343917846679
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,2,power_law_1.2,0.08050559997558594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,2,power_law_1.2,0.08631551742553711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,2,power_law_1.2,0.09880576133728028
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,2,power_law_1.2,0.11108736038208007
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,2,power_law_1.2,0.140382719039917
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,2,power_law_1.2,0.2390105628967285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,2,power_law_1.2,0.30154111862182614
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,2,power_law_1.2,0.17779712677001952
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,2,power_law_1.2,0.42935169219970704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,2,power_law_1.2,0.025395200252532956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,2,power_law_1.2,0.02553344011306763
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,2,power_law_1.2,0.03246975898742675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,2,power_law_1.2,0.03500672101974488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,2,power_law_1.2,0.042961921691894535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,2,power_law_1.2,0.04275455951690674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,4,2,power_law_1.2,0.021388800144195558
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,2,power_law_1.2,0.04316160202026367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,2,power_law_1.2,0.04873472213745117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,2,power_law_1.2,0.04952960014343262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,2,power_law_1.2,0.0497651195526123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,2,power_law_1.2,0.05178880214691162
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,2,power_law_1.2,0.05453184127807618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,2,power_law_1.2,0.05667200088500977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,2,power_law_1.2,1.0529779052734374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,2,power_law_1.2,0.546295051574707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,2,power_law_1.2,0.0599616003036499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,2,power_law_1.2,0.06202879905700683
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,2,power_law_1.2,0.06616064071655273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,2,power_law_1.2,0.08480256080627442
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,2,power_law_1.2,0.09346816062927246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,2,power_law_1.2,0.11347968101501464
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,2,power_law_1.2,0.1346777629852295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,2,power_law_1.2,0.22239999771118163
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,2,power_law_1.2,0.18228096008300781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,2,power_law_1.2,0.029304320812225344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,2,power_law_1.2,0.7850739288330078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,4,2,power_law_1.2,0.029231359958648683
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,4,2,power_law_1.2,0.029939200878143314
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,4,2,power_law_1.2,0.031116800308227537
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,4,2,power_law_1.2,0.03209471940994263
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,4,2,power_law_1.2,0.0379750394821167
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,4,2,power_law_1.2,0.03829504013061523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,2,power_law_1.2,0.31140607833862305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,4,2,power_law_1.2,0.03960448026657105
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,4,2,power_law_1.2,0.039296000003814696
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,4,2,power_law_1.2,0.040317440032958986
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,4,2,power_law_1.2,0.0408076810836792
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,4,2,power_law_1.2,0.04206592082977295
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,4,2,power_law_1.2,0.042389760017395015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,4,2,power_law_1.2,0.04973440170288086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,4,2,power_law_1.2,0.05133823871612549
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,2,power_law_1.2,0.3970406341552734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,4,2,power_law_1.01,0.39399166107177735
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,4,2,power_law_1.01,0.5867801666259765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,2,power_law_1.2,0.5775129699707031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,4,2,power_law_1.2,0.028504319190979004
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,4,2,power_law_1.2,0.05697535991668701
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,4,2,power_law_1.2,0.04643839836120606
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,4,2,power_law_1.2,0.06516992092132569
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,4,2,power_law_1.2,0.09797504425048828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,4,2,power_law_1.2,0.07213696002960204
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,4,2,power_law_1.2,0.13021696090698243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,8,balanced,0.029804799556732174
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,4,2,power_law_1.2,0.16434175491333008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,8,balanced,0.03126912117004395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,8,balanced,0.07658624172210693
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,8,balanced,0.12918399810791015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,8,balanced,0.15576959609985352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,8,balanced,0.16007551193237304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,8,balanced,0.05966976165771485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,8,balanced,0.16121984481811524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,8,balanced,0.16335359573364258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,8,balanced,0.17356927871704103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,8,balanced,0.16219776153564452
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,8,balanced,0.16939519882202148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,8,balanced,0.17346815109252928
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,8,balanced,0.18044160842895507
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,8,balanced,0.18847999572753907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,8,balanced,0.19536895751953126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,8,balanced,0.19422079086303712
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,2,power_law_1.2,0.7609190368652344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,8,balanced,0.24647424697875975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,8,balanced,0.25928960800170897
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,8,balanced,0.37493377685546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,8,balanced,0.43024768829345705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,8,balanced,0.5870950317382813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,8,balanced,0.7362649536132813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,4,2,power_law_1.2,0.21136640548706054
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,8,balanced,1.0701042938232423
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,8,balanced,0.030484480857849123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,8,balanced,0.031718399524688724
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,8,balanced,1.3960768127441407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,8,balanced,0.03503999948501587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,8,balanced,0.05065087795257568
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,8,balanced,0.08374655723571778
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,8,balanced,0.09290623664855957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,8,balanced,0.09482111930847167
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,8,balanced,0.09814144134521484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,8,balanced,0.10019071578979492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,8,balanced,0.10324735641479492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,8,balanced,0.09457535743713379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,8,balanced,0.09780863761901856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,8,balanced,0.09881600379943847
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,8,balanced,0.10283904075622559
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,8,balanced,0.10464127540588379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,8,balanced,0.10742015838623047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,8,balanced,0.12182911872863769
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,8,balanced,0.13959808349609376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,8,balanced,0.1542156791687012
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,8,balanced,2.974869689941406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,8,balanced,0.19050752639770507
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,8,balanced,2.134882507324219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,8,balanced,0.22394111633300778
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,8,balanced,0.31810047149658205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,8,balanced,0.3845772933959961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,8,balanced,0.5347622299194337
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,8,balanced,0.7052416229248047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,8,balanced,1.019240951538086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,8,balanced,1.3397952270507811
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,4,2,power_law_1.2,0.4126566314697266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,4,2,power_law_1.2,0.31697792053222656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,4,2,power_law_1.2,0.6194611358642578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,4,2,power_law_1.2,0.7608345794677734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,1,8,balanced,0.020811519622802734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,1,8,balanced,0.020211200714111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,1,8,balanced,0.050992641448974606
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,1,8,balanced,0.05121407985687256
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,1,8,balanced,0.0512499189376831
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,1,8,balanced,0.05146751880645752
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,1,8,balanced,0.05135615825653076
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,1,8,balanced,0.05219840049743653
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,1,8,balanced,0.05188223838806152
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,1,8,balanced,0.05286911964416504
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,1,8,balanced,0.052181758880615235
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,1,8,balanced,0.053939199447631835
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,1,8,balanced,0.03303936004638672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,1,8,balanced,0.054675202369689944
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,1,8,balanced,0.022338559627532957
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,1,8,balanced,0.05793536186218262
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,1,8,balanced,0.06023168087005615
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,1,8,balanced,0.05062528133392334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,1,8,balanced,0.05186304092407227
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,1,8,balanced,0.05178112030029297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,8,power_law_1.01,0.07482751846313476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,8,power_law_1.01,0.08089599609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,8,power_law_1.01,0.09783935546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,8,power_law_1.01,0.10086144447326659
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,8,power_law_1.01,0.09988991737365722
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,8,power_law_1.01,0.1352243232727051
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,8,power_law_1.01,0.14410112380981446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,8,power_law_1.01,0.1508198356628418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,8,power_law_1.01,0.15223039627075194
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,8,power_law_1.01,0.15631360054016114
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,8,power_law_1.01,0.16925823211669921
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,8,power_law_1.01,0.16631168365478516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,8,power_law_1.01,0.16025983810424804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,8,power_law_1.01,0.17823104858398436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,8,power_law_1.01,0.18884607315063479
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,8,power_law_1.01,0.2029734420776367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,8,power_law_1.01,0.2346124839782715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,8,power_law_1.01,0.277653751373291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,8,power_law_1.01,0.32621055603027344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,8,power_law_1.01,0.4321561431884765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,8,power_law_1.01,0.4836851119995117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,8,power_law_1.01,0.687857894897461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,8,power_law_1.01,0.8802393341064454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,8,power_law_1.01,1.3890419006347656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,8,power_law_1.01,1.777775421142578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,8,power_law_1.01,2.564185485839844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,1,8,balanced,0.05238016128540039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,8,power_law_1.01,0.04984576225280762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,8,power_law_1.01,0.06157824039459229
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,8,power_law_1.01,0.07059840202331544
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,1,8,balanced,0.08156031608581543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,8,power_law_1.01,0.06353024005889893
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,1,8,balanced,0.11124863624572753
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,1,8,balanced,0.13513471603393554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,8,power_law_1.01,0.05964032173156738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,8,power_law_1.01,0.08850815773010254
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,1,8,balanced,0.08438400268554688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,8,power_law_1.01,0.094967041015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,8,power_law_1.01,0.09288576126098633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,8,power_law_1.01,0.0960524845123291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,8,power_law_1.01,0.10080127716064453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,8,power_law_1.01,0.10060671806335449
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,8,power_law_1.01,0.10090111732482909
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,8,power_law_1.01,0.0970854377746582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,8,power_law_1.01,0.10238080024719239
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,8,power_law_1.01,0.10477696418762208
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,8,power_law_1.01,0.11054976463317871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,8,power_law_1.01,0.13970815658569335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,8,power_law_1.01,0.15630592346191405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,8,power_law_1.01,0.2168383979797363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,8,power_law_1.01,3.5797427368164065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,8,power_law_1.01,0.23887487411499025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,8,power_law_1.01,0.3383091354370117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,8,power_law_1.01,0.10043775558471679
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,8,power_law_1.01,0.44116992950439454
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,1,8,balanced,0.18780672073364257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,8,power_law_1.01,0.6095897674560546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,1,8,power_law_1.01,0.032257280349731444
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,1,8,power_law_1.01,0.04107391834259033
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,1,8,power_law_1.01,0.0448422384262085
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,1,8,power_law_1.01,0.03863679885864258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,8,power_law_1.01,0.7343283081054688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,1,8,power_law_1.01,0.04093567848205566
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,1,8,power_law_1.01,0.04857215881347656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,1,8,power_law_1.01,0.0494374418258667
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,1,8,power_law_1.01,0.04957695960998535
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,1,8,power_law_1.01,0.04973440170288086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,1,8,power_law_1.01,0.051326718330383304
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,1,8,power_law_1.01,0.05076479911804199
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,1,8,power_law_1.01,0.051429119110107425
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,8,power_law_1.01,1.2069913482666015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,1,8,power_law_1.01,0.053454079627990723
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,1,8,power_law_1.01,0.054120960235595705
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,1,8,power_law_1.01,0.054967041015624996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,8,power_law_1.01,1.4914714050292968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,1,8,power_law_1.01,0.052006402015686036
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,1,8,balanced,0.24593280792236327
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,1,8,power_law_1.01,0.07799680233001709
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,1,8,power_law_1.01,0.058015999794006345
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,1,8,power_law_1.01,0.08913151741027832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,1,8,power_law_1.01,0.13521280288696289
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,1,8,power_law_1.01,0.10053759574890138
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,8,power_law_1.2,0.07215231895446778
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,8,power_law_1.2,0.07537792205810548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,8,power_law_1.2,0.09283072471618652
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,8,power_law_1.2,0.08336128234863281
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,1,8,power_law_1.01,0.16637439727783204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,8,power_law_1.2,0.10758144378662109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,8,power_law_1.2,0.13689727783203126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,8,power_law_1.2,0.14337920188903808
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,1,8,power_law_1.01,0.06355967998504639
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,8,power_law_1.2,0.15548928260803224
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,1,8,power_law_1.01,0.22443519592285158
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,8,power_law_1.2,0.15733375549316406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,8,power_law_1.2,0.15854080200195314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,8,power_law_1.2,0.15803520202636717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,8,power_law_1.2,0.1719539260864258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,8,power_law_1.2,0.16320640563964844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,8,power_law_1.2,0.1808358383178711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,8,power_law_1.2,0.21728639602661132
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,8,power_law_1.2,0.22518272399902345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,8,power_law_1.2,0.2437491226196289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,8,power_law_1.2,0.2819302368164062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,8,power_law_1.2,0.32417663574218747
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,8,power_law_1.2,0.44796798706054686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,8,power_law_1.2,0.5958387374877929
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,8,power_law_1.2,0.7254707336425781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,8,power_law_1.2,1.007041244506836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,8,power_law_1.2,0.04821760177612305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,8,power_law_1.2,0.05043839931488037
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,8,power_law_1.2,1.4960421752929687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,8,power_law_1.2,0.06427775859832764
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,8,power_law_1.2,0.058728961944580084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,8,power_law_1.2,0.0582374382019043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,8,power_law_1.2,0.0891545581817627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,8,power_law_1.2,0.09053695678710938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,8,power_law_1.2,0.09484543800354003
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,8,power_law_1.2,0.10173312187194825
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,8,power_law_1.2,0.10436223983764648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,8,power_law_1.2,1.9400845336914063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,8,power_law_1.2,0.1003007984161377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,8,power_law_1.2,0.09807999610900879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,8,power_law_1.2,0.10178943634033204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,8,power_law_1.2,0.0980838394165039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,8,power_law_1.2,0.1043776035308838
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,8,power_law_1.2,0.10912639617919921
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,8,power_law_1.2,0.11643263816833496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,8,power_law_1.2,0.1444915199279785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,8,power_law_1.2,0.16810367584228517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,8,power_law_1.2,0.26255615234374996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,8,power_law_1.2,0.21831296920776366
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,8,power_law_1.2,0.4730112075805664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,8,power_law_1.2,0.3666368103027344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,8,power_law_1.2,2.8457907104492186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,8,power_law_1.2,0.8716287994384766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,1,8,power_law_1.2,0.032556800842285155
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,1,8,power_law_1.2,0.03607295989990235
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,1,8,power_law_1.2,0.03982592105865478
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,8,power_law_1.2,0.6986867523193359
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,1,8,power_law_1.2,0.04528639793395996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,1,8,power_law_1.2,0.03799295902252197
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,1,8,power_law_1.2,0.04878848075866699
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,1,8,power_law_1.2,0.04963200092315674
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,1,8,power_law_1.2,0.05101439952850342
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,1,8,power_law_1.2,0.051270399093627926
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,1,8,power_law_1.2,0.05165952205657959
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,1,8,power_law_1.2,0.05117695808410645
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,1,8,power_law_1.2,0.05178880214691162
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,1,8,power_law_1.2,0.05236991882324219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,8,power_law_1.2,1.26052734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,1,8,power_law_1.01,0.3208972930908203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,1,8,power_law_1.01,0.41465599060058594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,1,8,power_law_1.01,0.703746566772461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,8,power_law_1.2,4.145813903808594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,1,8,power_law_1.2,0.05454592227935791
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,1,8,power_law_1.2,0.054686717987060554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,8,power_law_1.2,1.6941325378417968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,1,8,power_law_1.2,0.05757440090179443
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,1,8,power_law_1.2,0.06091775894165039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,1,8,power_law_1.2,0.07280896186828614
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,1,8,power_law_1.2,0.08089344024658203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,1,8,power_law_1.2,0.0967078399658203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,4,balanced,0.028874239921569823
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,4,balanced,0.04200831890106201
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,4,balanced,0.05689856052398682
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,4,balanced,0.07843455791473389
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,4,balanced,0.1303859233856201
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,4,balanced,0.13688703536987307
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,4,balanced,0.14761216163635255
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,4,balanced,0.15158528327941895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,4,balanced,0.1532953643798828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,4,balanced,0.1575641632080078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,4,balanced,0.14994688034057618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,4,balanced,0.15246335983276366
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,4,balanced,0.15794943809509276
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,4,balanced,0.16030975341796877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,4,balanced,0.165032958984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,4,balanced,0.17018751144409178
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,4,balanced,0.1800217628479004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,4,balanced,0.20272640228271482
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,4,balanced,0.22696319580078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,4,balanced,0.3335820770263672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,1,8,power_law_1.2,0.11585023880004883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,4,balanced,0.38097793579101563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,1,8,power_law_1.2,0.17094144821166993
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,4,balanced,0.5224076843261718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,4,balanced,0.6559410858154296
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,1,8,power_law_1.2,0.23631744384765624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,4,balanced,0.9495718383789062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,4,balanced,0.031674880981445316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,4,balanced,1.2406195068359376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,4,balanced,0.031765758991241455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,4,balanced,0.039310081005096434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,4,balanced,0.05548031806945801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,4,balanced,0.08980223655700684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,4,balanced,0.0929804801940918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,4,balanced,0.1025881576538086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,4,balanced,0.10618111610412598
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,4,balanced,0.1076915168762207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,4,balanced,0.11152640342712403
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,4,balanced,1.8175871276855466
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,4,balanced,0.09986687660217285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,4,balanced,0.10306688308715821
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,4,balanced,0.10401280403137207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,4,balanced,0.10652799606323242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,4,balanced,0.10766336441040039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,4,balanced,0.10968704223632812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,4,balanced,0.11798912048339845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,4,balanced,0.13879039764404297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,4,balanced,2.403363800048828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,4,balanced,0.14887295722961427
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,4,balanced,0.18129152297973633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,4,balanced,0.211907844543457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,4,balanced,0.2947916793823242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,4,balanced,0.35036544799804686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,4,balanced,0.494890251159668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,4,balanced,0.6493055725097656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,4,balanced,0.9309760284423827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,4,balanced,1.2267890930175782
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,1,8,power_law_1.2,0.30616064071655275
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,1,8,power_law_1.2,0.4228736114501953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,1,8,power_law_1.2,0.6742707061767578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,2,4,balanced,0.022050559520721436
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,2,4,balanced,0.054511361122131355
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,2,4,balanced,0.05496448040008545
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,2,4,balanced,0.05486591815948486
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,2,4,balanced,0.023125760555267334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,2,4,balanced,0.05559423923492431
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,2,4,balanced,0.055107841491699214
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,2,4,balanced,0.054862079620361326
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,2,4,balanced,0.055509757995605466
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,2,4,balanced,0.02554368019104004
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,2,4,balanced,0.05640192031860351
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,2,4,balanced,0.057429761886596674
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,2,4,balanced,0.058314242362976075
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,2,4,balanced,0.06125696182250977
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,1,8,power_law_1.2,0.8529036712646484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,2,4,balanced,0.03673856019973755
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,2,4,balanced,0.062449917793273926
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,2,4,balanced,0.06639232158660888
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,2,4,balanced,0.07057919979095459
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,2,4,balanced,0.053300480842590335
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,2,4,balanced,0.053730559349060056
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,4,power_law_1.01,0.05472896099090576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,4,power_law_1.01,0.05847424030303955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,4,power_law_1.01,0.07267072200775146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,4,power_law_1.01,0.08140031814575195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,4,power_law_1.01,0.09446911811828614
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,4,power_law_1.01,0.11896191596984865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,4,power_law_1.01,0.13788928031921385
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,4,power_law_1.01,0.14384896278381348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,4,power_law_1.01,0.1476364803314209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,4,power_law_1.01,0.15224831581115722
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,4,power_law_1.01,0.15386624336242677
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,4,power_law_1.01,0.14812031745910645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,4,power_law_1.01,0.14894463539123534
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,4,power_law_1.01,0.15870847702026367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,4,power_law_1.01,0.1684889602661133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,4,power_law_1.01,0.1709619140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,4,power_law_1.01,0.19094528198242186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,4,power_law_1.01,0.2553651237487793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,4,power_law_1.01,0.28592384338378907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,4,power_law_1.01,0.3711552047729492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,4,power_law_1.01,0.44911361694335933
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,4,power_law_1.01,0.6136140823364258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,4,power_law_1.01,0.7674854278564454
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,2,4,balanced,0.05467391967773437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,4,power_law_1.01,1.107663345336914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,4,power_law_1.01,1.5363711547851562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,4,power_law_1.01,2.1242214965820314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,4,power_law_1.01,0.032529919147491454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,4,power_law_1.01,0.04773248195648193
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,4,power_law_1.01,0.05402368068695068
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,4,power_law_1.01,0.056491518020629884
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,4,power_law_1.01,2.71038330078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,4,power_law_1.01,0.0709875202178955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,4,power_law_1.01,0.08978560447692871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,4,power_law_1.01,0.09834624290466308
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,4,power_law_1.01,0.10142208099365235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,4,power_law_1.01,0.10598015785217285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,4,power_law_1.01,0.10928511619567871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,4,power_law_1.01,0.1075648021697998
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,2,4,balanced,0.17711360931396486
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,2,4,balanced,0.097194242477417
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,4,power_law_1.01,0.10129664421081544
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,2,4,balanced,0.1057804775238037
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,4,power_law_1.01,0.10416895866394042
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,4,power_law_1.01,0.10431743621826173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,4,power_law_1.01,0.10667136192321777
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,4,power_law_1.01,0.10942591667175292
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,2,4,balanced,0.25404672622680663
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,4,power_law_1.01,0.11874431610107421
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,4,power_law_1.01,0.14107263565063477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,4,power_law_1.01,0.1526694393157959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,4,power_law_1.01,0.20089599609375003
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,4,power_law_1.01,0.23589887619018554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,4,power_law_1.01,0.3164006423950195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,4,power_law_1.01,0.3872921752929687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,2,4,power_law_1.01,0.024360959529876706
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,4,power_law_1.01,0.5508172988891602
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,2,4,power_law_1.01,0.030522880554199217
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,2,4,power_law_1.01,0.032636160850524905
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,2,4,power_law_1.01,0.037978880405426026
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,2,4,power_law_1.01,0.03879168033599854
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,2,4,balanced,0.13994624137878417
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,2,4,power_law_1.01,0.05027840137481689
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,2,4,power_law_1.01,0.051595520973205564
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,2,4,power_law_1.01,0.0523199987411499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,4,power_law_1.01,0.9990745544433594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,2,4,power_law_1.01,0.053740801811218264
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,2,4,power_law_1.01,0.05305471897125245
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,2,4,power_law_1.01,0.053958401679992676
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,2,4,power_law_1.01,0.056053757667541504
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,2,4,power_law_1.01,0.05788288116455078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,2,4,power_law_1.01,0.05950079917907715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,4,power_law_1.01,1.3059519958496093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,2,4,power_law_1.01,0.06631423950195312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,2,4,power_law_1.01,0.07220992088317871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,4,power_law_1.01,0.7050457763671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,2,4,balanced,0.33269630432128905
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,2,4,power_law_1.01,0.062152957916259764
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,2,4,power_law_1.01,0.057034239768981934
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,2,4,power_law_1.01,0.10666624069213868
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,2,4,power_law_1.01,0.08156160354614259
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,2,4,power_law_1.01,0.12362624168395997
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,2,4,power_law_1.01,0.20939775466918947
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,4,power_law_1.2,0.05549952030181885
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,4,power_law_1.2,0.05963391780853271
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,2,4,power_law_1.01,0.17295103073120116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,4,power_law_1.2,0.06976255893707276
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,4,power_law_1.2,0.07181312084197998
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,2,4,power_law_1.01,0.2867353630065918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,4,power_law_1.2,0.08773759841918946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,4,power_law_1.2,0.13903103828430177
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,4,power_law_1.2,0.14622207641601564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,4,power_law_1.2,0.15363200187683107
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,4,power_law_1.2,0.15278207778930664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,4,power_law_1.2,0.15611136436462403
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,4,power_law_1.2,0.15114368438720702
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,4,power_law_1.2,0.1567296028137207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,4,power_law_1.2,0.17503616333007813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,4,power_law_1.2,0.175545597076416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,4,power_law_1.2,0.20529151916503907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,4,power_law_1.2,0.2609971237182617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,4,power_law_1.2,0.11873536109924317
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,4,power_law_1.2,0.2940108871459961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,4,power_law_1.2,0.3776947021484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,4,power_law_1.2,0.16296831130981446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,4,power_law_1.2,0.47534206390380856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,4,power_law_1.2,0.7943795013427735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,4,power_law_1.2,0.6329536056518554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,4,power_law_1.2,1.2014374542236328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,4,power_law_1.2,0.03292160034179688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,4,power_law_1.2,0.03910527944564819
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,4,power_law_1.2,0.05051904201507569
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,4,power_law_1.2,0.05976704120635986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,4,power_law_1.2,0.061224961280822755
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,4,power_law_1.2,0.09283455848693847
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,4,power_law_1.2,0.10028927803039552
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,4,power_law_1.2,0.10505087852478028
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,4,power_law_1.2,0.10619903564453126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,4,power_law_1.2,1.554897918701172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,4,power_law_1.2,0.10892671585083007
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,4,power_law_1.2,0.10330752372741699
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,4,power_law_1.2,0.10201472282409667
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,4,power_law_1.2,0.10324607849121095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,4,power_law_1.2,0.10660863876342774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,4,power_law_1.2,0.10889599800109864
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,4,power_law_1.2,0.11124608039855957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,4,power_law_1.2,0.11986175537109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,4,power_law_1.2,0.14304896354675295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,4,power_law_1.2,0.15526016235351564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,4,power_law_1.2,0.20312320709228518
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,4,power_law_1.2,0.24656639099121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,4,power_law_1.2,0.3261606216430664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,4,power_law_1.2,0.3999001693725586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,4,power_law_1.2,2.2335360717773436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,4,power_law_1.2,0.7245005035400391
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,2,4,power_law_1.2,0.024687359333038332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,2,4,power_law_1.2,0.028042240142822267
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,2,4,power_law_1.2,0.032463359832763675
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,2,4,power_law_1.2,0.035829761028289796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,4,power_law_1.2,0.5857907104492187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,2,4,power_law_1.2,0.03896960020065308
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,2,4,power_law_1.2,0.05107583999633789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,2,4,power_law_1.2,0.052055039405822755
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,2,4,power_law_1.2,0.05221759796142578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,2,4,power_law_1.2,0.0528652811050415
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,2,4,power_law_1.2,0.054271998405456545
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,2,4,power_law_1.2,0.05656576156616211
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,2,4,power_law_1.2,0.05604479789733887
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,2,4,power_law_1.2,0.05717887878417969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,2,4,power_law_1.2,0.060613121986389164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,4,power_law_1.2,1.0819225311279297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,2,4,power_law_1.01,0.3854912185668945
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,2,4,power_law_1.01,0.827504653930664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,2,4,power_law_1.01,0.5829452896118165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,4,power_law_1.2,3.032525939941406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,2,4,power_law_1.2,0.06260863780975343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,4,power_law_1.2,1.4132864379882812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,2,4,power_law_1.2,0.06550271987915039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,2,4,power_law_1.2,0.06895232200622559
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,2,4,power_law_1.2,0.08102527618408203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,2,4,power_law_1.2,0.10578816413879395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,2,4,power_law_1.2,0.1304089641571045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,2,balanced,0.026691839694976804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,2,balanced,0.042365441322326655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,2,balanced,0.055256319046020505
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,2,balanced,0.0748902416229248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,2,balanced,0.12157183647155763
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,2,balanced,0.12142975807189942
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,2,balanced,0.12389504432678222
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,2,balanced,0.12553471565246582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,2,balanced,0.14076031684875487
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,2,balanced,0.14141695976257324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,2,balanced,0.1430361557006836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,2,balanced,0.1475648021697998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,2,balanced,0.14844672203063963
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,2,balanced,0.15205760002136232
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,2,balanced,0.1552729606628418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,2,balanced,0.15973631858825682
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,2,balanced,0.1695372772216797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,2,balanced,0.1924287986755371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,2,balanced,0.21526784896850587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,2,balanced,0.3106547164916992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,2,balanced,0.3568588638305664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,2,balanced,0.4878630447387695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,2,balanced,0.6104422378540039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,2,balanced,0.8876825714111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,2,balanced,1.153876495361328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,2,balanced,1.6927897644042968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,2,4,power_law_1.2,0.16033920288085937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,2,balanced,0.02925823926925659
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,2,balanced,0.030969600677490234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,2,balanced,0.04077184200286865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,2,balanced,2.236492767333984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,2,balanced,0.06327936172485352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,2,balanced,0.09723391532897949
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,2,balanced,0.0984985637664795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,2,balanced,0.09934080123901368
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,2,balanced,0.10013440132141113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,2,balanced,0.11430656433105468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,2,balanced,0.11543040275573731
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,2,4,power_law_1.2,0.21436159133911134
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,2,balanced,0.11693951606750488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,2,balanced,0.11815423965454101
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,2,balanced,0.11935232162475587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,2,balanced,0.1217625617980957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,2,balanced,0.12349056243896483
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,2,balanced,0.12552063941955566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,2,balanced,0.13131903648376464
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,2,balanced,0.14272000312805175
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,2,balanced,0.15729408264160155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,2,balanced,0.17990528106689455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,2,balanced,0.20959871292114257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,2,balanced,0.29885568618774416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,2,balanced,0.3488780975341797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,2,balanced,0.49212158203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,2,4,power_law_1.2,0.2782924842834473
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,2,balanced,0.6489024353027344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,2,4,power_law_1.2,0.3814156723022461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,2,balanced,0.9350784301757813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,2,balanced,1.2333106994628906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,2,4,power_law_1.2,0.5831321716308594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,2,4,power_law_1.2,0.7702438354492187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,4,2,balanced,0.028856320381164553
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,4,2,balanced,0.06883200168609618
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,4,2,balanced,0.06900352001190185
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,4,2,balanced,0.04668799877166748
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,4,2,balanced,0.06876031875610351
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,4,2,balanced,0.06869247913360596
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,4,2,balanced,0.06911871910095216
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,4,2,balanced,0.07044608116149902
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,4,2,balanced,0.03374975919723511
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,4,2,balanced,0.07107711791992187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,4,2,balanced,0.07217023849487306
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,4,2,balanced,0.07397503852844238
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,4,2,balanced,0.07537280082702637
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,4,2,balanced,0.07905151844024658
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,4,2,balanced,0.06715007781982421
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,4,2,balanced,0.08277119636535644
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,4,2,balanced,0.08972800254821778
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,4,2,balanced,0.0971878433227539
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,4,2,balanced,0.029799680709838867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,4,2,balanced,0.06724224090576172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,4,2,balanced,0.0680460786819458
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,2,power_law_1.01,0.035617280006408694
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,2,power_law_1.01,0.04730495929718018
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,2,power_law_1.01,0.060568318367004395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,2,power_law_1.01,0.07073919773101807
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,2,4,power_law_1.2,1.1327948760986328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,2,power_law_1.01,0.08121472358703613
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,2,power_law_1.01,0.10975104331970216
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,2,power_law_1.01,0.10722175598144532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,2,power_law_1.01,0.11456000328063964
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,2,power_law_1.01,0.13524607658386228
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,2,power_law_1.01,0.13734016418457032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,2,power_law_1.01,0.14049792289733887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,2,power_law_1.01,0.14417535781860352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,2,power_law_1.01,0.1443404769897461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,2,power_law_1.01,0.15321599960327148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,2,power_law_1.01,0.15735039710998536
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,2,power_law_1.01,0.1631385612487793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,2,power_law_1.01,0.1759270477294922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,2,power_law_1.01,0.24522367477416993
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,2,power_law_1.01,0.2685759925842285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,2,power_law_1.01,0.3216345596313476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,2,power_law_1.01,0.4110464096069336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,2,power_law_1.01,0.5340044784545899
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,2,power_law_1.01,0.673364486694336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,2,power_law_1.01,0.9728793334960937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,2,power_law_1.01,1.2789401245117187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,2,power_law_1.01,1.8252044677734376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,4,2,balanced,0.13759488105773926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,2,power_law_1.01,0.030108160972595215
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,4,2,balanced,0.2734476852416992
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,4,2,balanced,0.21134719848632813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,4,2,balanced,0.1555788803100586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,2,power_law_1.01,0.03885695934295654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,2,power_law_1.01,0.048046078681945804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,2,power_law_1.01,0.05730175971984863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,2,power_law_1.01,0.06647679805755616
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,2,power_law_1.01,0.08858367919921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,2,power_law_1.01,0.09415040016174317
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,2,power_law_1.01,0.10892671585083007
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,2,power_law_1.01,0.11158143997192382
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,2,power_law_1.01,0.11364607810974121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,2,power_law_1.01,0.11707136154174805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,4,2,balanced,0.39426944732666014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,2,power_law_1.01,0.11812352180480956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,2,power_law_1.01,2.411295928955078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,2,power_law_1.01,0.12187135696411133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,2,power_law_1.01,0.12284288406372071
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,2,power_law_1.01,0.12516863822937013
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,2,power_law_1.01,0.14937600135803222
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,2,power_law_1.01,0.13257599830627442
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,2,power_law_1.01,0.16597759246826171
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,2,power_law_1.01,0.20576128005981445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,2,power_law_1.01,0.24650367736816406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,2,power_law_1.01,0.32142463684082034
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,2,power_law_1.01,0.38901119232177733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,2,power_law_1.01,0.09603584289550782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,2,power_law_1.01,0.5482534408569336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,4,2,power_law_1.01,0.03010688066482544
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,4,2,power_law_1.01,0.033634560108184816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,2,power_law_1.01,0.7050521850585938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,4,2,power_law_1.01,0.0384768009185791
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,4,2,power_law_1.01,0.044445438385009764
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,4,2,power_law_1.01,0.04999680042266846
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,4,2,power_law_1.01,0.06443264007568358
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,4,2,power_law_1.01,0.06541056156158447
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,4,2,power_law_1.01,0.0658188819885254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,2,power_law_1.01,0.9914342498779296
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,4,2,power_law_1.01,0.06695168018341065
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,4,2,power_law_1.01,0.06912896156311035
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,4,2,power_law_1.01,0.07007872104644776
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,4,2,power_law_1.01,0.07255296230316162
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,4,2,power_law_1.01,0.07336319923400879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,2,power_law_1.01,1.3220684814453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,4,2,power_law_1.01,0.07672704219818115
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,4,2,power_law_1.01,0.07904128074645997
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,4,2,power_law_1.01,0.08601471900939942
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,4,2,balanced,0.5160780715942382
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,4,2,power_law_1.01,0.06278528213500977
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,4,2,power_law_1.01,0.10069631576538085
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,4,2,power_law_1.01,0.181712646484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,4,2,power_law_1.01,0.1414899158477783
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,4,2,power_law_1.01,0.11775103569030762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,2,power_law_1.2,0.03333247900009155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,2,power_law_1.2,0.04735487937927246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,2,power_law_1.2,0.05967872142791748
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,4,2,power_law_1.01,0.24389888763427736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,2,power_law_1.2,0.06606336116790772
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,2,power_law_1.2,0.07809792041778565
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,4,2,power_law_1.01,0.30824960708618165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,2,power_law_1.2,0.11160063743591309
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,2,power_law_1.2,0.11117568016052246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,2,power_law_1.2,0.1216089630126953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,2,power_law_1.2,0.13985152244567872
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,2,power_law_1.2,0.13828736305236816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,2,power_law_1.2,0.1451136016845703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,2,power_law_1.2,0.1472435188293457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,2,power_law_1.2,0.14985471725463867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,2,power_law_1.2,0.15470208168029784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,2,power_law_1.2,0.1625484848022461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,2,power_law_1.2,0.16787200927734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,2,power_law_1.2,0.18420480728149413
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,2,power_law_1.2,0.24445568084716798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,2,power_law_1.2,0.2732825660705566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,2,power_law_1.2,0.3262886428833008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,2,power_law_1.2,0.42019584655761716
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,2,power_law_1.2,0.6833805084228516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,2,power_law_1.2,0.9958566284179687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,2,power_law_1.2,0.5449164962768555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,2,power_law_1.2,0.030105600357055663
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,2,power_law_1.2,0.03886847972869873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,2,power_law_1.2,0.04451583862304688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,2,power_law_1.2,0.05772031784057617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,2,power_law_1.2,1.2998463439941408
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,2,power_law_1.2,0.06156799793243408
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,2,power_law_1.2,0.08996479988098145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,2,power_law_1.2,0.09567744255065917
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,2,power_law_1.2,0.09856512069702147
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,2,power_law_1.2,0.11175552368164063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,2,power_law_1.2,0.11419520378112794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,2,power_law_1.2,0.11399423599243164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,2,power_law_1.2,0.11782143592834474
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,2,power_law_1.2,0.11913984298706055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,2,power_law_1.2,0.12333056449890137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,2,power_law_1.2,0.1238745594024658
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,2,power_law_1.2,0.1268825626373291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,2,power_law_1.2,0.13240960121154785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,2,power_law_1.2,0.15232768058776855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,2,power_law_1.2,0.16720767974853518
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,2,power_law_1.2,0.2080998420715332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,2,power_law_1.2,0.24991615295410155
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,4,2,power_law_1.01,0.4886924743652344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,2,power_law_1.2,0.3263052749633789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,2,power_law_1.2,0.3964531326293945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,2,power_law_1.2,1.8865023803710936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,2,power_law_1.2,0.7157708740234374
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,4,2,power_law_1.2,0.030105600357055663
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,4,2,power_law_1.2,0.0339302396774292
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,4,2,power_law_1.2,0.03705471992492676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,2,power_law_1.2,0.5639193725585938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,4,2,power_law_1.2,0.04272640228271484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,4,2,power_law_1.2,0.04584447860717773
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,4,2,power_law_1.2,0.06255743980407716
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,4,2,power_law_1.2,0.06428927898406982
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,4,2,power_law_1.2,0.06507775783538819
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,4,2,power_law_1.2,0.066626558303833
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,4,2,power_law_1.2,0.06766719818115234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,4,2,power_law_1.2,0.06950016021728515
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,4,2,power_law_1.2,0.07112832069396972
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,4,2,power_law_1.2,0.0722815990447998
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,4,2,power_law_1.2,0.07627520084381104
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,2,power_law_1.2,1.0067327880859376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,4,2,power_law_1.01,0.5822592163085938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,4,2,power_law_1.01,1.2384550476074219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,4,2,power_law_1.01,0.9477490997314453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,4,2,power_law_1.2,0.1347980785369873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,2,power_law_1.2,2.5216549682617186
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,4,2,power_law_1.2,0.07965695858001709
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,4,2,power_law_1.2,0.08260352134704589
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,2,power_law_1.2,1.3616026306152345
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,4,2,power_law_1.2,0.09381888389587403
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,4,2,power_law_1.2,0.1260646438598633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,4,2,power_law_1.2,0.17092479705810545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,8,1,balanced,0.0348908805847168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,8,1,balanced,0.050786561965942376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,8,1,balanced,0.06661632061004638
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,8,1,balanced,0.0876921558380127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,8,1,balanced,0.14020928382873535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,8,1,balanced,0.14636223793029784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,8,1,balanced,0.138603515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,8,1,balanced,0.14013119697570803
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,4,2,power_law_1.2,0.2086476707458496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,8,1,balanced,0.1414342403411865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,8,1,balanced,0.14270591735839844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,8,1,balanced,0.1437728023529053
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,8,1,balanced,0.17545791625976564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,8,1,balanced,0.17770624160766602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,8,1,balanced,0.18263744354248046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,8,1,balanced,0.18540735244750978
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,8,1,balanced,0.18939071655273437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,8,1,balanced,0.19811967849731443
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,8,1,balanced,0.21675840377807618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,8,1,balanced,0.23660608291625979
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,8,1,balanced,0.3336659240722656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,8,1,balanced,0.37391742706298825
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,8,1,balanced,0.5002444839477539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,8,1,balanced,0.61727294921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,8,1,balanced,0.881712646484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,8,1,balanced,1.1366342163085936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,8,1,balanced,1.668113250732422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,8,1,balanced,0.036917760372161865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,8,1,balanced,2.1710733032226566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,8,1,balanced,0.041886081695556646
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,8,1,balanced,0.054580478668212896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,8,1,balanced,0.08166912078857422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,8,1,balanced,0.13862784385681154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,8,1,balanced,0.1374022388458252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,8,1,balanced,0.13849023818969725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,8,1,balanced,0.14013567924499512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,8,1,balanced,0.14063936233520508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,8,1,balanced,0.14272640228271485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,8,1,balanced,0.16336063385009764
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,8,1,balanced,0.16270143508911133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,8,1,balanced,0.16598783493041994
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,8,1,balanced,0.16545791625976564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,8,1,balanced,0.16878528594970704
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,4,2,power_law_1.2,0.34636161804199217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,8,1,balanced,0.1730067253112793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,8,1,balanced,0.18410175323486327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,8,1,balanced,0.13710911750793459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,8,1,balanced,0.1963532829284668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,8,1,balanced,0.21737728118896485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,8,1,balanced,0.24185535430908206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,8,1,balanced,0.3663455963134766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,4,2,power_law_1.2,0.43558399200439457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,8,1,balanced,0.41089534759521484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,8,1,balanced,0.5807526397705078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,8,1,balanced,0.7399667358398437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,8,1,balanced,1.0844230651855469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,8,1,balanced,1.425345916748047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,4,2,power_law_1.2,0.8868479919433593
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,4,2,power_law_1.2,0.6714956665039062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,8,1,balanced,0.04198912143707275
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,8,1,balanced,0.04470655918121338
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,8,1,balanced,0.09409088134765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,8,1,balanced,0.09361151695251466
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,8,1,balanced,0.050472321510314944
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,8,1,balanced,0.0677516794204712
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,8,1,balanced,0.09445183753967286
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,8,1,balanced,0.09533120155334472
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,8,1,balanced,0.09627776145935059
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,8,1,balanced,0.0972652816772461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,8,1,balanced,0.09829952239990233
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,8,1,balanced,0.10024895668029785
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,8,1,balanced,0.10156031608581544
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,8,1,balanced,0.10415360450744629
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,8,1,balanced,0.11345536231994628
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,8,1,balanced,0.10704000473022461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,8,1,balanced,0.11927040100097656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,8,1,balanced,0.1331264019012451
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,8,1,balanced,0.1473516845703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,8,1,balanced,0.09373632431030274
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,4,2,power_law_1.2,1.252099838256836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,8,1,balanced,0.09311424255371094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,8,1,power_law_1.01,0.03616447925567627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,8,1,power_law_1.01,0.054650239944458015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,8,1,power_law_1.01,0.06655680179595948
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,8,1,power_law_1.01,0.07758272171020508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,8,1,power_law_1.01,0.09177536010742188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,8,1,power_law_1.01,0.12360447883605956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,8,1,power_law_1.01,0.12868736267089845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,8,1,power_law_1.01,0.13251328468322754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,8,1,power_law_1.01,0.13878335952758789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,8,1,power_law_1.01,0.14193984031677248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,8,1,power_law_1.01,0.15224255561828612
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,8,1,power_law_1.01,0.17304256439208984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,8,1,power_law_1.01,0.17661056518554688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,8,1,power_law_1.01,0.18315263748168947
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,8,1,power_law_1.01,0.18810623168945312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,8,1,power_law_1.01,0.1918022346496582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,8,1,power_law_1.01,0.20416448593139647
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,8,1,power_law_1.01,0.2559449577331543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,8,1,power_law_1.01,0.29110464096069333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,8,1,power_law_1.01,0.33898750305175784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,8,1,power_law_1.01,0.4053126525878906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,8,1,power_law_1.01,0.5316940689086914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,8,1,power_law_1.01,0.6635648345947265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,8,1,power_law_1.01,0.9187468719482421
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,8,1,power_law_1.01,1.178652801513672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,8,1,balanced,0.21703680038452147
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,8,1,balanced,0.24786752700805664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,8,1,power_law_1.01,0.037344000339508056
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,8,1,balanced,0.3461561584472656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,8,1,power_law_1.01,0.04526400089263916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,8,1,balanced,0.44781246185302737
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,8,1,power_law_1.01,0.05785344123840332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,8,1,power_law_1.01,0.07235712051391602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,8,1,power_law_1.01,0.09477375984191895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,8,1,power_law_1.01,1.695166778564453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,4,2,power_law_1.2,1.6121292114257812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,8,1,power_law_1.01,0.13219327926635743
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,8,1,power_law_1.01,0.1362784004211426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,8,1,power_law_1.01,0.13988544464111327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,8,1,power_law_1.01,0.15905344009399414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,8,1,power_law_1.01,2.2193145751953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,8,1,power_law_1.01,0.16078144073486328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,8,1,balanced,0.6481171417236328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,8,1,balanced,0.8513017272949218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,8,1,power_law_1.01,0.1312512016296387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,8,1,power_law_1.01,0.1681216049194336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,8,1,power_law_1.01,0.18190015792846678
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,8,1,power_law_1.01,0.19664703369140626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,8,1,power_law_1.01,0.2571622467041016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,8,1,power_law_1.01,0.3064672088623047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,8,1,power_law_1.01,0.38306049346923826
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,8,1,power_law_1.01,0.16561855316162108
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,8,1,power_law_1.01,0.16956607818603514
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,8,1,power_law_1.01,0.46633342742919925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,8,1,power_law_1.01,0.22386112213134765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,8,1,power_law_1.01,0.6385772705078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,8,1,power_law_1.01,0.8084429168701173
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,8,1,power_law_1.01,0.04243711948394775
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,8,1,power_law_1.01,0.04682688236236572
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,8,1,power_law_1.01,0.05128960132598877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,8,1,power_law_1.01,1.1457759857177734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,8,1,power_law_1.01,0.13739263534545898
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,8,1,power_law_1.01,0.08523008346557617
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,8,1,power_law_1.01,0.06773695945739747
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,8,1,power_law_1.01,0.0889907169342041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,8,1,power_law_1.01,1.4874777221679687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,8,1,power_law_1.01,0.09045248031616211
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,8,1,power_law_1.01,0.0944320011138916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,8,1,power_law_1.01,0.09836607933044433
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,8,1,power_law_1.01,0.1007852840423584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,8,1,power_law_1.01,0.10344384193420411
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,8,1,power_law_1.01,0.10983167648315431
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,8,1,power_law_1.01,0.12455552101135255
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,8,1,power_law_1.01,0.11676159858703614
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,8,1,power_law_1.01,0.12218303680419922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,8,1,power_law_1.01,0.05916031837463379
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,8,1,power_law_1.01,0.0935807991027832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,8,1,power_law_1.01,0.16134719848632811
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,8,1,power_law_1.01,0.1355628776550293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,8,1,power_law_1.01,0.19067136764526366
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,8,1,power_law_1.01,0.24963520050048826
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,8,1,power_law_1.01,0.313185920715332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,8,1,power_law_1.2,0.03452287912368775
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,8,1,power_law_1.01,0.4435756683349609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,8,1,power_law_1.2,0.05445184230804443
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,8,1,power_law_1.2,0.06454144001007081
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,8,1,power_law_1.2,0.07559232234954834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,8,1,power_law_1.2,0.08829376220703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,8,1,power_law_1.2,0.12443776130676269
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,8,1,power_law_1.2,0.13127424240112304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,8,1,power_law_1.2,0.13603391647338867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,8,1,power_law_1.2,0.14046143531799316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,8,1,power_law_1.2,0.14377663612365724
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,8,1,power_law_1.2,0.1560582447052002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,8,1,power_law_1.2,0.17512447357177735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,8,1,power_law_1.2,0.17801279067993164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,8,1,power_law_1.2,0.18439807891845703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,8,1,power_law_1.2,0.18936191558837892
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,8,1,power_law_1.2,0.19364416122436523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,8,1,power_law_1.2,0.20458240509033204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,8,1,power_law_1.2,0.2669171142578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,8,1,power_law_1.2,0.29352191925048826
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,8,1,power_law_1.2,0.34187328338623046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,8,1,power_law_1.2,0.40616382598876954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,8,1,power_law_1.2,0.5338604736328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,8,1,power_law_1.2,0.6645037078857422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,8,1,power_law_1.2,0.9231871795654296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,8,1,power_law_1.2,1.1840735626220702
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,8,1,power_law_1.2,0.03696063995361328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,8,1,power_law_1.01,0.5822272109985351
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,8,1,power_law_1.2,0.04646336078643799
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,8,1,power_law_1.01,0.8545350646972656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,8,1,power_law_1.2,0.05640448093414306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,8,1,power_law_1.2,0.08883711814880371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,8,1,power_law_1.2,0.12489343643188477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,8,1,power_law_1.2,0.12956543922424318
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,8,1,power_law_1.2,0.13316096305847167
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,8,1,power_law_1.01,1.129949417114258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,8,1,power_law_1.2,0.13639231681823732
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,8,1,power_law_1.2,0.13732416152954102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,8,1,power_law_1.2,0.13984767913818358
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,8,1,power_law_1.2,0.1602521514892578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,8,1,power_law_1.2,0.16101951599121095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,8,1,power_law_1.2,0.1670489692687988
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,8,1,power_law_1.2,0.1689190483093262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,8,1,power_law_1.2,0.17406015396118163
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,8,1,power_law_1.2,0.18198272705078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,8,1,power_law_1.2,0.20693567276000974
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,8,1,power_law_1.2,0.2278879928588867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,8,1,power_law_1.01,1.7131782531738282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,8,1,power_law_1.2,0.2650163269042969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,8,1,power_law_1.2,0.06859903812408448
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,8,1,power_law_1.2,0.31426048278808594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,8,1,power_law_1.01,2.3066278076171876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,8,1,power_law_1.2,0.3854374313354492
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,8,1,power_law_1.2,0.04265984058380127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,8,1,power_law_1.2,0.47261310577392573
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,8,1,power_law_1.2,0.04684864044189453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,8,1,power_law_1.2,0.051258239746093746
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,8,1,power_law_1.2,0.056979198455810544
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,8,1,power_law_1.2,0.06513023853302002
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,8,1,power_law_1.2,0.08636927604675293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,8,1,power_law_1.2,0.08980863571166993
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,8,1,power_law_1.2,0.09202239990234376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,8,1,power_law_1.2,0.09463104248046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,8,1,power_law_1.2,0.6419026947021484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,8,1,power_law_1.2,0.09619839668273926
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,8,1,power_law_1.2,0.1000556755065918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,8,1,power_law_1.2,0.10431872367858888
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,8,1,power_law_1.2,0.10762495994567871
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,8,1,power_law_1.2,0.11507583618164063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,8,1,power_law_1.2,0.1317568016052246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,8,1,power_law_1.2,0.8140844726562501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,8,1,power_law_1.2,1.6987930297851563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,8,1,power_law_1.2,0.12401535987854004
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,8,1,power_law_1.2,0.149302396774292
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,8,1,power_law_1.2,0.18802303314208985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,8,1,power_law_1.2,1.14797119140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,8,1,power_law_1.2,0.22951551437377932
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,8,1,power_law_1.2,0.3168409538269043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,8,1,power_law_1.2,2.2235066223144533
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,8,1,power_law_1.2,0.4087059020996094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,1,8,balanced,0.06621568202972412
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,1,8,balanced,0.07924736022949219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,1,8,balanced,0.0961087989807129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,1,8,balanced,0.16855808258056643
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,1,8,balanced,0.3245913696289063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,1,8,balanced,0.40738174438476565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,1,8,balanced,0.4051903915405274
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,8,1,power_law_1.2,0.598276481628418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,1,8,balanced,0.41150337219238275
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,1,8,balanced,0.41550464630126954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,1,8,balanced,0.42029567718505856
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,8,1,power_law_1.2,0.7908480072021484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,1,8,balanced,0.4226982498168946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,1,8,balanced,0.4210700988769531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,1,8,balanced,0.43647872924804687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,1,8,balanced,0.4473855972290039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,1,8,balanced,0.4599795150756836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,1,8,balanced,0.49241470336914067
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,1,8,balanced,0.5298112106323243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,1,8,balanced,0.44452606201171874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,1,8,balanced,0.576358413696289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,8,1,power_law_1.2,1.487368927001953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,1,8,balanced,0.8037273406982421
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,1,8,balanced,0.8779341125488281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,1,8,balanced,1.2462566375732422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,1,8,balanced,1.5426777648925782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,1,8,balanced,2.1588735961914063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,1,8,balanced,0.03835520029067993
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,1,8,balanced,0.045664000511169436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,1,8,balanced,0.060970239639282234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,8,1,power_law_1.2,1.1818425750732422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,1,8,balanced,0.10541312217712402
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,1,8,balanced,3.1574719238281252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,1,8,balanced,0.16501760482788086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,1,8,balanced,0.20576000213623047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,1,8,balanced,0.21052160263061523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,1,8,balanced,0.21634559631347655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,1,8,balanced,0.21714303970336912
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,1,8,balanced,0.22107007980346677
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,1,8,balanced,0.21101440429687499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,1,8,balanced,0.2140428733825684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,1,8,balanced,0.22111232757568358
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,1,8,balanced,0.22264448165893555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,1,8,balanced,0.22477951049804687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,1,8,balanced,0.24603904724121092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,1,8,balanced,4.848558044433593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,1,8,balanced,0.280577278137207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,1,8,balanced,0.2960588836669922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,1,8,balanced,0.21634048461914063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,1,8,balanced,0.34609279632568357
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,1,8,balanced,0.3996083068847656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,1,8,balanced,0.566805763244629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,1,8,balanced,0.6702976226806641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,1,8,balanced,0.9368563079833985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,1,8,balanced,1.2054707336425783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,1,8,balanced,6.561610107421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,1,8,balanced,1.7040281677246092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,1,8,balanced,2.230362854003906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,8,1,power_law_1.2,1.601312713623047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,8,1,power_law_1.2,2.4659884643554686
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,1,8,balanced,0.02360960006713867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,1,8,balanced,0.025588479042053226
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,1,8,balanced,0.11126015663146974
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,1,8,balanced,0.11148159980773927
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,1,8,balanced,0.11174912452697754
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,1,8,balanced,0.04240896224975586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,1,8,balanced,0.11165056228637696
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,1,8,balanced,0.11213824272155762
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,1,8,balanced,0.11236607551574707
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,1,8,balanced,0.11287296295166016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,1,8,balanced,0.11303680419921876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,1,8,balanced,0.11394047737121582
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,1,8,balanced,0.11522815704345704
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,1,8,balanced,0.0941977596282959
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,1,8,balanced,0.11631999969482423
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,1,8,balanced,0.11944576263427735
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,1,8,balanced,0.12317440032958984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,1,8,balanced,0.1286195182800293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,1,8,balanced,0.059216642379760744
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,1,8,balanced,0.11038592338562012
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,1,8,balanced,0.11120767593383789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,1,8,power_law_1.01,0.1423078441619873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,1,8,power_law_1.01,0.2759424018859863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,1,8,power_law_1.01,0.2083776092529297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,1,8,power_law_1.01,0.30477056503295896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,1,8,power_law_1.01,0.27823488235473637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,1,8,power_law_1.01,0.3713216018676758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,1,8,power_law_1.01,0.38385280609130856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,1,8,power_law_1.01,0.38132865905761715
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,1,8,balanced,0.32891006469726564
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,1,8,balanced,0.18978048324584962
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,1,8,balanced,0.17610111236572265
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,1,8,balanced,0.42292606353759765
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,1,8,balanced,0.2518515205383301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,1,8,power_law_1.01,0.3948454284667969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,1,8,power_law_1.01,0.40197376251220707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,1,8,power_law_1.01,0.412408332824707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,1,8,power_law_1.01,0.4147520065307617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,1,8,power_law_1.01,0.41883903503417963
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,1,8,power_law_1.01,0.463267822265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,1,8,power_law_1.01,0.4611328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,1,8,power_law_1.01,0.5113433456420898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,1,8,power_law_1.01,0.5453401565551758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,1,8,power_law_1.01,0.6221465682983398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,1,8,power_law_1.01,0.6985894775390624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,1,8,balanced,0.5624358367919922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,1,8,power_law_1.01,0.8585497283935547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,1,8,power_law_1.01,1.061006088256836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,1,8,power_law_1.01,0.10337792396545412
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,1,8,power_law_1.01,1.7115814208984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,1,8,power_law_1.01,0.1548300838470459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,1,8,power_law_1.01,1.8504013061523437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,1,8,power_law_1.01,0.12346624374389648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,1,8,power_law_1.01,0.14219136238098146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,1,8,power_law_1.01,0.15161600112915039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,1,8,power_law_1.01,0.17805952072143555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,1,8,power_law_1.01,0.1890764808654785
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,8,1,power_law_1.2,3.2862911987304684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,1,8,power_law_1.01,0.1863257598876953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,1,8,power_law_1.01,0.1909222412109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,1,8,power_law_1.01,0.1968435287475586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,1,8,power_law_1.01,0.21001983642578126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,1,8,power_law_1.01,0.19994367599487303
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,1,8,power_law_1.01,2.9905355834960936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,1,8,power_law_1.01,0.20789888381958005
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,1,8,power_law_1.01,3.2199346923828123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,1,8,power_law_1.01,0.2146060752868652
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,1,8,power_law_1.01,0.21762176513671877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,1,8,power_law_1.01,0.2329497528076172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,1,8,power_law_1.01,0.2657868766784668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,1,8,power_law_1.01,0.29938943862915035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,1,8,power_law_1.01,0.36652671813964843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,1,8,power_law_1.01,0.44132350921630864
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,1,8,power_law_1.01,0.5693926239013672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,1,8,power_law_1.01,0.20821632385253905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,1,8,power_law_1.01,0.7521087646484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,1,8,power_law_1.01,0.06026495933532715
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,1,8,power_law_1.01,0.07640960216522216
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,1,8,power_law_1.01,0.09175552368164062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,1,8,power_law_1.01,1.0196351623535156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,1,8,power_law_1.01,0.0910848045349121
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,1,8,power_law_1.01,0.09471487998962402
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,1,8,power_law_1.01,0.09604479789733886
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,1,8,power_law_1.01,1.2723379516601563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,1,8,power_law_1.01,5.024605407714843
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,1,8,power_law_1.01,0.09773311614990235
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,1,8,power_law_1.01,0.10100607872009278
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,1,8,power_law_1.01,0.10208255767822265
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,1,8,power_law_1.01,0.10380672454833983
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,1,8,power_law_1.01,0.10673791885375976
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,1,8,power_law_1.01,0.10815999984741212
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,1,8,power_law_1.01,1.8362380981445312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,1,8,power_law_1.01,0.11142656326293945
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,1,8,power_law_1.01,0.11049599647521972
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,1,8,power_law_1.01,0.11362048149108887
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,1,8,power_law_1.01,0.1716044807434082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,1,8,power_law_1.01,2.533749694824219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,1,8,power_law_1.01,7.154352416992188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,1,8,power_law_1.01,0.09802111625671386
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,1,8,power_law_1.01,0.13198464393615722
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,1,8,power_law_1.01,0.11986175537109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,1,8,power_law_1.01,0.14653311729431154
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,1,8,power_law_1.01,0.21176191329956057
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,1,8,power_law_1.2,0.14272512435913084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,1,8,power_law_1.2,0.1862131118774414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,1,8,power_law_1.2,0.21279743194580075
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,1,8,power_law_1.01,0.2560063934326172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,1,8,power_law_1.2,0.3058995246887207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,1,8,power_law_1.2,0.28491775512695316
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,1,8,power_law_1.01,0.3213849639892578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,1,8,power_law_1.2,0.3661606216430664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,1,8,power_law_1.2,0.38507518768310545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,1,8,power_law_1.2,0.39201919555664066
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,1,8,power_law_1.01,0.5140326309204102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,1,8,power_law_1.2,0.40050048828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,1,8,power_law_1.2,0.4142323303222656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,1,8,power_law_1.2,0.410113296508789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,1,8,power_law_1.2,0.44984191894531256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,1,8,power_law_1.2,0.48432254791259766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,1,8,power_law_1.2,0.5073664093017578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,1,8,power_law_1.2,0.5257164764404296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,1,8,power_law_1.2,0.3941721725463867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,1,8,power_law_1.2,0.5548416137695312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,1,8,power_law_1.2,0.6313919830322265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,1,8,power_law_1.2,0.7806259155273437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,1,8,power_law_1.2,0.9930048370361328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,1,8,power_law_1.2,1.2031641387939453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,1,8,power_law_1.2,1.7141746520996093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,1,8,power_law_1.2,2.161623077392578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,1,8,power_law_1.2,0.10383232116699219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,1,8,power_law_1.2,0.11971839904785156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,1,8,power_law_1.2,0.13895936012268068
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,1,8,power_law_1.2,0.1506662368774414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,1,8,power_law_1.2,0.14616959571838378
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,1,8,power_law_1.2,0.17562623977661135
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,1,8,power_law_1.2,3.4089370727539063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,1,8,power_law_1.2,0.1731520080566406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,1,8,power_law_1.2,0.19879423141479494
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,1,8,power_law_1.2,0.19393280029296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,1,8,power_law_1.2,0.19991680145263674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,1,8,power_law_1.2,0.20674047470092774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,1,8,power_law_1.2,0.20885759353637695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,1,8,power_law_1.2,0.208538875579834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,1,8,power_law_1.2,0.2166579246520996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,1,8,power_law_1.2,0.2219161605834961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,1,8,power_law_1.2,0.22379392623901367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,1,8,power_law_1.2,0.24656255722045897
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,1,8,power_law_1.2,6.26195556640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,1,8,power_law_1.2,0.28797952651977543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,1,8,power_law_1.2,0.31761152267456055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,1,8,power_law_1.2,4.223155212402344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,1,8,power_law_1.2,0.4058124923706055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,1,8,power_law_1.2,0.46496128082275395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,1,8,power_law_1.2,0.6608064270019531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,1,8,power_law_1.2,0.7923712158203126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,1,8,power_law_1.2,2.0251315307617186
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,1,8,power_law_1.2,0.06104703903198242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,1,8,power_law_1.2,1.1180697631835936
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,1,8,power_law_1.2,0.07001855850219726
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,1,8,power_law_1.2,0.08247936248779297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,1,8,power_law_1.2,0.0880627155303955
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,1,8,power_law_1.2,0.08515071868896484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,1,8,power_law_1.2,0.09209600448608399
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,1,8,power_law_1.2,0.09910400390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,1,8,power_law_1.2,0.09871999740600586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,1,8,power_law_1.2,1.565180206298828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,1,8,power_law_1.2,0.1036569595336914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,1,8,power_law_1.2,0.1061747169494629
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,1,8,power_law_1.2,0.10659968376159668
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,1,8,power_law_1.2,0.10362624168395995
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,1,8,power_law_1.2,0.10524800300598144
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,1,8,power_law_1.2,0.11763456344604492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,1,8,power_law_1.2,2.7643148803710935
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,1,8,power_law_1.2,0.11689855575561522
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,1,8,power_law_1.2,0.14745984077453614
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,1,8,power_law_1.01,0.5463040161132813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,1,8,power_law_1.01,0.9598873901367189
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,1,8,power_law_1.01,1.188884506225586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,1,8,power_law_1.2,9.922301635742187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,1,8,power_law_1.2,0.12931839942932127
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,1,8,power_law_1.2,0.11228799819946289
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,1,8,power_law_1.2,0.16355072021484376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,1,8,power_law_1.2,0.22113279342651365
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,2,4,balanced,0.061489920616149905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,2,4,balanced,0.07380864143371582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,2,4,balanced,0.09284480094909668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,2,4,balanced,0.14844672203063963
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,2,4,balanced,0.3114944076538086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,2,4,balanced,0.33882495880126956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,2,4,balanced,0.4011801528930664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,2,4,balanced,0.4040755081176758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,2,4,balanced,0.4061248016357422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,2,4,balanced,0.41079040527343746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,2,4,balanced,0.4063027191162109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,2,4,balanced,0.41109760284423824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,2,4,balanced,0.4171878433227539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,2,4,balanced,0.42504192352294917
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,2,4,balanced,0.4318041610717773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,2,4,balanced,0.44197631835937495
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,2,4,balanced,0.4581785583496094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,1,8,power_law_1.2,0.24558847427368163
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,2,4,balanced,0.49137409210205074
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,1,8,power_law_1.2,0.38371070861816403
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,2,4,balanced,0.526484489440918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,2,4,balanced,0.7450457763671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,1,8,power_law_1.2,0.40654464721679684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,2,4,balanced,0.8148236846923828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,2,4,balanced,1.1203826904296874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,2,4,balanced,1.4245683288574218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,2,4,balanced,1.8990246582031252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,2,4,balanced,0.03598207950592041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,2,4,balanced,0.04451583862304688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,2,4,balanced,2.5672460937499997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,2,4,balanced,0.06581503868103027
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,2,4,balanced,0.10924160003662109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,2,4,balanced,0.17206016540527344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,2,4,balanced,0.19692928314208985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,2,4,balanced,0.2151795196533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,2,4,balanced,0.21781887054443358
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,2,4,balanced,0.22030080795288084
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,1,8,power_law_1.2,0.6892838287353515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,2,4,balanced,0.21999616622924806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,2,4,balanced,0.21303808212280276
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,2,4,balanced,3.609940490722656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,2,4,balanced,0.21626623153686522
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,2,4,balanced,0.21445119857788084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,2,4,balanced,0.22072704315185546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,2,4,balanced,0.22132480621337888
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,2,4,balanced,0.2255628776550293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,2,4,balanced,0.24146688461303709
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,2,4,balanced,0.2715635108947754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,2,4,balanced,0.2813363265991211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,2,4,balanced,0.3288652801513672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,2,4,balanced,0.3836006546020508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,2,4,balanced,0.5307072067260743
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,2,4,balanced,0.6191872024536134
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,2,4,balanced,0.8837337493896484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,2,4,balanced,5.068815307617188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,2,4,balanced,1.1186713409423827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,2,4,balanced,1.556689910888672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,2,4,balanced,2.040264892578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,1,8,power_law_1.2,0.8635263824462891
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,1,8,power_law_1.2,1.270851821899414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,1,8,power_law_1.2,1.6868109130859374
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,2,4,balanced,0.025241599082946774
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,2,4,balanced,0.11640447616577147
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,2,4,balanced,0.11717375755310058
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,2,4,balanced,0.11715071678161622
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,2,4,balanced,0.04320511817932129
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,2,4,balanced,0.028577280044555665
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,2,4,balanced,0.11733504295349122
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,2,4,balanced,0.11667072296142578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,2,4,balanced,0.11781503677368163
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,2,4,balanced,0.062136321067810064
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,2,4,balanced,0.12011391639709472
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,2,4,balanced,0.11986175537109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,2,4,balanced,0.12197631835937499
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,2,4,balanced,0.1255116844177246
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,2,4,balanced,0.12958847999572753
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,2,4,balanced,0.13643648147583007
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,2,4,balanced,0.14323072433471679
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,2,4,balanced,0.09815039634704589
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,2,4,balanced,0.11552512168884277
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,2,4,balanced,0.11541888236999512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,2,4,power_law_1.01,0.09342975616455078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,2,4,power_law_1.01,0.132860164642334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,2,4,power_law_1.01,0.14041728019714356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,2,4,power_law_1.01,0.20252544403076173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,2,4,power_law_1.01,0.23629568099975584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,2,4,power_law_1.01,0.31703296661376956
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,2,4,balanced,0.1150476837158203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,2,4,power_law_1.01,0.3775987243652344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,2,4,power_law_1.01,0.37550720214843747
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,2,4,power_law_1.01,0.3919500732421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,2,4,power_law_1.01,0.39302143096923825
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,2,4,power_law_1.01,0.39529598236083985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,2,4,power_law_1.01,0.39870849609375003
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,2,4,power_law_1.01,0.40554878234863284
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,2,4,power_law_1.01,0.4139212799072266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,2,4,power_law_1.01,0.42113792419433593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,2,4,power_law_1.01,0.4541120147705078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,2,4,power_law_1.01,0.5155545425415039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,2,4,power_law_1.01,0.5651468658447266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,2,4,power_law_1.01,0.6365708923339843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,2,4,power_law_1.01,0.8030860900878907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,2,4,power_law_1.01,0.9226367950439454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,2,4,power_law_1.01,1.2102642822265626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,2,4,power_law_1.01,1.6363737487792966
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,2,4,power_law_1.01,2.329228820800781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,2,4,balanced,0.2002726364135742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,2,4,balanced,0.21585664749145508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,2,4,power_law_1.01,2.8911065673828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,2,4,balanced,0.2958297538757324
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,2,4,balanced,0.5026188659667968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,2,4,power_law_1.01,0.08871808052062988
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,2,4,balanced,0.38749439239501954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,2,4,power_law_1.01,0.088919038772583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,2,4,power_law_1.01,0.12348031997680664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,2,4,power_law_1.01,0.17106943130493163
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,2,4,power_law_1.01,0.20058368682861327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,2,4,power_law_1.01,0.20461568832397461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,2,4,power_law_1.01,0.0627507209777832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,2,4,power_law_1.01,0.2059584045410156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,2,4,power_law_1.01,0.20898048400878907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,2,4,power_law_1.01,0.21264896392822266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,2,4,power_law_1.01,0.20364160537719728
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,2,4,power_law_1.01,0.2078835105895996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,2,4,power_law_1.01,0.2112396812438965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,2,4,power_law_1.01,4.202560119628906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,2,4,power_law_1.01,0.2131328010559082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,2,4,power_law_1.01,0.21694591522216794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,2,4,power_law_1.01,0.13657343864440918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,2,4,power_law_1.01,0.22893823623657225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,2,4,power_law_1.01,0.27718143463134765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,2,4,power_law_1.01,0.2998233604431152
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,2,4,power_law_1.01,0.357760009765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,2,4,power_law_1.01,0.4135910415649414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,2,4,power_law_1.01,0.5535987091064454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,2,4,power_law_1.01,0.6762380981445313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,2,4,power_law_1.01,0.045920000076293946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,2,4,power_law_1.01,0.9506534576416016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,2,4,power_law_1.01,0.06238719940185546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,2,4,power_law_1.01,0.058152961730957034
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,2,4,power_law_1.01,1.2001254272460937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,2,4,power_law_1.01,0.06942207813262939
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,2,4,power_law_1.01,0.08094592094421386
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,2,4,power_law_1.01,0.10603263854980469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,2,4,power_law_1.01,0.1038105583190918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,2,4,power_law_1.01,0.10209152221679688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,2,4,power_law_1.01,0.10766464233398439
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,2,4,power_law_1.01,0.10868864059448242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,2,4,power_law_1.01,0.10845696449279785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,2,4,power_law_1.01,1.720780792236328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,2,4,power_law_1.01,0.1113599967956543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,2,4,power_law_1.01,0.12239999771118164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,2,4,power_law_1.01,0.12348928451538085
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,2,4,power_law_1.01,0.12181504249572754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,2,4,power_law_1.01,2.1981132507324217
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,2,4,power_law_1.01,0.193621768951416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,2,4,power_law_1.01,5.722269287109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,2,4,power_law_1.01,0.11290240287780762
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,2,4,balanced,0.682968292236328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,2,4,power_law_1.01,0.13476863861083985
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,2,4,power_law_1.01,0.23491455078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,2,4,power_law_1.01,0.16507904052734373
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,2,4,power_law_1.01,0.1465702438354492
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,2,4,power_law_1.01,0.3112396812438965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,2,4,power_law_1.2,0.09414655685424804
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,2,4,power_law_1.01,0.3907852935791015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,2,4,power_law_1.2,0.1113036823272705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,2,4,power_law_1.2,0.1430732822418213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,2,4,power_law_1.2,0.18215936660766602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,2,4,power_law_1.2,0.2435327911376953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,2,4,power_law_1.2,0.325082893371582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,2,4,power_law_1.2,0.3785932922363281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,2,4,power_law_1.2,0.38325374603271484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,2,4,power_law_1.01,0.5870374298095703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,2,4,power_law_1.2,0.39211776733398435
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,2,4,power_law_1.2,0.39432449340820314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,2,4,power_law_1.2,0.4017804718017578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,2,4,power_law_1.2,0.39580032348632815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,2,4,power_law_1.2,0.41410049438476565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,2,4,power_law_1.2,0.42313598632812505
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,2,4,power_law_1.2,0.4283366394042969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,2,4,power_law_1.2,0.4619609451293945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,2,4,power_law_1.2,0.48279296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,2,4,power_law_1.2,0.5735884857177734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,2,4,power_law_1.2,0.6522509002685547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,2,4,power_law_1.2,0.8420198059082031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,2,4,power_law_1.2,1.0374822235107422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,2,4,power_law_1.2,1.3182655334472657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,2,4,power_law_1.2,1.731173095703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,2,4,power_law_1.2,0.06312191963195801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,2,4,power_law_1.2,0.07794303894042968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,2,4,power_law_1.2,0.09724032402038574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,2,4,power_law_1.2,2.558659820556641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,2,4,power_law_1.2,0.11939328193664552
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,2,4,power_law_1.2,0.1313049602508545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,2,4,power_law_1.2,0.17021055221557618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,2,4,power_law_1.2,0.20006528854370118
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,2,4,power_law_1.01,0.7448461151123047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,2,4,power_law_1.2,0.20446592330932617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,2,4,power_law_1.2,0.2074649620056152
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,2,4,power_law_1.2,0.21050495147705078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,2,4,power_law_1.2,0.21036928176879882
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,2,4,power_law_1.2,0.20805631637573244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,2,4,power_law_1.2,0.20827903747558593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,2,4,power_law_1.2,0.21648895263671877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,2,4,power_law_1.2,0.2172979164123535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,2,4,power_law_1.2,0.22124671936035156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,2,4,power_law_1.2,0.2424038314819336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,2,4,power_law_1.2,0.2877363204956055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,2,4,power_law_1.2,0.3065356826782227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,2,4,power_law_1.2,0.37644161224365236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,2,4,power_law_1.2,0.4452454376220703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,2,4,power_law_1.2,3.1178765869140626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,2,4,power_law_1.2,0.5821913528442383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,2,4,power_law_1.2,0.7279974365234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,2,4,power_law_1.2,1.0090380859374999
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,2,4,power_law_1.2,0.045401601791381835
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,2,4,power_law_1.2,0.05081984043121338
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,2,4,power_law_1.2,0.058849282264709465
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,2,4,power_law_1.2,0.07181951999664307
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,2,4,power_law_1.2,0.07833983898162841
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,2,4,power_law_1.2,1.816204833984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,2,4,power_law_1.2,4.804261169433594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,2,4,power_law_1.2,6.065623168945312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,2,4,power_law_1.2,0.10578816413879395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,2,4,power_law_1.2,0.10421504020690917
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,2,4,power_law_1.2,0.10738304138183594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,2,4,power_law_1.2,0.11024383544921874
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,2,4,power_law_1.2,0.10741503715515137
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,2,4,power_law_1.2,0.11141887664794922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,2,4,power_law_1.2,0.11479680061340332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,2,4,power_law_1.2,0.115600643157959
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,2,4,power_law_1.2,0.115731201171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,2,4,power_law_1.2,1.2326348876953124
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,2,4,power_law_1.2,0.1377356815338135
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,2,4,power_law_1.2,2.3537049865722657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,2,4,power_law_1.2,0.30497791290283205
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,2,4,power_law_1.01,1.0970995330810547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,2,4,power_law_1.01,1.4128460693359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,2,4,power_law_1.2,0.1229759979248047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,2,4,power_law_1.2,0.12983551979064942
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,2,4,power_law_1.2,0.1827084732055664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,2,4,power_law_1.2,0.16370559692382813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,2,4,power_law_1.2,0.2369049644470215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,4,2,balanced,0.05798143863677978
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,4,2,balanced,0.07007872104644776
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,4,2,balanced,0.09102335929870606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,4,2,balanced,0.14674176216125487
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,4,2,balanced,0.31106815338134763
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,4,2,balanced,0.33205886840820314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,4,2,balanced,0.3350822448730469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,4,2,balanced,0.3364044952392578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,4,2,balanced,0.33869056701660155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,4,2,balanced,0.39097854614257815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,4,2,balanced,0.39426815032958984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,4,2,balanced,0.399153938293457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,4,2,balanced,0.4021260833740234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,4,2,balanced,0.40792064666748046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,4,2,balanced,0.412677116394043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,4,2,balanced,0.419415054321289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,4,2,balanced,0.4367987060546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,4,2,balanced,0.47039104461669917
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,4,2,balanced,0.5072025680541993
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,4,2,balanced,0.7076863861083984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,4,2,balanced,0.7665510559082032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,4,2,balanced,1.0573990631103516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,4,2,balanced,1.3949235534667968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,4,2,balanced,1.7997325134277342
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,2,4,power_law_1.2,0.4001971054077148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,4,2,balanced,2.4962777709960937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,4,2,balanced,0.03947391986846924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,4,2,balanced,0.048648958206176755
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,4,2,balanced,0.0676262378692627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,4,2,balanced,0.11755647659301757
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,4,2,balanced,0.18317951202392577
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,4,2,balanced,0.20909952163696288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,4,2,balanced,3.4953381347656247
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,4,2,balanced,0.21776895523071288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,4,2,balanced,0.2156697654724121
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,2,4,power_law_1.2,0.7267135620117188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,4,2,balanced,0.23498111724853515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,4,2,balanced,0.21240320205688476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,4,2,balanced,4.440776977539063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,4,2,balanced,0.23581823348999023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,4,2,balanced,0.23867776870727536
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,4,2,balanced,0.2408511924743652
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,4,2,balanced,0.24549503326416017
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,4,2,balanced,0.244714241027832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,2,4,power_law_1.2,0.5183795166015626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,4,2,balanced,0.24788095474243166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,4,2,balanced,0.2586611175537109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,4,2,balanced,0.28548608779907225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,4,2,balanced,0.30510976791381833
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,4,2,balanced,0.331644172668457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,4,2,balanced,0.3799948883056641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,4,2,balanced,0.554444808959961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,4,2,balanced,0.6308889770507813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,4,2,balanced,0.8991040039062501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,4,2,balanced,1.1451058959960938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,4,2,balanced,1.5605824279785157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,4,2,balanced,2.066967010498047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,2,4,power_law_1.2,1.0535667419433594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,2,4,power_law_1.2,1.5400192260742187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,2,4,power_law_1.2,2.0819827270507814
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,4,2,balanced,0.0325708794593811
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,4,2,balanced,0.03585920095443725
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,4,2,balanced,0.13260416030883787
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,4,2,balanced,0.13240320205688477
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,4,2,balanced,0.13312255859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,4,2,balanced,0.1349952030181885
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,4,2,balanced,0.13442432403564453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,4,2,balanced,0.1377843189239502
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,4,2,balanced,0.13827967643737793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,4,2,balanced,0.14027903556823731
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,4,2,balanced,0.1436953639984131
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,4,2,balanced,0.14878080368041993
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,4,2,balanced,0.15500800132751463
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,4,2,balanced,0.0721446418762207
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,4,2,balanced,0.16595199584960937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,4,2,balanced,0.17757183074951172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,4,2,balanced,0.051459841728210456
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,4,2,balanced,0.1120025634765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,4,2,balanced,0.13191424369812013
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,4,2,balanced,0.13209728240966795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,4,2,power_law_1.01,0.07132031917572021
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,4,2,power_law_1.01,0.08807295799255371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,4,2,power_law_1.01,0.10141440391540528
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,4,2,power_law_1.01,0.14864128112792968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,4,2,power_law_1.01,0.19469823837280273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,4,2,power_law_1.01,0.3083724784851074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,4,2,power_law_1.01,0.31614591598510744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,4,2,power_law_1.01,0.3210611343383789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,4,2,power_law_1.01,0.31943296432495116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,4,2,power_law_1.01,0.36582145690917967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,4,2,power_law_1.01,0.37321086883544924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,4,2,power_law_1.01,0.38530174255371097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,4,2,power_law_1.01,0.39215103149414066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,4,2,power_law_1.01,0.39361793518066407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,4,2,power_law_1.01,0.3972518539428711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,4,2,power_law_1.01,0.4190822219848632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,4,2,power_law_1.01,0.47120128631591796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,4,2,power_law_1.01,0.5324288177490234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,4,2,power_law_1.01,0.5818713760375976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,4,2,power_law_1.01,0.7267660522460938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,4,2,power_law_1.01,0.8507405090332032
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,4,2,balanced,0.13610879898071288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,4,2,power_law_1.01,1.0992742156982422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,4,2,balanced,0.28783359527587893
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,4,2,balanced,0.2608563232421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,4,2,balanced,0.41136001586914067
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,4,2,power_law_1.01,1.356793670654297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,4,2,power_law_1.01,0.04528639793395996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,4,2,power_law_1.01,1.998954315185547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,4,2,balanced,0.5302758407592774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,4,2,power_law_1.01,0.06336895942687988
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,4,2,power_law_1.01,0.07296127796173095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,4,2,power_law_1.01,0.11502079963684082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,4,2,power_law_1.01,0.1351679992675781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,4,2,power_law_1.01,0.1846118354797363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,4,2,power_law_1.01,0.19152383804321288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,4,2,power_law_1.01,0.19364992141723633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,4,2,power_law_1.01,0.19627904891967773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,4,2,power_law_1.01,0.21956096649169923
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,4,2,power_law_1.01,0.22524032592773438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,4,2,power_law_1.01,2.5589068603515623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,4,2,power_law_1.01,0.22775680541992188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,4,2,power_law_1.01,0.23196800231933593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,4,2,power_law_1.01,0.23500799179077148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,4,2,power_law_1.01,0.23648000717163087
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,4,2,power_law_1.01,0.23966335296630858
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,4,2,balanced,0.7120563507080078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,4,2,power_law_1.01,0.25145984649658204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,4,2,power_law_1.01,0.28199808120727543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,4,2,power_law_1.01,0.32517505645751954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,4,2,power_law_1.01,0.36590721130371096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,4,2,power_law_1.01,0.43698944091796876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,4,2,power_law_1.01,3.5582656860351562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,4,2,power_law_1.01,0.5568793487548829
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,4,2,power_law_1.01,0.6831385803222656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,4,2,power_law_1.01,0.03608448028564453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,4,2,power_law_1.01,0.938256607055664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,4,2,power_law_1.01,0.05441152095794678
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,4,2,power_law_1.01,0.05844736099243164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,4,2,power_law_1.01,0.06999680042266845
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,4,2,power_law_1.01,0.08506367683410646
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,4,2,power_law_1.01,1.1751974487304688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,4,2,power_law_1.01,0.11812864303588869
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,4,2,power_law_1.01,0.12018815994262695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,4,2,power_law_1.01,0.12240639686584473
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,4,2,power_law_1.01,0.119683837890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,4,2,power_law_1.01,0.12502143859863282
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,4,2,power_law_1.01,0.13077759742736816
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,4,2,power_law_1.01,0.13430656433105467
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,4,2,power_law_1.01,0.13483263969421386
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,4,2,power_law_1.01,1.6571212768554688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,4,2,power_law_1.01,0.13847807884216307
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,4,2,power_law_1.01,0.14401791572570802
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,4,2,power_law_1.01,0.20967679977416992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,4,2,power_law_1.01,2.1102796936035153
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,4,2,power_law_1.01,4.704335327148438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,4,2,balanced,0.9701465606689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,4,2,power_law_1.01,0.16016767501831056
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,4,2,power_law_1.01,0.26556159973144533
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,4,2,power_law_1.01,0.15042943954467775
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,4,2,power_law_1.01,0.17880319595336913
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,4,2,power_law_1.01,0.3178073692321777
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,4,2,power_law_1.01,0.6046591949462891
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,4,2,power_law_1.01,0.43477630615234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,4,2,power_law_1.2,0.07115520000457763
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,4,2,power_law_1.2,0.08055935859680176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,4,2,power_law_1.2,0.10830080032348632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,4,2,power_law_1.2,0.13590271949768068
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,4,2,power_law_1.2,0.16394367218017578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,4,2,power_law_1.2,0.3030284881591797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,4,2,power_law_1.2,0.31729663848876954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,4,2,power_law_1.2,0.3215974426269531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,4,2,power_law_1.2,0.3214438247680664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,4,2,power_law_1.2,0.369378547668457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,4,2,power_law_1.2,0.37399166107177734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,4,2,power_law_1.2,0.38199424743652344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,4,2,power_law_1.2,0.3914700698852539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,4,2,power_law_1.2,0.4080217742919922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,4,2,power_law_1.2,0.41719936370849614
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,4,2,power_law_1.2,0.45613567352294926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,4,2,power_law_1.2,0.5022655868530274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,4,2,power_law_1.2,0.5419033432006837
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,4,2,power_law_1.2,0.5866854476928711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,4,2,power_law_1.2,0.7286643218994141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,4,2,power_law_1.2,0.8636736297607422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,4,2,power_law_1.2,1.1111795043945312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,4,2,power_law_1.2,2.0615615844726562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,4,2,power_law_1.2,1.5165708923339845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,4,2,power_law_1.2,0.046465277671813965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,4,2,power_law_1.2,0.06051583766937256
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,4,2,power_law_1.01,0.779004135131836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,4,2,power_law_1.2,0.07604991912841796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,4,2,power_law_1.2,0.10873727798461914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,4,2,power_law_1.2,0.13179903984069824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,4,2,power_law_1.2,0.19375616073608398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,4,2,power_law_1.2,0.19081600189208986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,4,2,power_law_1.2,0.19271936416625976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,4,2,power_law_1.2,0.1994483184814453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,4,2,power_law_1.2,0.22255487442016603
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,4,2,power_law_1.2,2.555988464355469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,4,2,power_law_1.2,3.7764901733398437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,4,2,power_law_1.2,0.2261631965637207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,4,2,power_law_1.2,0.22671232223510743
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,4,2,power_law_1.2,0.2329497528076172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,4,2,power_law_1.2,0.2422617530822754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,4,2,power_law_1.2,0.2425139236450195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,4,2,power_law_1.2,0.24558719635009765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,4,2,power_law_1.2,0.26553728103637697
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,4,2,power_law_1.2,0.2989401626586914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,4,2,power_law_1.2,0.32902271270751954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,4,2,power_law_1.2,0.37256065368652347
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,4,2,power_law_1.2,0.4492633438110351
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,4,2,power_law_1.2,0.5706329727172852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,4,2,power_law_1.2,0.7016294097900391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,4,2,power_law_1.2,0.9673177337646484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,4,2,power_law_1.2,0.03634047985076904
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,4,2,power_law_1.2,0.04548480033874512
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,4,2,power_law_1.2,0.05906047821044922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,4,2,power_law_1.2,1.6979327392578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,4,2,power_law_1.2,0.07057536125183106
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,4,2,power_law_1.2,0.07918208122253419
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,4,2,power_law_1.2,0.11615360260009766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,4,2,power_law_1.2,0.11780991554260253
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,4,2,power_law_1.2,0.12207232475280763
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,4,2,power_law_1.2,0.12606719970703123
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,4,2,power_law_1.2,0.12685183525085447
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,4,2,power_law_1.2,0.13404288291931152
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,4,2,power_law_1.2,0.13311871528625488
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,4,2,power_law_1.2,0.1364748764038086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,4,2,power_law_1.2,0.14524031639099122
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,4,2,power_law_1.2,0.15379199981689454
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,4,2,power_law_1.2,0.16801023483276367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,4,2,power_law_1.2,1.191171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,4,2,power_law_1.2,4.94181884765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,4,2,power_law_1.2,2.2460287475585936
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,4,2,power_law_1.01,0.9962534332275391
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,4,2,power_law_1.01,1.506429443359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,4,2,power_law_1.01,1.987791290283203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,4,2,power_law_1.2,0.16124160766601564
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,4,2,power_law_1.2,0.20885759353637695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,4,2,power_law_1.2,0.2376153564453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,4,2,power_law_1.2,0.32089599609374997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,8,1,balanced,0.06996479988098145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,8,1,balanced,0.08364992141723633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,8,1,balanced,0.10561792373657226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,8,1,balanced,0.15947839736938477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,8,1,balanced,0.32478527069091795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,8,1,balanced,0.3559833526611328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,8,1,balanced,0.3562150573730469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,8,1,balanced,0.3577344131469727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,8,1,balanced,0.35919742584228515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,8,1,balanced,0.36178688049316404
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,8,1,balanced,0.3657145690917969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,8,1,balanced,0.3692038345336914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,8,1,balanced,0.4276665496826172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,8,1,balanced,0.43427391052246095
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,4,2,power_law_1.2,0.41491329193115234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,8,1,balanced,0.43834625244140624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,8,1,balanced,0.4451353454589844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,8,1,balanced,0.4598400115966797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,8,1,balanced,0.49032638549804686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,8,1,balanced,0.5242195129394531
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,4,2,power_law_1.2,0.6022208023071289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,8,1,balanced,0.731945571899414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,8,1,balanced,0.7831878662109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,4,2,power_law_1.2,0.7407193756103515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,8,1,balanced,1.0941235351562502
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,8,1,balanced,1.3664544677734374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,8,1,balanced,1.8196006774902345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,8,1,balanced,0.04880320072174073
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,8,1,balanced,0.06129087924957275
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,8,1,balanced,0.08853440284729004
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,4,2,power_law_1.2,1.1220953369140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,8,1,balanced,0.14166272163391114
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,8,1,balanced,2.3934783935546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,8,1,balanced,0.23205696105957033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,8,1,balanced,0.2713030433654785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,8,1,balanced,0.27725055694580075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,8,1,balanced,0.27494848251342774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,8,1,balanced,0.27634687423706056
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,8,1,balanced,0.27907648086547854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,8,1,balanced,0.2821542358398438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,8,1,balanced,0.2843071937561035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,8,1,balanced,0.3420684814453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,8,1,balanced,3.5386163330078126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,8,1,balanced,0.3455686569213867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,8,1,balanced,0.3469152069091797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,8,1,balanced,0.34932350158691405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,8,1,balanced,0.3679558563232422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,8,1,balanced,0.3765542221069336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,8,1,balanced,0.39294464111328126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,8,1,balanced,0.42965953826904296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,8,1,balanced,0.46687423706054687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,8,1,balanced,0.7043840026855468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,8,1,balanced,0.7824877166748048
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,8,1,balanced,4.590858154296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,8,1,balanced,1.1033657836914064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,8,1,balanced,1.4265216064453123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,8,1,balanced,1.9164965820312498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,8,1,balanced,2.557145538330078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,4,2,power_law_1.2,1.4216281127929686
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,4,2,power_law_1.2,2.3707968139648434
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,8,1,balanced,0.052078080177307126
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,8,1,balanced,0.047196159362792964
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,8,1,balanced,0.06779327869415283
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,8,1,balanced,0.17091520309448244
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,8,1,balanced,0.17066240310668945
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,8,1,balanced,0.1711065673828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,8,1,balanced,0.17332799911499025
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,8,1,balanced,0.17369279861450196
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,8,1,balanced,0.09490112304687501
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,8,1,balanced,0.17980607986450195
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,8,1,balanced,0.17542272567749023
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,8,1,balanced,0.1797587203979492
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,8,1,balanced,0.14418304443359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,8,1,balanced,0.18402816772460937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,8,1,balanced,0.19012928009033203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,8,1,balanced,0.2001510429382324
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,8,1,balanced,0.2107788848876953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,8,1,balanced,0.23241727828979492
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,8,1,balanced,0.25481216430664066
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,4,2,power_law_1.2,3.0251507568359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,8,1,balanced,0.16863040924072265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,8,1,power_law_1.01,0.06767680168151854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,8,1,power_law_1.01,0.08324095726013184
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,8,1,power_law_1.01,0.1053171157836914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,8,1,power_law_1.01,0.14734016418457033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,8,1,power_law_1.01,0.19435712814331055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,8,1,power_law_1.01,0.33746368408203126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,8,1,power_law_1.01,0.33640064239501954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,8,1,power_law_1.01,0.3422476959228516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,8,1,power_law_1.01,0.34580734252929685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,8,1,power_law_1.01,0.34617984771728516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,8,1,power_law_1.01,0.3560992050170898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,8,1,power_law_1.01,0.3834880065917969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,8,1,balanced,0.16906112670898438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,8,1,power_law_1.01,0.4153689575195313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,8,1,power_law_1.01,0.42940544128417973
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,8,1,power_law_1.01,0.4395398330688477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,8,1,power_law_1.01,0.4566214370727539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,8,1,power_law_1.01,0.5225964736938477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,8,1,power_law_1.01,0.5630611038208008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,8,1,power_law_1.01,0.5992313766479492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,8,1,power_law_1.01,0.7057222747802735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,8,1,power_law_1.01,0.8515436553955078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,8,1,power_law_1.01,1.0810975646972656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,8,1,power_law_1.01,1.3579379272460936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,8,1,power_law_1.01,1.892772521972656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,8,1,power_law_1.01,2.4858656311035157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,8,1,power_law_1.01,3.5047915649414065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,8,1,power_law_1.01,4.621479187011719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,8,1,power_law_1.01,0.06339647769927978
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,8,1,balanced,0.43769088745117185
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,8,1,power_law_1.01,0.04630784034729004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,8,1,power_law_1.01,0.11551487922668457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,8,1,power_law_1.01,0.16515647888183593
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,8,1,balanced,0.6311155319213867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,8,1,balanced,0.8520365142822266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,8,1,power_law_1.01,0.244703369140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,8,1,power_law_1.01,0.23534847259521485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,8,1,power_law_1.01,0.2558534431457519
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,8,1,power_law_1.01,0.08691712379455566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,8,1,power_law_1.01,0.26629823684692383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,8,1,power_law_1.01,0.2600806427001953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,8,1,power_law_1.01,0.267128963470459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,8,1,power_law_1.01,0.27203840255737305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,8,1,power_law_1.01,0.3176691246032715
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,8,1,balanced,1.153148193359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,8,1,power_law_1.01,0.33314624786376956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,8,1,power_law_1.01,0.3350143814086914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,8,1,power_law_1.01,0.3377395248413086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,8,1,power_law_1.01,0.35582145690917966
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,8,1,balanced,0.3885273742675781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,8,1,balanced,1.5487794494628906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,8,1,power_law_1.01,0.38520576477050783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,8,1,power_law_1.01,0.40691646575927737
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,8,1,power_law_1.01,0.4756499099731445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,8,1,power_law_1.01,0.5525990295410156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,8,1,power_law_1.01,0.6932876586914063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,8,1,power_law_1.01,0.8500441741943359
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,8,1,power_law_1.01,0.047607679367065434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,8,1,power_law_1.01,1.125156478881836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,8,1,power_law_1.01,0.05504831790924072
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,8,1,power_law_1.01,0.06898687839508057
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,8,1,power_law_1.01,1.3975244140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,8,1,power_law_1.01,0.08424575805664063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,8,1,power_law_1.01,0.10179519653320312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,8,1,power_law_1.01,0.15421952247619627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,8,1,power_law_1.01,1.9616461181640623
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,8,1,power_law_1.01,0.1483955192565918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,8,1,power_law_1.01,0.1579423999786377
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,8,1,power_law_1.01,0.16142208099365235
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,8,1,power_law_1.01,0.16717632293701173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,8,1,power_law_1.01,2.529929656982422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,8,1,power_law_1.01,0.17321535110473635
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,8,1,power_law_1.01,0.176428165435791
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,8,1,power_law_1.01,0.18913856506347657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,8,1,power_law_1.01,0.20304639816284178
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,8,1,power_law_1.01,0.27053504943847656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,8,1,power_law_1.01,0.31257152557373047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,8,1,power_law_1.01,0.19676031112670897
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,8,1,power_law_1.01,0.14173055648803712
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,8,1,power_law_1.01,0.22632511138916014
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,8,1,power_law_1.01,0.5067750549316405
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,8,1,power_law_1.01,0.39959743499755856
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,8,1,power_law_1.01,0.698884506225586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,8,1,power_law_1.2,0.06813888072967529
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,8,1,power_law_1.2,0.08338175773620606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,8,1,power_law_1.2,0.1032857608795166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,8,1,power_law_1.2,0.13511039733886718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,8,1,power_law_1.2,0.17703615188598634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,8,1,power_law_1.2,0.32417663574218747
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,8,1,power_law_1.2,0.33553985595703123
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,8,1,power_law_1.01,0.9122361755371093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,8,1,power_law_1.2,0.3416659164428711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,8,1,power_law_1.2,0.34677055358886716
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,8,1,power_law_1.01,1.3367910766601563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,8,1,power_law_1.2,0.3508691024780274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,8,1,power_law_1.2,0.3660543823242187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,8,1,power_law_1.2,0.39766590118408207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,8,1,power_law_1.2,0.4217958450317383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,8,1,power_law_1.2,0.4665932846069336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,8,1,power_law_1.2,0.43475776672363275
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,8,1,power_law_1.2,0.4801728057861328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,8,1,power_law_1.2,0.5294457626342773
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,8,1,power_law_1.01,1.7752000427246095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,8,1,power_law_1.2,0.5638431930541993
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,8,1,power_law_1.2,0.602336654663086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,8,1,power_law_1.2,0.7150099182128906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,8,1,power_law_1.2,0.8383840179443359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,8,1,power_law_1.2,1.1175552368164063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,8,1,power_law_1.2,0.04769216060638427
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,8,1,power_law_1.2,1.3626675415039062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,8,1,power_law_1.2,0.062255358695983885
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,8,1,power_law_1.2,0.08522303581237793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,8,1,power_law_1.2,0.10730496406555176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,8,1,power_law_1.2,1.8839987182617186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,8,1,power_law_1.2,0.1602809524536133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,8,1,power_law_1.2,0.23390464782714843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,8,1,power_law_1.2,0.2530905532836914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,8,1,power_law_1.2,0.2586521530151367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,8,1,power_law_1.2,0.2635763168334961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,8,1,power_law_1.2,0.27049215316772457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,8,1,power_law_1.2,2.4239852905273436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,8,1,power_law_1.2,0.2466329574584961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,8,1,power_law_1.2,0.2747551918029785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,8,1,power_law_1.2,0.3349017715454102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,8,1,power_law_1.2,0.34392318725585935
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,8,1,power_law_1.2,0.34241600036621095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,8,1,power_law_1.2,0.35261825561523436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,8,1,power_law_1.2,0.36724033355712893
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,8,1,power_law_1.2,0.38650878906249997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,8,1,power_law_1.2,0.41611583709716793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,8,1,power_law_1.2,0.4876972961425781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,8,1,power_law_1.2,0.5648960113525391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,8,1,power_law_1.2,0.7034489440917968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,8,1,power_law_1.2,0.8364979553222657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,8,1,power_law_1.2,1.1338591766357422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,8,1,power_law_1.2,0.04764416217803955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,8,1,power_law_1.2,1.406815948486328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,8,1,power_law_1.2,0.05504960060119629
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,8,1,power_law_1.2,0.06730815887451172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,8,1,power_law_1.2,3.4421652221679686
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,8,1,power_law_1.2,0.07953663825988769
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,8,1,power_law_1.2,0.09799039840698243
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,8,1,power_law_1.2,0.14101823806762695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,8,1,power_law_1.2,0.148405122756958
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,8,1,power_law_1.2,0.15374079704284668
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,8,1,power_law_1.2,0.16094655990600587
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,8,1,power_law_1.2,0.16548927307128908
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,8,1,power_law_1.2,0.1723027229309082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,8,1,power_law_1.2,0.17894016265869142
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,8,1,power_law_1.2,0.1862835121154785
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,8,1,power_law_1.2,0.19700735092163085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,8,1,power_law_1.2,1.9728134155273438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,8,1,power_law_1.2,4.523429260253907
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,8,1,power_law_1.01,2.696477355957031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,8,1,power_law_1.01,3.614433898925781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,8,1,power_law_1.2,2.5669683837890624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,8,1,power_law_1.2,0.22569536209106444
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,8,1,power_law_1.2,0.21386112213134764
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,8,1,power_law_1.2,0.3116748809814453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,8,1,power_law_1.2,0.25143999099731446
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,8,1,power_law_1.2,0.37484672546386716
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,8,1,power_law_1.2,0.5152243041992188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,8,balanced,0.07107071876525879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,8,balanced,0.08081791877746583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,8,balanced,0.09828224182128906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,8,balanced,0.14762368202209472
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,8,balanced,0.2993523216247559
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,8,balanced,0.5236300659179688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,8,1,power_law_1.2,0.6622994995117187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,8,balanced,0.7437350463867187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,8,balanced,0.7945702362060547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,8,1,power_law_1.2,0.9867743682861327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,8,balanced,0.8083980560302735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,8,balanced,0.8225279998779296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,8,balanced,0.8059238433837891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,8,balanced,0.8174336242675782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,8,balanced,0.8292953491210937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,8,balanced,0.8393497467041016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,8,balanced,0.8564339447021485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,8,balanced,0.8620275115966797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,8,balanced,0.8719039916992187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,8,balanced,0.9313062286376953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,8,balanced,0.9633331298828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,8,balanced,1.0371737670898438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,8,balanced,1.1271372985839843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,8,balanced,1.298297576904297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,8,balanced,1.8010214233398436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,8,balanced,2.141099548339844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,8,balanced,2.9933068847656252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,8,balanced,0.042315521240234376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,8,balanced,0.04946176052093506
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,8,balanced,0.06809216022491456
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,8,balanced,0.09791744232177735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,8,balanced,0.1646937561035156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,8,balanced,0.28694528579711914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,8,balanced,4.450681457519531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,8,balanced,0.39327999114990236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,8,balanced,0.4187417602539062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,8,balanced,0.422716178894043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,8,balanced,0.4230080032348633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,8,balanced,0.4160563278198242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,8,balanced,0.4194060897827148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,8,balanced,0.42263935089111326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,8,balanced,0.42534015655517576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,8,balanced,0.42649982452392576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,8,balanced,6.677084350585938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,8,balanced,0.4391743850708008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,8,balanced,0.45416961669921874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,8,balanced,0.48645759582519527
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,8,balanced,0.5033075332641601
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,8,balanced,0.5621132659912109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,8,balanced,0.6036761474609376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,8,balanced,0.7074470520019531
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,8,1,power_law_1.2,1.3277011108398438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,8,balanced,0.8189900970458984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,8,balanced,0.9955891418457032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,8,balanced,1.35164794921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,8,balanced,1.7436006164550784
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,8,1,power_law_1.2,2.0591961669921877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,8,balanced,2.3312332153320314
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,8,1,power_law_1.2,2.800074157714844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,8,1,power_law_1.2,4.3360421752929685
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,8,1,power_law_1.2,5.925161743164063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,1,8,balanced,0.027167999744415285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,1,8,balanced,0.2248179244995117
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,1,8,balanced,0.02878335952758789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,1,8,balanced,0.2253900718688965
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,1,8,balanced,0.22589183807373048
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,1,8,balanced,0.22598783493041993
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,1,8,balanced,0.22570880889892578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,1,8,balanced,0.03774336099624634
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,1,8,balanced,0.2284326362609863
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,1,8,balanced,0.22711936950683595
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,1,8,balanced,0.22881919860839844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,1,8,balanced,0.09063679695129394
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,1,8,balanced,0.22951423645019534
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,1,8,balanced,0.23247232437133789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,1,8,balanced,0.23449344635009767
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,1,8,balanced,0.05745791912078857
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,1,8,balanced,0.23991039276123044
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,1,8,balanced,0.25358207702636715
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,1,8,balanced,0.2632473564147949
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,1,8,balanced,0.15542655944824219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,1,8,balanced,0.28043647766113283
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,8,power_law_1.01,0.1472921562194824
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,1,8,balanced,0.24359296798706054
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,8,power_law_1.01,0.29746047973632816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,8,power_law_1.01,0.521475830078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,8,power_law_1.01,0.4500454330444336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,8,power_law_1.01,0.5235993576049804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,8,power_law_1.01,0.630255355834961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,8,power_law_1.01,0.6670515441894531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,8,power_law_1.01,0.7289625549316406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,8,power_law_1.01,0.7432371520996094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,8,power_law_1.01,0.7523699188232422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,8,power_law_1.01,0.7163980865478516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,8,power_law_1.01,0.7326412963867187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,8,power_law_1.01,0.7462425231933594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,8,power_law_1.01,0.7273471832275391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,8,power_law_1.01,0.7766822052001954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,8,power_law_1.01,0.8075852966308593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,8,power_law_1.01,0.935815658569336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,8,power_law_1.01,0.7625651550292969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,8,power_law_1.01,0.962060775756836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,8,power_law_1.01,1.0877286529541015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,8,power_law_1.01,1.2906930541992188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,1,8,balanced,0.22565504074096682
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,8,power_law_1.01,1.7009266662597657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,8,power_law_1.01,1.936414794921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,8,power_law_1.01,2.639434204101563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,8,power_law_1.01,0.0887168025970459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,8,power_law_1.01,0.15019007682800295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,8,power_law_1.01,3.3370739746093747
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,8,power_law_1.01,0.2654425621032715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,8,power_law_1.01,0.23983488082885743
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,8,power_law_1.01,0.28672767639160157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,8,power_law_1.01,0.34571903228759765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,8,power_law_1.01,0.33898750305175784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,8,power_law_1.01,4.935498352050781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,8,power_law_1.01,0.36403968811035153
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,8,power_law_1.01,0.3930099105834961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,8,power_law_1.01,0.38636417388916017
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,8,power_law_1.01,0.34544384002685546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,8,power_law_1.01,0.34782848358154295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,8,power_law_1.01,0.35620353698730467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,8,power_law_1.01,0.3743231964111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,8,power_law_1.01,0.38687103271484374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,8,power_law_1.01,0.37885696411132813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,8,power_law_1.01,0.40181632995605465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,8,power_law_1.01,0.4294707107543945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,8,power_law_1.01,6.521699829101562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,8,power_law_1.01,0.4603750228881836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,8,power_law_1.01,0.5222284698486328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,8,power_law_1.01,0.5746700668334961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,8,power_law_1.01,0.701239013671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,8,power_law_1.01,0.8149504089355469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,8,power_law_1.01,1.0723980712890626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,8,power_law_1.01,1.403864288330078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,1,8,power_law_1.01,0.06047103881835938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,1,8,power_law_1.01,0.0949452781677246
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,1,8,power_law_1.01,0.15728256225585938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,8,power_law_1.01,1.8807040405273436
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,1,8,power_law_1.01,0.13248000144958497
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,1,8,power_law_1.01,0.16193920135498047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,1,8,power_law_1.01,0.20089855194091796
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,1,8,power_law_1.01,0.19121408462524414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,1,8,power_law_1.01,0.1906982421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,1,8,power_law_1.01,0.1848921585083008
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,1,8,power_law_1.01,0.18347135543823243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,8,power_law_1.01,2.5103552246093748
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,1,8,power_law_1.01,0.18886783599853516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,1,8,power_law_1.01,0.18667648315429688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,1,8,power_law_1.01,0.20305919647216797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,1,8,power_law_1.01,0.18774015426635743
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,1,8,power_law_1.01,0.194202880859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,1,8,power_law_1.01,0.19601024627685545
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,1,8,power_law_1.01,0.20050048828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,1,8,power_law_1.01,0.2126412773132324
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,1,8,balanced,0.2493324851989746
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,1,8,balanced,0.3963353729248047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,1,8,balanced,0.5923225784301758
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,1,8,balanced,0.4391334533691406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,1,8,power_law_1.01,0.28053760528564453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,1,8,power_law_1.01,0.23051263809204103
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,1,8,power_law_1.01,0.2532569694519043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,8,power_law_1.2,0.14624256134033203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,1,8,power_law_1.01,0.3392793655395508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,8,power_law_1.2,0.2967475128173828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,8,power_law_1.2,0.3774016189575195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,8,power_law_1.2,0.4734489440917969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,8,power_law_1.2,0.5761075210571289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,8,power_law_1.2,0.637015037536621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,8,power_law_1.2,0.7330048370361328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,8,power_law_1.2,0.7046963500976562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,8,power_law_1.2,0.7326783752441406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,8,power_law_1.2,0.7208332824707031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,8,power_law_1.2,0.6910297393798828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,1,8,power_law_1.01,0.3832191848754883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,8,power_law_1.2,0.7110284423828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,8,power_law_1.2,0.7164710235595704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,8,power_law_1.2,0.7407769775390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,8,power_law_1.2,0.7755827331542969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,8,power_law_1.2,0.7602444458007812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,8,power_law_1.2,0.8173363494873047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,8,power_law_1.2,0.9461170959472657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,8,power_law_1.2,0.9897907257080079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,8,power_law_1.2,1.1580416107177736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,8,power_law_1.2,1.3959231567382813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,8,power_law_1.2,1.767713317871094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,8,power_law_1.2,2.2852902221679687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,1,8,power_law_1.01,0.6954316711425781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,8,power_law_1.2,0.08860287666320801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,8,power_law_1.2,0.15009920120239256
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,1,8,power_law_1.01,0.5284646224975587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,8,power_law_1.2,2.9589132690429687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,8,power_law_1.2,0.19079551696777344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,1,8,power_law_1.01,0.978476791381836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,8,power_law_1.2,0.23402496337890627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,8,power_law_1.2,0.3028979110717773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,8,power_law_1.2,0.36020606994628906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,8,power_law_1.2,0.2739737510681152
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,8,power_law_1.2,0.3731955337524414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,8,power_law_1.2,0.312936954498291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,8,power_law_1.2,0.3269926452636719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,8,power_law_1.2,0.3360230255126953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,8,power_law_1.2,0.3459417724609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,8,power_law_1.2,0.34302719116210934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,8,power_law_1.2,0.3692915344238281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,8,power_law_1.2,0.3858969497680664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,8,power_law_1.2,0.3577382278442383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,8,power_law_1.2,0.39494014739990235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,8,power_law_1.2,3.8356198120117186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,8,power_law_1.2,0.43611774444580076
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,8,power_law_1.2,0.4718783950805664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,8,power_law_1.2,0.5219980621337891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,8,power_law_1.2,0.6009830474853516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,8,power_law_1.2,0.7427686309814453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,8,power_law_1.2,5.63219482421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,8,power_law_1.2,0.9102925109863282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,8,power_law_1.2,1.2044620513916016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,1,8,power_law_1.2,0.060648961067199705
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,1,8,power_law_1.2,0.09493760108947755
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,1,8,power_law_1.2,0.11495936393737793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,1,8,power_law_1.2,0.13369855880737305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,1,8,power_law_1.2,0.16880895614624022
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,8,power_law_1.2,1.4776191711425781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,1,8,power_law_1.2,0.20817407608032226
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,1,8,power_law_1.2,0.19868160247802735
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,1,8,power_law_1.2,0.18275583267211915
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,1,8,power_law_1.2,0.1656230354309082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,8,power_law_1.2,2.0719757080078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,1,8,power_law_1.2,0.17815807342529297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,1,8,power_law_1.2,0.17533056259155272
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,1,8,power_law_1.2,0.18069120407104494
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,8,power_law_1.2,7.476375122070313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,1,8,power_law_1.2,0.20050432205200197
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,1,8,power_law_1.2,0.1767616081237793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,1,8,power_law_1.2,0.18067712783813478
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,1,8,power_law_1.2,0.21550464630126953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,8,power_law_1.2,2.6297586059570315
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,1,8,power_law_1.2,0.2378611183166504
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,1,8,power_law_1.2,0.33180545806884765
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,1,8,power_law_1.2,0.18920703887939455
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,1,8,power_law_1.01,1.1782144165039061
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,1,8,power_law_1.2,0.4225791931152344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,1,8,power_law_1.2,0.24567935943603514
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,4,balanced,0.06275199890136719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,1,8,power_law_1.2,0.28245504379272457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,4,balanced,0.07534848213195801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,4,balanced,0.09688447952270508
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,1,8,power_law_1.2,0.5157158279418945
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,1,8,power_law_1.2,0.6610483551025391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,4,balanced,0.14414848327636717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,4,balanced,0.2957401657104492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,4,balanced,0.5285478210449218
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,1,8,power_law_1.2,0.843909149169922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,4,balanced,0.739552001953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,4,balanced,0.7470246124267578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,4,balanced,0.7433574676513672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,4,balanced,0.7345331573486329
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,4,balanced,0.8038118743896485
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,1,8,power_law_1.2,1.3210917663574218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,4,balanced,0.8031053161621093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,4,balanced,0.8034687805175782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,4,balanced,0.8159257507324218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,4,balanced,0.8248588562011718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,4,balanced,0.8468966674804687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,4,balanced,0.8683776092529296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,4,balanced,0.88724609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,4,balanced,0.9210086059570312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,4,balanced,0.9955187225341797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,4,balanced,1.075683822631836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,4,balanced,1.2457279968261719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,4,balanced,0.039129600524902344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,4,balanced,0.04612224102020264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,4,balanced,0.0654963207244873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,4,balanced,1.725566711425781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,4,balanced,0.10439423561096191
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,4,balanced,0.16915071487426758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,4,balanced,0.29508735656738283
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,4,balanced,2.0168984985351566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,4,balanced,0.40031871795654295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,4,balanced,0.4021555328369141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,4,balanced,0.40731647491455075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,4,balanced,2.7044454956054684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,4,balanced,0.4433881759643555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,4,balanced,0.43991039276123045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,4,balanced,0.44567169189453126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,4,balanced,0.45018241882324217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,4,balanced,0.45173889160156244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,4,balanced,3.731969299316406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,4,balanced,0.45843967437744143
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,4,balanced,0.4732748794555664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,4,balanced,0.5048486328124999
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,4,balanced,0.5668991851806641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,4,balanced,0.6155430221557617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,4,balanced,0.4072934341430664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,4,balanced,0.6861273956298828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,4,balanced,0.7886694335937501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,4,balanced,5.580149536132812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,4,balanced,0.5175551986694336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,4,balanced,0.9707878112792969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,4,balanced,1.3062413024902344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,4,balanced,1.7126541137695312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,4,balanced,2.219696655273437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,1,8,power_law_1.2,1.5869503784179688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,2,4,balanced,0.0339084792137146
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,2,4,balanced,0.037747199535369876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,2,4,balanced,0.045121278762817386
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,2,4,balanced,0.24074495315551755
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,2,4,balanced,0.0635481595993042
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,2,4,balanced,0.2401702308654785
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,2,4,balanced,0.2411686325073242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,2,4,balanced,0.24147712707519534
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,2,4,balanced,0.09901439666748046
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,2,4,balanced,0.24132863998413087
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,2,4,balanced,0.24310911178588865
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,2,4,balanced,0.1668070411682129
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,2,4,balanced,0.24419200897216795
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,2,4,balanced,0.24398591995239255
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,2,4,balanced,0.24730752944946288
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,2,4,balanced,0.2500556755065918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,2,4,balanced,0.25525760650634766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,2,4,balanced,0.26138624191284177
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,2,4,balanced,0.26945407867431637
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,2,4,balanced,0.282480640411377
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,2,4,balanced,0.2969011116027832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,2,4,balanced,0.32283519744873046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,4,power_law_1.01,0.09655424118041991
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,4,power_law_1.01,0.14003071784973145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,4,power_law_1.01,0.2905907249450684
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,2,4,balanced,0.23990144729614257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,4,power_law_1.01,0.29377792358398436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,4,power_law_1.01,0.3823846435546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,4,power_law_1.01,0.48080513000488284
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,4,power_law_1.01,0.48473087310791013
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,4,power_law_1.01,0.592567024230957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,4,power_law_1.01,0.6131929779052735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,4,power_law_1.01,0.6032166290283203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,4,power_law_1.01,0.6778803253173828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,4,power_law_1.01,0.6890137481689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,4,power_law_1.01,0.6904422760009765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,4,power_law_1.01,0.7039347076416016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,4,power_law_1.01,0.7260211181640626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,4,power_law_1.01,0.7132556915283204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,4,power_law_1.01,0.7544793701171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,4,power_law_1.01,0.8384563446044921
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,4,power_law_1.01,0.9307366180419923
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,4,power_law_1.01,1.0300531005859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,4,power_law_1.01,1.1846336364746093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,4,power_law_1.01,1.4832179260253906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,2,4,balanced,0.23999359130859377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,4,power_law_1.01,1.7719602966308592
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,4,power_law_1.01,0.06192512035369872
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,4,power_law_1.01,0.09010175704956055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,4,power_law_1.01,2.5417459106445315
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,4,power_law_1.01,0.1505907154083252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,4,power_law_1.01,0.16599296569824218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,4,power_law_1.01,3.0180645751953126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,4,power_law_1.01,0.22007808685302735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,4,power_law_1.01,0.27896703720092775
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,4,power_law_1.01,0.2780659294128418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,4,power_law_1.01,0.34233985900878905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,4,power_law_1.01,0.3221516799926758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,4,power_law_1.01,0.31460992813110356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,4,power_law_1.01,0.365431022644043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,4,power_law_1.01,0.3847859191894531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,4,power_law_1.01,0.36385665893554686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,4,power_law_1.01,0.36838401794433595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,4,power_law_1.01,0.3807308959960937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,4,power_law_1.01,0.40022014617919927
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,4,power_law_1.01,0.41554817199707034
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,4,power_law_1.01,0.45574401855468755
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,4,power_law_1.01,4.24927734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,4,power_law_1.01,0.4679679870605469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,4,power_law_1.01,5.648167724609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,4,power_law_1.01,0.508716812133789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,4,power_law_1.01,0.5711219024658203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,4,power_law_1.01,0.7003865814208984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,4,power_law_1.01,0.8349375915527343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,4,power_law_1.01,1.106836471557617
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,2,4,power_law_1.01,0.0509887981414795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,4,power_law_1.01,1.3209805297851562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,2,4,power_law_1.01,0.06876160144805908
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,2,4,power_law_1.01,0.09788031578063965
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,2,4,power_law_1.01,0.1050649642944336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,4,power_law_1.01,1.8247232055664064
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,2,4,power_law_1.01,0.12820096015930177
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,2,4,power_law_1.01,0.1628607940673828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,2,4,power_law_1.01,0.1678054428100586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,4,power_law_1.01,2.3019340515136717
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,2,4,power_law_1.01,0.199102725982666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,2,4,power_law_1.01,0.1913920021057129
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,2,4,power_law_1.01,0.19957887649536132
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,2,4,power_law_1.01,0.21123327255249022
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,2,4,power_law_1.01,0.21353343963623045
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,2,4,power_law_1.01,0.21416448593139648
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,2,4,power_law_1.01,0.22336896896362304
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,2,4,power_law_1.01,0.22214399337768556
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,2,4,power_law_1.01,0.21384191513061523
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,2,4,power_law_1.01,0.22361215591430664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,2,4,power_law_1.01,0.23467775344848635
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,2,4,balanced,0.5278771209716797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,2,4,balanced,0.45968257904052734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,2,4,balanced,0.7301977539062501
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,2,4,power_law_1.01,0.26553216934204105
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,2,4,power_law_1.01,0.4921868896484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,2,4,power_law_1.01,0.38418689727783206
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,2,4,power_law_1.01,0.29596160888671874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,4,power_law_1.2,0.09610495567321778
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,4,power_law_1.2,0.13988608360290528
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,4,power_law_1.2,0.20798080444335937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,4,power_law_1.2,0.2925388717651367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,4,power_law_1.2,0.37675006866455074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,4,power_law_1.2,0.4653260803222656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,4,power_law_1.2,0.5202048110961914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,4,power_law_1.2,0.5702092742919922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,4,power_law_1.2,0.6055744171142579
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,2,4,power_law_1.01,0.31769855499267574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,4,power_law_1.2,0.5877670288085938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,4,power_law_1.2,0.6610368347167969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,4,power_law_1.2,0.6704307556152344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,4,power_law_1.2,0.6989478302001954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,4,power_law_1.2,0.7000601959228516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,4,power_law_1.2,0.7391219329833985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,4,power_law_1.2,0.7354700469970703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,4,power_law_1.2,0.7659340667724609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,4,power_law_1.2,0.89889404296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,4,power_law_1.2,0.9517938995361328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,4,power_law_1.2,1.0778482818603516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,4,power_law_1.2,1.2414796447753906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,4,power_law_1.2,1.5990489196777342
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,4,power_law_1.2,1.900042266845703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,4,power_law_1.2,2.6493402099609376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,4,power_law_1.2,0.06235648155212402
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,4,power_law_1.2,0.08941311836242676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,4,power_law_1.2,3.25928955078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,4,power_law_1.2,0.1263155174255371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,4,power_law_1.2,0.15952639579772948
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,4,power_law_1.2,0.21139711380004883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,4,power_law_1.2,0.26977279663085935
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,4,power_law_1.2,0.26938240051269535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,4,power_law_1.2,0.34045310974121096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,4,power_law_1.2,4.949502868652344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,4,power_law_1.2,0.3493721771240234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,2,4,power_law_1.01,0.6593228912353515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,4,power_law_1.2,0.32484222412109376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,4,power_law_1.2,0.35106689453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,4,power_law_1.2,0.3460006332397461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,4,power_law_1.2,0.36522750854492186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,4,power_law_1.2,0.38824192047119144
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,2,4,power_law_1.01,0.872574691772461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,4,power_law_1.2,0.376258544921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,4,power_law_1.2,5.87907958984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,4,power_law_1.2,0.4028467178344727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,4,power_law_1.2,0.43763584136962896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,4,power_law_1.2,0.41974655151367185
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,2,4,power_law_1.01,1.1147942352294922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,4,power_law_1.2,0.4643686294555664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,4,power_law_1.2,0.553752326965332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,4,power_law_1.2,0.6147289657592774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,4,power_law_1.2,0.7251634979248047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,2,4,power_law_1.2,0.04969344139099121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,4,power_law_1.2,0.8750374603271485
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,2,4,power_law_1.2,0.06883584022521973
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,4,power_law_1.2,1.1442662048339844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,2,4,power_law_1.2,0.08846336364746094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,2,4,power_law_1.2,0.10388352394104003
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,2,4,power_law_1.2,0.12839679718017577
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,4,power_law_1.2,1.4393164062500001
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,2,4,power_law_1.2,0.16357759475708006
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,2,4,power_law_1.2,0.18686080932617188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,2,4,power_law_1.2,0.18804992675781249
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,2,4,power_law_1.2,0.19143552780151368
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,4,power_law_1.2,1.9138021850585936
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,2,4,power_law_1.2,0.19831935882568358
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,2,4,power_law_1.2,0.19912704467773437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,2,4,power_law_1.2,0.19876480102539062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,2,4,power_law_1.2,0.22344064712524414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,2,4,power_law_1.2,0.21513343811035157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,2,4,power_law_1.2,0.2225587272644043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,2,4,power_law_1.2,0.2313817596435547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,2,4,power_law_1.2,0.15812479972839355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,4,power_law_1.2,2.512806396484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,2,4,power_law_1.01,1.5262451171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,2,4,power_law_1.2,0.2553868865966797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,2,4,power_law_1.2,0.29026687622070313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,2,4,power_law_1.2,0.34298240661621093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,4,2,balanced,0.06100480079650879
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,2,4,power_law_1.2,0.6615026855468751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,2,balanced,0.07492608070373535
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,2,4,power_law_1.2,0.4713727951049805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,2,balanced,0.09385087966918945
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,2,4,power_law_1.2,0.3862041473388672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,2,balanced,0.13960063934326172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,2,balanced,0.29755903244018556
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,2,balanced,0.5276710510253906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,2,balanced,0.7399910736083984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,2,balanced,0.7432972717285156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,2,balanced,0.7355264282226562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,2,balanced,0.7489228820800781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,2,balanced,0.7467750549316406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,2,balanced,0.7529958343505859
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,2,4,power_law_1.2,0.890218276977539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,2,balanced,0.7415513610839843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,2,balanced,0.8201932525634765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,2,balanced,0.8270988464355469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,2,balanced,0.8235507202148437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,2,balanced,0.8412915039062501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,2,balanced,0.8787955474853515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,2,balanced,0.9167040252685548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,2,balanced,0.99059326171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,2,balanced,1.0579814147949218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,2,balanced,1.2614131164550781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,2,balanced,1.7186842346191404
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,2,balanced,0.04643968105316162
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,2,balanced,1.9949452209472658
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,2,balanced,0.048660478591918944
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,2,balanced,0.07147136211395264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,2,balanced,0.10560895919799804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,2,balanced,0.18777727127075194
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,2,balanced,0.3308902359008789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,2,balanced,2.7334130859375003
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,2,balanced,0.457182731628418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,2,balanced,0.45957504272460936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,2,balanced,0.4644825744628906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,2,balanced,3.6617803955078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,2,balanced,0.46361728668212887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,2,balanced,0.4667020797729492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,2,balanced,0.4710092926025391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,2,balanced,0.48194049835205083
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,2,balanced,0.5286796951293946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,2,balanced,0.5308083343505859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,2,balanced,0.5383385467529297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,2,balanced,0.5436710357666016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,2,balanced,0.5635763168334961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,2,balanced,0.5916569519042969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,2,balanced,0.623034896850586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,2,balanced,0.6715596771240234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,2,balanced,5.625484619140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,2,balanced,0.7500940704345703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,2,balanced,0.8321971130371093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,2,balanced,0.9906969451904297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,2,balanced,1.3851033020019532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,2,balanced,1.721943054199219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,2,balanced,2.2987571716308595
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,2,4,power_law_1.2,1.1415692901611327
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,2,4,power_law_1.2,1.8939903259277344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,2,4,power_law_1.2,1.9729959106445314
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,4,2,balanced,0.05003136157989502
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,4,2,balanced,0.05414783954620361
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,4,2,balanced,0.06203775882720948
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,4,2,balanced,0.08246527671813965
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,4,2,balanced,0.28970239639282225
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,4,2,balanced,0.2909440040588379
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,4,2,balanced,0.29347455978393555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,4,2,balanced,0.2938163185119629
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,4,2,balanced,0.2942720031738281
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,4,2,balanced,0.12123647689819336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,4,2,balanced,0.2959334373474121
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,4,2,balanced,0.29751039505004884
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,4,2,balanced,0.2902387237548828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,4,2,balanced,0.3006028747558594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,4,2,balanced,0.30836351394653316
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,4,2,balanced,0.31375871658325194
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,4,2,balanced,0.32853504180908205
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,4,2,balanced,0.3396006393432617
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,4,2,balanced,0.3659699249267578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,4,2,balanced,0.19933696746826174
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,4,2,balanced,0.39514625549316407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,4,2,balanced,0.4445913696289063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,4,2,power_law_1.01,0.07499392032623291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,2,power_law_1.01,0.0956595230102539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,2,power_law_1.01,0.13870592117309571
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,2,power_law_1.01,0.1823846435546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,2,power_law_1.01,0.3358975982666016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,2,power_law_1.01,0.39074047088623043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,2,power_law_1.01,0.44703231811523436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,2,power_law_1.01,0.5506867218017578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,2,power_law_1.01,0.5647449493408203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,2,power_law_1.01,0.5711872100830078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,2,power_law_1.01,0.5945817565917968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,2,power_law_1.01,0.609062385559082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,2,power_law_1.01,0.615731201171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,2,power_law_1.01,0.7195609283447266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,2,power_law_1.01,0.734576644897461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,2,power_law_1.01,0.7379750061035156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,2,power_law_1.01,0.7777190399169922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,2,power_law_1.01,0.8327295684814453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,2,power_law_1.01,0.9147801971435546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,2,power_law_1.01,1.0561945343017578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,2,power_law_1.01,1.1569318389892578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,2,power_law_1.01,1.4513177490234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,2,power_law_1.01,1.7593650817871094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,2,power_law_1.01,2.2879321289062498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,2,power_law_1.01,2.8776397705078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,2,power_law_1.01,0.046407680511474605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,2,power_law_1.01,0.06688000202178955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,2,power_law_1.01,0.09751680374145508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,2,power_law_1.01,0.14247167587280274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,2,power_law_1.01,4.264257202148437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,2,power_law_1.01,0.2035443115234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,2,power_law_1.01,0.2674227142333984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,2,power_law_1.01,0.29502847671508786
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,2,power_law_1.01,0.3578854370117187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,2,power_law_1.01,0.3531558227539063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,2,power_law_1.01,5.079132080078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,2,power_law_1.01,0.3686822509765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,2,power_law_1.01,0.37861248016357424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,2,power_law_1.01,0.3996134567260742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,2,power_law_1.01,0.39606529235839844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,2,power_law_1.01,0.45065471649169925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,2,power_law_1.01,0.46026752471923826
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,4,2,balanced,0.2913420867919922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,2,power_law_1.01,0.46536960601806643
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,2,power_law_1.01,0.4783526229858398
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,4,2,balanced,0.29144319534301755
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,2,power_law_1.01,0.4883430480957031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,2,power_law_1.01,0.5340480041503907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,2,power_law_1.01,0.5842035293579102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,2,power_law_1.01,0.6468569946289062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,2,power_law_1.01,0.7823526763916016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,2,power_law_1.01,0.8870963287353515
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,4,2,power_law_1.01,0.05353600025177002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,2,power_law_1.01,1.153311996459961
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,4,2,power_law_1.01,0.06892672061920166
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,4,2,power_law_1.01,0.0892147159576416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,2,power_law_1.01,1.409233856201172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,4,2,power_law_1.01,0.09546879768371583
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,4,2,power_law_1.01,0.12385279655456542
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,4,2,power_law_1.01,0.1591001605987549
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,4,2,power_law_1.01,0.1636582374572754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,2,power_law_1.01,1.8893414306640626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,4,2,power_law_1.01,0.21166463851928713
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,4,2,power_law_1.01,0.21288192749023435
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,4,2,power_law_1.01,0.22000640869140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,4,2,power_law_1.01,0.23286144256591798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,2,power_law_1.01,2.3936358642578126
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,4,2,power_law_1.01,0.24064256668090822
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,4,2,power_law_1.01,0.23699712753295898
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,4,2,power_law_1.01,0.24941951751708985
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,4,2,power_law_1.01,0.26220544815063473
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,4,2,power_law_1.01,0.27208703994750977
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,4,2,power_law_1.01,0.2735193634033203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,4,2,power_law_1.01,0.3128268814086914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,4,2,balanced,0.6145087814331054
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,4,2,balanced,0.7430502319335938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,4,2,balanced,1.024490203857422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,4,2,power_law_1.01,0.3399641418457031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,4,2,power_law_1.01,0.38127742767333983
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,4,2,power_law_1.2,0.0743398380279541
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,4,2,power_law_1.01,0.4349504089355468
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,4,2,power_law_1.01,0.6947929382324218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,2,power_law_1.2,0.09421695709228516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,4,2,power_law_1.01,0.5603827285766602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,2,power_law_1.2,0.12331775665283203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,2,power_law_1.2,0.17985664367675783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,2,power_law_1.2,0.31979263305664063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,2,power_law_1.2,0.3751180648803711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,2,power_law_1.2,0.4097958374023437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,2,power_law_1.2,0.5434918212890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,2,power_law_1.2,0.556864013671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,2,power_law_1.2,0.5727564620971679
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,2,power_law_1.2,0.6015423965454102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,2,power_law_1.2,0.6159552001953126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,2,power_law_1.2,0.6232012939453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,2,power_law_1.2,0.7061414337158203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,2,power_law_1.2,0.7235826873779297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,2,power_law_1.2,0.7406758117675781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,2,power_law_1.2,0.7752076721191405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,2,power_law_1.2,0.8577958679199218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,2,power_law_1.2,0.9304000091552733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,2,power_law_1.2,1.060094757080078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,2,power_law_1.2,1.1961100769042967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,2,power_law_1.2,1.5083264160156251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,2,power_law_1.2,1.7885848999023437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,2,power_law_1.2,2.3727935791015624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,4,2,power_law_1.01,0.8515213012695313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,4,2,power_law_1.01,1.1770687866210938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,2,power_law_1.2,0.046012158393859866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,2,power_law_1.2,3.03080078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,4,2,power_law_1.01,1.636607971191406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,2,power_law_1.2,0.06655360221862792
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,2,power_law_1.2,0.08544896125793458
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,2,power_law_1.2,0.1403123188018799
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,2,power_law_1.2,0.18932224273681642
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,2,power_law_1.2,0.25663232803344727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,2,power_law_1.2,0.27719680786132816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,2,power_law_1.2,0.34832000732421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,2,power_law_1.2,0.36908798217773436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,2,power_law_1.2,0.3609868621826172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,2,power_law_1.2,0.3772364807128906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,2,power_law_1.2,0.38690814971923826
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,2,power_law_1.2,0.4007526397705078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,2,power_law_1.2,4.539466247558594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,2,power_law_1.2,0.43192577362060547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,2,power_law_1.2,0.4385011291503906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,2,power_law_1.2,0.46449535369873046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,2,power_law_1.2,0.4612531280517578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,2,power_law_1.2,0.4942374420166016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,2,power_law_1.2,0.5466444778442383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,2,power_law_1.2,0.6009932708740234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,2,power_law_1.2,5.213212280273437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,2,power_law_1.2,0.6639103698730469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,2,power_law_1.2,0.8069631958007812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,4,2,power_law_1.2,0.05476096153259278
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,2,power_law_1.2,0.9164915466308594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,4,2,power_law_1.2,0.06911488056182861
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,4,2,power_law_1.2,0.08167936325073241
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,2,power_law_1.2,1.2089241790771486
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,4,2,power_law_1.2,0.09855232238769532
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,4,2,power_law_1.2,0.11672703742980957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,2,power_law_1.2,1.4636889648437499
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,4,2,power_law_1.2,0.14548607826232912
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,4,2,power_law_1.2,0.16554624557495118
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,4,2,power_law_1.2,0.2000294494628906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,4,2,power_law_1.2,0.20968448638916018
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,4,2,power_law_1.2,0.22045183181762695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,4,2,power_law_1.2,0.226560001373291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,2,power_law_1.2,1.9144639587402346
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,4,2,power_law_1.2,0.23633279800415039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,4,2,power_law_1.2,0.23347583770751953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,4,2,power_law_1.2,0.2491865539550781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,4,2,power_law_1.2,0.25954687118530273
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,4,2,power_law_1.2,0.27222528457641604
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,4,2,power_law_1.2,0.28115840911865236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,2,power_law_1.2,2.513514251708984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,4,2,power_law_1.2,0.3633907318115234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,4,2,power_law_1.2,0.5520358276367188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,4,2,power_law_1.01,2.149049530029297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,4,2,power_law_1.2,0.3369638442993164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,8,1,balanced,0.07244287967681885
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,8,1,balanced,0.08692416191101074
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,4,2,power_law_1.2,0.4342246246337891
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,4,2,power_law_1.2,0.7448038482666016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,8,1,balanced,0.10857919692993163
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,8,1,balanced,0.15553791999816896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,8,1,balanced,0.3153171157836914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,4,2,power_law_1.2,0.9341222381591796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,8,1,balanced,0.5620671844482421
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,4,2,power_law_1.2,1.2574297332763673
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,8,1,balanced,0.7865074920654297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,4,2,power_law_1.2,1.6920614624023436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,8,1,balanced,0.7880953979492188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,8,1,balanced,0.7827763366699219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,8,1,balanced,0.793207015991211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,8,1,balanced,0.7946905517578126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,8,1,balanced,0.7960985565185548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,8,1,balanced,0.8099533081054687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,8,1,balanced,0.8020114898681641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,8,1,balanced,0.8164441680908203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,8,1,balanced,0.8101107025146485
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,4,2,power_law_1.2,2.5366490173339846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,8,1,balanced,0.9358470153808593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,8,1,balanced,0.9739116668701172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,8,1,balanced,0.992899169921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,8,1,balanced,1.0560492706298827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,8,1,balanced,1.1350150299072266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,8,1,balanced,1.3003584289550782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,8,1,balanced,0.05480063915252685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,8,1,balanced,0.06632832050323487
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,8,1,balanced,0.08936575889587403
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,8,1,balanced,0.13111231803894044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,8,1,balanced,0.22669439315795897
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,8,1,balanced,1.8186770629882811
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,8,1,balanced,0.4182579040527344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,8,1,balanced,2.083660125732422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,8,1,balanced,0.6108998489379883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,8,1,balanced,0.6163584136962891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,8,1,balanced,0.6134624099731445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,8,1,balanced,0.6194527816772462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,8,1,balanced,2.8336993408203126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,8,1,balanced,0.6224102401733399
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,8,1,balanced,0.6259577560424805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,8,1,balanced,0.6312044906616211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,8,1,balanced,0.6378828811645507
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,8,1,balanced,0.6500460815429687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,8,1,balanced,0.6519308471679688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,8,1,balanced,0.7331033325195312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,8,1,balanced,3.737353515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,8,1,balanced,0.7504473876953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,8,1,balanced,0.7679449462890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,8,1,balanced,0.8099129486083985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,8,1,balanced,0.8485715484619141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,8,1,balanced,0.9287417602539062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,8,1,balanced,1.0061478424072265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,8,1,balanced,1.1710905456542968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,8,1,balanced,5.762896118164063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,8,1,balanced,1.7223539733886717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,8,1,balanced,2.068053741455078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,8,1,balanced,2.790943908691406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,4,2,power_law_1.2,3.3322329711914067
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,8,1,balanced,0.07879936218261718
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,8,1,balanced,0.07898816108703613
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,8,1,balanced,0.08074624061584472
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,8,1,balanced,0.07989376068115235
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,8,1,balanced,0.11242112159729004
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,8,1,balanced,0.08454143524169921
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,8,1,balanced,0.11942399978637694
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,8,1,balanced,0.09149375915527344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,8,1,balanced,0.13564096450805666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,8,1,balanced,0.09867199897766113
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,8,1,balanced,0.10616703987121583
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,8,1,balanced,0.18388992309570312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,8,1,balanced,0.20579776763916016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,8,1,balanced,0.2506438446044922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,8,1,balanced,0.43502529144287105
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,8,1,balanced,0.34319873809814455
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,8,1,balanced,0.6213843154907226
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,8,1,balanced,0.30053312301635743
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,8,1,balanced,0.7929440307617187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,8,1,balanced,1.159442596435547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,8,1,balanced,1.534020538330078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,8,1,power_law_1.01,0.07017727851867675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,8,1,power_law_1.01,0.08674431800842285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,8,1,power_law_1.01,0.10903103828430176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,8,1,power_law_1.01,0.15555839538574218
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,8,1,balanced,2.270615692138672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,8,1,power_law_1.01,0.2906937599182129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,8,1,power_law_1.01,0.40094848632812496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,8,1,power_law_1.01,0.43423809051513673
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,8,1,power_law_1.01,0.5635404968261719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,8,1,balanced,2.9899859619140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,8,1,power_law_1.01,0.5847660827636718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,8,1,power_law_1.01,0.5938278579711914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,8,1,power_law_1.01,0.6128806304931641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,8,1,power_law_1.01,0.6425068664550782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,8,1,power_law_1.01,0.6170662307739259
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,8,1,power_law_1.01,0.692008285522461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,8,1,power_law_1.01,0.7293977355957031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,8,1,balanced,4.4375341796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,8,1,power_law_1.01,0.7778361511230469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,8,1,power_law_1.01,0.8627193450927735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,8,1,power_law_1.01,0.9224678039550781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,8,1,power_law_1.01,1.0186431884765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,8,1,power_law_1.01,1.1096134185791016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,8,1,power_law_1.01,1.239508514404297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,8,1,power_law_1.01,1.5177267456054688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,8,1,power_law_1.01,1.7445114135742188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,8,1,power_law_1.01,0.05500671863555908
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,8,1,power_law_1.01,2.3034854125976563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,8,1,power_law_1.01,0.06462207794189453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,8,1,power_law_1.01,0.13201855659484862
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,8,1,power_law_1.01,0.21588224411010742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,8,1,power_law_1.01,2.799205627441406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,8,1,power_law_1.01,0.08928832054138183
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,8,1,power_law_1.01,0.3198687934875488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,8,1,power_law_1.01,0.3822592163085937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,8,1,power_law_1.01,3.8883001708984373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,8,1,power_law_1.01,0.45616062164306637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,8,1,power_law_1.01,0.47110462188720703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,8,1,power_law_1.01,0.4844179153442383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,8,1,power_law_1.01,0.5007161712646484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,8,1,power_law_1.01,0.5059385681152344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,8,1,power_law_1.01,0.5273075103759766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,8,1,power_law_1.01,0.5425043106079102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,8,1,power_law_1.01,5.0399038696289065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,8,1,power_law_1.01,0.5473740768432618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,8,1,power_law_1.01,0.5677286529541016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,8,1,power_law_1.01,0.6338675308227539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,8,1,power_law_1.01,0.6601567840576171
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,8,1,power_law_1.01,0.69032958984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,8,1,power_law_1.01,0.7616038513183594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,8,1,power_law_1.01,0.8355609893798828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,8,1,power_law_1.01,0.9762022399902344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,8,1,power_law_1.01,1.1176544189453126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,8,1,power_law_1.01,1.4049363708496094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,8,1,power_law_1.01,1.6912197875976562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,8,1,power_law_1.01,2.2796313476562498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,8,1,power_law_1.01,2.8606854248046876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,8,1,power_law_1.01,0.07873536109924316
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,8,1,power_law_1.01,0.07885824203491211
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,8,1,power_law_1.01,0.07956543922424317
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,8,1,power_law_1.01,0.08114496231079102
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,8,1,power_law_1.01,0.08444800376892089
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,8,1,power_law_1.01,0.0914560031890869
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,8,1,power_law_1.01,0.0988377571105957
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,8,1,power_law_1.01,0.1053657627105713
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,8,1,power_law_1.01,0.11328639984130859
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,8,1,power_law_1.01,0.11889535903930663
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,8,1,power_law_1.01,0.13853759765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,8,1,power_law_1.01,0.18365631103515626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,8,1,power_law_1.01,0.20280832290649414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,8,1,power_law_1.01,0.300631046295166
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,8,1,power_law_1.01,0.24911808013916015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,8,1,power_law_1.01,1.1560883331298828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,8,1,power_law_1.01,0.34426174163818357
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,8,1,power_law_1.01,0.4354278564453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,8,1,power_law_1.01,0.6182412719726562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,8,1,power_law_1.01,0.7930963134765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,8,1,power_law_1.01,1.5325581359863283
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,8,1,power_law_1.01,2.265379180908203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,8,1,power_law_1.01,2.993347778320312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,8,1,power_law_1.2,0.069651198387146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,8,1,power_law_1.2,0.08714176177978515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,8,1,power_law_1.2,0.10911104202270508
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,8,1,power_law_1.01,4.463219299316406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,8,1,power_law_1.2,0.23999744415283203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,8,1,power_law_1.2,0.42478721618652343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,8,1,power_law_1.2,0.37542465209960935
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,8,1,power_law_1.2,0.5524300765991211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,8,1,power_law_1.2,0.5652755355834961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,8,1,power_law_1.2,0.5996102523803711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,8,1,power_law_1.2,0.6178681564331054
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,8,1,power_law_1.2,0.6396582412719727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,8,1,power_law_1.2,0.6953075408935547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,8,1,power_law_1.2,0.7357484436035155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,8,1,power_law_1.2,0.7813362884521484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,8,1,power_law_1.2,0.8618233489990235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,8,1,power_law_1.2,0.9669439697265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,8,1,power_law_1.2,1.0231231689453124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,8,1,power_law_1.2,1.1186962890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,8,1,power_law_1.2,1.2628460693359376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,8,1,power_law_1.2,0.5832204818725586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,8,1,power_law_1.2,1.5183775329589844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,8,1,power_law_1.2,1.8199162292480469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,8,1,power_law_1.2,0.15272128105163574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,8,1,power_law_1.2,2.3090074157714846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,8,1,power_law_1.2,0.04975615978240967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,8,1,power_law_1.2,3.905594787597656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,8,1,power_law_1.2,0.08531328201293945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,8,1,power_law_1.2,0.05956352233886719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,8,1,power_law_1.2,0.13303680419921876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,8,1,power_law_1.2,0.2074060821533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,8,1,power_law_1.2,0.3668646240234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,8,1,power_law_1.2,0.4509964752197265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,8,1,power_law_1.2,0.30410112380981447
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,8,1,power_law_1.2,0.46334079742431644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,8,1,power_law_1.2,5.042796020507812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,8,1,power_law_1.2,0.47378433227539063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,8,1,power_law_1.2,0.4865011215209961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,8,1,power_law_1.2,0.5184640121459961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,8,1,power_law_1.2,0.5640633773803712
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,8,1,power_law_1.2,0.5025638580322266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,8,1,power_law_1.2,0.5375980758666993
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,8,1,power_law_1.2,0.548223991394043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,8,1,power_law_1.2,2.8256985473632814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,8,1,power_law_1.2,0.6362131118774415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,8,1,power_law_1.2,0.6722220611572266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,8,1,power_law_1.2,0.7055232238769531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,8,1,power_law_1.2,0.7912646484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,8,1,power_law_1.2,0.8544761657714843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,8,1,power_law_1.2,1.0134015655517579
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,8,1,power_law_1.2,1.1595692443847656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,8,1,power_law_1.2,1.4551148986816407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,8,1,power_law_1.2,1.736702117919922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,8,1,power_law_1.2,0.07868735790252686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,8,1,power_law_1.2,2.324165802001953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,8,1,power_law_1.2,2.9110571289062497
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,8,1,power_law_1.2,0.08111616134643554
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,8,1,power_law_1.2,0.09133760452270508
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,8,1,power_law_1.2,0.09767359733581543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,8,1,power_law_1.2,0.10566975593566894
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,8,1,power_law_1.2,0.11879103660583497
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,8,1,power_law_1.2,0.11210240364074706
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,8,1,power_law_1.2,0.13789631843566894
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,8,1,power_law_1.2,0.1782182312011719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,8,1,power_law_1.2,0.20602624893188476
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,8,1,power_law_1.2,0.2512928009033203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,8,1,power_law_1.2,0.07962687969207763
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,8,1,power_law_1.2,0.08453375816345214
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,8,1,power_law_1.2,0.0787443208694458
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,8,1,power_law_1.2,0.3011884880065918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,8,1,power_law_1.2,0.34434368133544924
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,8,1,power_law_1.2,0.4333292770385742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,8,1,power_law_1.2,0.6186028671264648
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,8,1,power_law_1.2,0.7988134765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,8,1,power_law_1.2,1.1616204833984374
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,8,1,power_law_1.2,1.5306297302246095
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,8,1,power_law_1.2,2.2585433959960937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,8,1,power_law_1.2,2.992637939453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,8,balanced,0.056131839752197266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,8,balanced,0.0997811222076416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,8,balanced,0.18417280197143554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,8,balanced,0.2170854377746582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,8,balanced,0.22045568466186521
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,8,balanced,0.029648640155792237
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,8,balanced,0.22788608551025388
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,8,balanced,0.21743488311767578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,8,balanced,0.22417280197143558
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,8,balanced,0.22498432159423828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,8,balanced,0.2338764762878418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,8,balanced,0.24072832107543946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,8,balanced,0.24718719482421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,8,balanced,0.2579315185546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,8,balanced,0.2857612800598145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,8,balanced,0.31016447067260744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,8,balanced,0.3553676986694336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,8,1,power_law_1.2,4.482740478515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,8,balanced,0.40165760040283205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,8,balanced,0.5497804641723633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,8,balanced,0.9071731567382812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,8,balanced,1.167452163696289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,8,balanced,1.776741180419922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,8,balanced,0.6485363006591797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,8,balanced,0.03011840105056763
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,8,balanced,2.4374732971191406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,8,balanced,0.04349823951721191
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,8,balanced,0.07340288162231445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,8,balanced,0.031820800304412845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,8,balanced,0.22216960906982425
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,8,balanced,0.12607104301452637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,8,balanced,0.13051136016845705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,8,balanced,0.12569087982177735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,8,balanced,0.030132479667663574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,8,balanced,0.11859968185424805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,8,balanced,0.12346495628356932
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,8,balanced,0.12468352317810058
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,8,balanced,0.11217280387878419
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,8,balanced,0.12798463821411132
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,8,balanced,0.08590975761413575
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,8,balanced,0.12978431701660156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,8,balanced,0.14460800170898438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,8,balanced,0.15898624420166016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,8,balanced,0.17360767364501953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,8,balanced,0.20512256622314454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,8,balanced,0.23597951889038088
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,8,balanced,0.29924736022949217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,8,balanced,0.3605120086669922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,8,balanced,0.1280294418334961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,8,balanced,0.5102374267578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,8,balanced,0.6302566528320312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,8,balanced,0.8955942535400391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,8,balanced,0.13264512062072753
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,8,balanced,1.1748786926269532
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,1,8,balanced,0.022767360210418704
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,1,8,balanced,0.021584639549255373
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,1,8,balanced,0.06694399833679199
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,1,8,balanced,0.06662015914916992
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,1,8,balanced,0.06709375858306885
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,1,8,balanced,0.06745215892791748
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,1,8,balanced,0.06747136116027833
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,1,8,balanced,0.06743167877197266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,1,8,balanced,0.06802432060241699
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,1,8,balanced,0.022016000747680665
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,1,8,balanced,0.06811007976531983
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,1,8,balanced,0.068122878074646
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,1,8,balanced,0.06853375911712647
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,1,8,balanced,0.06858367919921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,1,8,balanced,0.07017600059509277
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,1,8,balanced,0.07172351837158203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,1,8,balanced,0.07372159957885742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,1,8,balanced,0.07619328022003173
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,1,8,balanced,0.0805299186706543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,1,8,balanced,0.08494336128234863
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,1,8,balanced,0.06921599864959717
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,1,8,balanced,0.04292096138000488
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,1,8,balanced,0.029605119228363036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,8,power_law_1.01,0.061537280082702636
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,8,power_law_1.01,0.09402239799499512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,8,power_law_1.01,0.13678848266601562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,8,power_law_1.01,0.16048896789550782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,8,power_law_1.01,0.17920896530151367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,8,power_law_1.01,0.1830400085449219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,8,power_law_1.01,0.20930688858032226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,8,power_law_1.01,0.2191705513000488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,8,power_law_1.01,0.21899776458740233
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,8,power_law_1.01,0.21968896865844725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,8,power_law_1.01,0.22196352005004885
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,8,power_law_1.01,0.22391168594360353
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,8,power_law_1.01,0.21417472839355467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,8,power_law_1.01,0.22113407135009763
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,8,power_law_1.01,0.22381568908691407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,8,power_law_1.01,0.23061887741088868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,8,power_law_1.01,0.2414656066894531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,8,power_law_1.01,0.2701606369018555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,8,power_law_1.01,0.3236249542236328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,8,power_law_1.01,0.3664896011352539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,8,power_law_1.01,0.41627777099609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,8,power_law_1.01,0.570568962097168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,8,power_law_1.01,0.7250611114501953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,8,power_law_1.01,0.9921215820312501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,8,power_law_1.01,1.3363200378417968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,8,power_law_1.01,1.8136384582519534
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,1,8,balanced,0.11666815757751466
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,1,8,balanced,0.1250547218322754
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,1,8,balanced,0.2132172775268555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,1,8,balanced,0.16656639099121093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,8,power_law_1.01,0.084136962890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,8,power_law_1.01,2.531226806640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,8,power_law_1.01,0.08155391693115235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,8,power_law_1.01,0.07024896144866943
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,8,power_law_1.01,0.10716287612915039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,8,power_law_1.01,0.11352831840515136
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,8,power_law_1.01,0.1154368019104004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,8,power_law_1.01,0.11049087524414063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,8,power_law_1.01,0.1252889633178711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,8,power_law_1.01,0.1272396755218506
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,8,power_law_1.01,0.11721343994140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,8,power_law_1.01,0.12360063552856446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,8,power_law_1.01,0.12143487930297851
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,8,power_law_1.01,0.12327679634094238
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,8,power_law_1.01,0.12543871879577637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,8,power_law_1.01,0.12569984436035156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,8,power_law_1.01,0.04363647937774658
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,8,power_law_1.01,0.1442585563659668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,8,power_law_1.01,0.16127744674682618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,8,power_law_1.01,0.20734720230102538
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,8,power_law_1.01,0.22932096481323244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,8,power_law_1.01,0.2909427261352539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,8,power_law_1.01,0.3681241607666016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,8,power_law_1.01,0.4604467010498047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,8,power_law_1.01,0.6192089462280274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,8,power_law_1.01,0.7698355102539063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,1,8,power_law_1.01,0.028816640377044678
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,1,8,power_law_1.01,0.044060158729553225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,8,power_law_1.01,1.1212889862060547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,1,8,power_law_1.01,0.048474879264831544
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,1,8,power_law_1.01,0.0524671983718872
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,1,8,power_law_1.01,0.057992959022521974
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,1,8,power_law_1.01,0.06181248188018799
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,1,8,power_law_1.01,0.06317311763763428
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,1,8,power_law_1.01,0.06046207904815674
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,1,8,power_law_1.01,0.061898241043090815
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,1,8,power_law_1.01,0.0636787223815918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,1,8,power_law_1.01,0.06111743927001954
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,1,8,power_law_1.01,0.06243072032928467
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,1,8,power_law_1.01,0.06595967769622803
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,1,8,power_law_1.01,0.0649241590499878
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,1,8,power_law_1.01,0.062277121543884276
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,1,8,power_law_1.01,0.06540671825408936
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,1,8,power_law_1.01,0.07582079887390136
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,1,8,power_law_1.01,0.08766079902648925
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,1,8,power_law_1.01,0.06209023952484131
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,8,power_law_1.01,0.0776473617553711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,1,8,power_law_1.01,0.07244800090789795
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,1,8,power_law_1.01,0.10098943710327149
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,1,8,power_law_1.01,0.12744064331054689
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,1,8,power_law_1.01,0.14677248001098633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,8,power_law_1.2,0.09220735549926758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,8,power_law_1.2,0.0985536003112793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,8,power_law_1.2,0.061479678153991704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,8,power_law_1.2,0.14618495941162107
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,8,power_law_1.2,0.17545728683471679
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,8,power_law_1.2,0.18084991455078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,8,power_law_1.2,0.21409791946411133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,8,power_law_1.2,0.2164031982421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,8,power_law_1.2,0.21674112319946287
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,8,power_law_1.2,0.21934976577758789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,8,power_law_1.2,0.22718719482421873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,8,power_law_1.2,0.2221708869934082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,1,8,power_law_1.01,0.20590848922729493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,8,power_law_1.2,0.2193075180053711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,8,power_law_1.2,0.22316415786743163
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,8,power_law_1.2,0.22708351135253907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,8,power_law_1.2,0.24208768844604492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,8,power_law_1.2,0.2702720069885254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,8,power_law_1.2,0.2147622489929199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,8,power_law_1.2,0.33263359069824217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,8,power_law_1.2,0.3833446502685547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,8,power_law_1.2,0.47098369598388673
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,8,power_law_1.2,0.6482009887695312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,8,power_law_1.2,0.8146431732177735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,8,power_law_1.2,0.9882150268554687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,8,power_law_1.2,1.4357363891601562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,8,power_law_1.2,2.079636535644531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,8,power_law_1.2,0.07000703811645508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,8,power_law_1.2,0.07027711868286132
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,8,power_law_1.2,0.08240256309509278
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,8,power_law_1.2,2.7482394409179687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,8,power_law_1.2,0.08720128059387207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,8,power_law_1.2,0.10967935562133788
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,8,power_law_1.2,0.10587648391723632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,8,power_law_1.2,0.10374655723571777
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,8,power_law_1.2,0.11556480407714845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,8,power_law_1.2,0.11699711799621582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,8,power_law_1.2,0.126627836227417
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,8,power_law_1.2,0.12450431823730468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,8,power_law_1.2,0.12381312370300293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,8,power_law_1.2,0.12162176132202149
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,8,power_law_1.2,0.12387968063354493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,8,power_law_1.2,0.12427007675170898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,8,power_law_1.2,0.12987648010253908
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,8,power_law_1.2,0.043773441314697265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,8,power_law_1.2,0.14405247688293457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,8,power_law_1.2,0.17392639160156248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,8,power_law_1.2,0.21614719390869142
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,8,power_law_1.2,0.24922367095947268
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,8,power_law_1.2,0.32243072509765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,1,8,power_law_1.01,0.25053312301635744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,8,power_law_1.2,0.38811264038085935
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,8,power_law_1.2,0.5234175872802734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,1,8,power_law_1.2,0.028619520664215088
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,8,power_law_1.2,0.6936537933349609
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,1,8,power_law_1.2,0.042225918769836425
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,1,8,power_law_1.2,0.043921918869018556
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,1,8,power_law_1.2,0.054142718315124516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,1,8,power_law_1.2,0.05643904209136963
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,8,power_law_1.2,1.2840103149414062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,1,8,power_law_1.2,0.05755008220672607
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,1,8,power_law_1.2,0.06052991867065429
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,1,8,power_law_1.2,0.0562559986114502
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,1,8,power_law_1.2,0.061999359130859376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,1,8,power_law_1.2,0.06075776100158692
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,1,8,power_law_1.2,0.06162559986114502
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,1,8,power_law_1.2,0.061224961280822755
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,1,8,power_law_1.2,0.06513792037963867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,1,8,power_law_1.2,0.06535168170928955
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,1,8,power_law_1.2,0.06721280097961427
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,1,8,power_law_1.2,0.06939648151397705
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,1,8,power_law_1.2,0.07081471920013428
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,1,8,power_law_1.2,0.08667648315429688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,8,power_law_1.2,0.9049855804443359
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,1,8,power_law_1.01,0.36305023193359376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,1,8,power_law_1.01,0.49763328552246094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,1,8,power_law_1.2,0.0784819221496582
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,1,8,power_law_1.2,0.09815551757812499
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,1,8,power_law_1.2,0.11414527893066406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,2,4,balanced,0.02644479990005493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,2,4,balanced,0.026679039001464844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,2,4,balanced,0.044097280502319335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,2,4,balanced,0.06022784233093261
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,2,4,balanced,0.09736960411071778
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,2,4,balanced,0.17486848831176757
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,2,4,balanced,0.17624576568603517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,2,4,balanced,0.1772979164123535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,2,4,balanced,0.20357120513916013
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,2,4,balanced,0.20537216186523438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,2,4,balanced,0.2083020782470703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,2,4,balanced,0.2099839973449707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,2,4,balanced,0.21378816604614256
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,1,8,power_law_1.2,0.16417280197143555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,2,4,balanced,0.21775232315063478
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,2,4,balanced,0.22313983917236327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,2,4,balanced,0.23581695556640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,2,4,balanced,0.25465087890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,2,4,balanced,0.22802047729492186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,2,4,balanced,0.2745932769775391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,2,4,balanced,0.31097600936889647
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,2,4,balanced,0.3522713470458984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,2,4,balanced,0.4844927978515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,2,4,balanced,0.5703359985351563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,2,4,balanced,0.793930206298828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,2,4,balanced,1.0188787078857422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,2,4,balanced,0.027640318870544432
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,2,4,balanced,0.02806272029876709
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,2,4,balanced,1.48313720703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,2,4,balanced,0.03143807888031006
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,1,8,power_law_1.2,0.2928998374938965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,2,4,balanced,0.04765696048736572
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,2,4,balanced,0.07497856140136719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,2,4,balanced,0.11639807701110841
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,2,4,balanced,0.11765248298645019
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,1,8,power_law_1.2,0.42611328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,2,4,balanced,0.1287987232208252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,2,4,balanced,0.12954624176025392
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,2,4,balanced,1.942942657470703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,2,4,balanced,0.12970239639282227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,2,4,balanced,0.13052160263061524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,2,4,balanced,0.13161727905273438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,2,4,balanced,0.13271679878234863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,2,4,balanced,0.13501440048217775
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,2,4,balanced,0.13833087921142578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,2,4,balanced,0.14458880424499512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,2,4,balanced,0.16093439102172852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,2,4,balanced,0.16921472549438477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,2,4,balanced,0.19882368087768554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,2,4,balanced,0.22540416717529296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,2,4,balanced,0.26659072875976564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,2,4,balanced,0.12150400161743165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,2,4,balanced,0.3231769561767578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,2,4,balanced,0.4686067199707031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,2,4,balanced,0.5626854324340821
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,2,4,balanced,0.7992793273925781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,2,4,balanced,1.0487500762939452
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,1,8,power_law_1.2,0.20042240142822268
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,1,8,power_law_1.2,0.5385715103149413
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,1,8,power_law_1.2,0.7644493103027343
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,2,4,balanced,0.025699839591979978
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,2,4,balanced,0.07246335983276367
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,2,4,balanced,0.07341951847076415
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,2,4,balanced,0.02662528038024902
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,2,4,balanced,0.0281062388420105
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,2,4,balanced,0.07337215900421143
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,2,4,balanced,0.034174718856811524
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,2,4,balanced,0.07411839962005615
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,2,4,balanced,0.07383935928344727
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,2,4,balanced,0.07416063785552979
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,2,4,balanced,0.07452799797058104
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,2,4,balanced,0.07477248191833497
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,2,4,balanced,0.07648640155792237
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,2,4,balanced,0.048846077919006345
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,2,4,balanced,0.07786880016326905
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,2,4,balanced,0.07962240219116211
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,2,4,balanced,0.08255999565124512
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,2,4,balanced,0.08737792015075684
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,2,4,balanced,0.09264896392822267
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,2,4,balanced,0.09964159965515137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,2,4,power_law_1.01,0.045688319206237796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,2,4,power_law_1.01,0.05933695793151855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,2,4,power_law_1.01,0.07800320148468018
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,2,4,power_law_1.01,0.09116031646728515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,2,4,power_law_1.01,0.12293120384216309
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,2,4,power_law_1.01,0.14403840065002443
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,2,4,power_law_1.01,0.16700799942016603
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,2,4,power_law_1.01,0.1763199996948242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,2,4,power_law_1.01,0.19746559143066406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,2,4,power_law_1.01,0.20147584915161132
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,2,4,balanced,0.07111167907714844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,2,4,power_law_1.01,0.19755008697509766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,2,4,power_law_1.01,0.2034444808959961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,2,4,power_law_1.01,0.21360639572143553
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,2,4,power_law_1.01,0.21574655532836914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,2,4,power_law_1.01,0.2218841552734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,2,4,power_law_1.01,0.22616575241088865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,2,4,power_law_1.01,0.259946231842041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,2,4,power_law_1.01,0.28582527160644533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,2,4,power_law_1.01,0.3488332748413086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,2,4,power_law_1.01,0.4002342224121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,2,4,power_law_1.01,0.19231744766235354
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,2,4,power_law_1.01,0.5234201431274415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,2,4,power_law_1.01,0.6583872222900391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,2,4,power_law_1.01,0.8796173095703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,2,4,power_law_1.01,1.1185011291503906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,2,4,balanced,0.07060991764068604
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,2,4,power_law_1.01,1.6032025146484377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,2,4,power_law_1.01,0.046625280380249025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,2,4,power_law_1.01,0.05211135864257812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,2,4,power_law_1.01,0.07139711856842042
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,2,4,power_law_1.01,2.1535462951660156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,2,4,power_law_1.01,0.07074687957763673
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,2,4,balanced,0.07303679943084716
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,2,4,power_law_1.01,0.09847807884216309
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,2,4,power_law_1.01,0.1096332836151123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,2,4,power_law_1.01,0.11583999633789062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,2,4,power_law_1.01,0.1113804817199707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,2,4,power_law_1.01,0.11403519630432128
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,2,4,power_law_1.01,0.11750656127929689
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,2,4,power_law_1.01,0.12207488059997558
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,2,4,power_law_1.01,0.12691967964172363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,2,4,power_law_1.01,0.12925824165344238
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,2,4,power_law_1.01,0.13233152389526368
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,2,4,power_law_1.01,0.13724543571472167
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,2,4,power_law_1.01,0.1570995235443115
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,2,4,power_law_1.01,0.16752639770507813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,2,4,power_law_1.01,0.20097663879394528
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,2,4,power_law_1.01,0.22935935974121097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,2,4,power_law_1.01,0.29491840362548827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,2,4,power_law_1.01,0.34693248748779293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,2,4,power_law_1.01,0.4869388961791993
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,2,4,power_law_1.01,0.028853759765624997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,2,4,power_law_1.01,0.627243537902832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,2,4,power_law_1.01,0.02876672029495239
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,2,4,power_law_1.01,0.034942719936370846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,2,4,power_law_1.01,0.8667289733886718
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,2,4,power_law_1.01,0.04235904216766358
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,2,4,power_law_1.01,0.046996479034423826
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,2,4,power_law_1.01,0.05247231960296631
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,2,4,power_law_1.01,1.1559871673583983
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,2,4,power_law_1.01,0.055253758430480956
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,2,4,power_law_1.01,0.0650713586807251
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,2,4,power_law_1.01,0.06299263954162597
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,2,4,power_law_1.01,0.06790527820587158
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,2,4,power_law_1.01,0.06310400009155273
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,2,4,power_law_1.01,0.06338687896728515
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,2,4,power_law_1.01,0.06630784034729004
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,2,4,balanced,0.2692403221130371
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,2,4,balanced,0.13910400390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,2,4,power_law_1.01,0.06776319980621338
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,2,4,power_law_1.01,0.06926591873168945
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,2,4,power_law_1.01,0.07224959850311279
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,2,4,power_law_1.01,0.07454336166381836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,2,4,balanced,0.20989055633544923
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,2,4,balanced,0.1528870391845703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,2,4,power_law_1.01,0.0706329584121704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,2,4,power_law_1.01,0.09056384086608886
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,2,4,power_law_1.01,0.08383359909057617
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,2,4,power_law_1.01,0.08648832321166992
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,2,4,power_law_1.01,0.10501888275146484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,2,4,power_law_1.01,0.15095935821533205
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,2,4,power_law_1.01,0.19122175216674803
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,2,4,power_law_1.01,0.26343679428100586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,2,4,power_law_1.2,0.04578432083129883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,2,4,power_law_1.2,0.05911935806274414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,2,4,power_law_1.2,0.06903295993804932
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,2,4,power_law_1.2,0.08776320457458496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,2,4,power_law_1.2,0.12407423973083495
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,2,4,power_law_1.2,0.1725555229187012
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,2,4,power_law_1.2,0.1748531150817871
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,2,4,power_law_1.01,0.12416511535644532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,2,4,power_law_1.2,0.19543296813964844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,2,4,power_law_1.2,0.19823488235473632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,2,4,power_law_1.2,0.19023231506347654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,2,4,power_law_1.2,0.20327808380126955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,2,4,power_law_1.2,0.20460927963256834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,2,4,power_law_1.2,0.20299007415771486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,2,4,power_law_1.2,0.21527168273925784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,2,4,power_law_1.2,0.22255359649658205
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,2,4,power_law_1.01,0.349620475769043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,2,4,power_law_1.2,0.2283955192565918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,2,4,power_law_1.2,0.25398656845092776
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,2,4,power_law_1.2,0.2924172782897949
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,2,4,power_law_1.2,0.3574720001220703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,2,4,power_law_1.2,0.41693183898925784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,2,4,power_law_1.2,0.5535142517089844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,2,4,power_law_1.2,0.6697702026367187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,2,4,power_law_1.2,0.030188798904418945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,2,4,power_law_1.2,0.04491392135620117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,2,4,power_law_1.2,1.2168370819091796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,2,4,power_law_1.2,0.050019841194152835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,2,4,power_law_1.2,0.07140096187591552
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,2,4,power_law_1.2,0.0741593599319458
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,2,4,power_law_1.2,1.6920690917968748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,2,4,power_law_1.2,0.0844927978515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,2,4,power_law_1.2,0.9409101104736328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,2,4,power_law_1.2,0.09517696380615234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,2,4,power_law_1.2,0.11400575637817383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,2,4,power_law_1.2,0.11476736068725586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,2,4,power_law_1.2,0.12392576217651367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,2,4,power_law_1.2,2.195924530029297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,2,4,power_law_1.2,0.12305407524108887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,2,4,power_law_1.2,0.12858752250671385
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,2,4,power_law_1.2,0.1192140769958496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,2,4,power_law_1.2,0.13094655990600584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,2,4,power_law_1.2,0.13289088249206543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,2,4,power_law_1.2,0.13957119941711427
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,2,4,power_law_1.2,0.15554944038391114
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,2,4,power_law_1.2,0.17617279052734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,2,4,power_law_1.2,0.20766975402832033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,2,4,power_law_1.2,0.24058111190795897
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,2,4,power_law_1.2,0.3077491188049316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,2,4,power_law_1.2,0.11649408340454101
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,2,4,power_law_1.2,0.37437824249267576
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,2,4,power_law_1.2,0.028482561111450193
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,2,4,power_law_1.2,0.4995929718017578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,2,4,power_law_1.2,0.6388851165771484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,2,4,power_law_1.2,0.03461375951766968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,2,4,power_law_1.2,0.03862272024154663
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,2,4,power_law_1.2,0.04533631801605224
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,2,4,power_law_1.2,0.9134540557861328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,2,4,power_law_1.2,0.05418623924255371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,2,4,power_law_1.2,0.14530688285827636
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,2,4,power_law_1.2,0.06515583992004395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,2,4,power_law_1.2,0.061792001724243165
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,2,4,power_law_1.2,0.06113408088684082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,2,4,power_law_1.2,1.1597132873535156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,2,4,power_law_1.2,0.06950143814086915
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,2,4,power_law_1.2,0.06380928039550782
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,2,4,power_law_1.2,0.06589312076568603
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,2,4,power_law_1.2,0.0667852783203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,2,4,power_law_1.2,0.07203968048095703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,2,4,power_law_1.2,0.07115776062011718
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,2,4,power_law_1.2,0.0737113618850708
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,2,4,power_law_1.2,0.0496230411529541
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,2,4,power_law_1.01,0.47257854461669924
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,2,4,power_law_1.01,0.6237580871582031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,2,4,power_law_1.2,0.07976319789886474
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,2,4,power_law_1.2,0.09064064025878907
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,2,4,power_law_1.2,0.09801471710205079
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,2,4,power_law_1.2,0.14928000450134277
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,2,4,power_law_1.2,0.11993344306945801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,4,2,balanced,0.033337600231170654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,4,2,balanced,0.042328319549560546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,4,2,balanced,0.060672001838684084
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,2,4,power_law_1.2,0.18743936538696288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,4,2,balanced,0.09636480331420898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,4,2,balanced,0.17776384353637695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,4,2,balanced,0.17748479843139647
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,4,2,balanced,0.1775040054321289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,4,2,balanced,0.17841152191162107
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,4,2,balanced,0.1783679962158203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,4,2,balanced,0.18040576934814453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,4,2,balanced,0.20867200851440432
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,4,2,balanced,0.21075967788696287
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,4,2,balanced,0.21246335983276365
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,4,2,balanced,0.21736576080322267
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,4,2,balanced,0.02316672086715698
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,4,2,balanced,0.21801855087280275
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,4,2,balanced,0.22682367324829103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,4,2,balanced,0.24272384643554687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,4,2,balanced,0.2619289588928223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,4,2,balanced,0.2961126327514648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,4,2,balanced,0.3342758560180664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,4,2,balanced,0.4467340850830078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,4,2,balanced,0.5270233535766602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,4,2,balanced,0.734940185546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,4,2,balanced,0.9383424377441406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,4,2,balanced,1.3609472656250001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,4,2,balanced,0.02636415958404541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,4,2,balanced,0.02852479934692383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,4,2,balanced,0.032624640464782716
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,4,2,balanced,1.775694122314453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,4,2,balanced,0.08717951774597169
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,4,2,balanced,0.13246848106384276
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,4,2,balanced,0.13153152465820311
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,4,2,balanced,0.1327500820159912
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,4,2,balanced,0.13396096229553223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,4,2,balanced,0.13569151878356933
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,4,2,balanced,0.047688961029052734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,4,2,balanced,0.15393792152404787
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,4,2,balanced,0.15499648094177246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,4,2,balanced,0.1565721607208252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,4,2,balanced,0.1329088020324707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,4,2,balanced,0.15810943603515626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,4,2,balanced,0.16017919540405273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,4,2,balanced,0.1634547233581543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,4,2,balanced,0.18382335662841798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,4,2,balanced,0.20275455474853515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,4,2,balanced,0.2275929641723633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,4,2,balanced,0.2698726463317871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,4,2,balanced,0.31265792846679685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,4,2,balanced,0.4658355331420898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,4,2,balanced,0.1720627212524414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,4,2,balanced,0.5499993515014648
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,2,4,power_law_1.2,0.2569753646850586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,4,2,balanced,0.7817203521728515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,4,2,balanced,1.0273741149902345
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,2,4,power_law_1.2,0.3496870422363281
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,2,4,power_law_1.2,0.5834688186645508
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,2,4,power_law_1.2,0.7593433380126953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,2,4,power_law_1.2,1.0696857452392579
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,4,2,balanced,0.03831552028656006
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,4,2,balanced,0.09174271583557128
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,4,2,balanced,0.09139200210571288
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,4,2,balanced,0.09188991546630859
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,4,2,balanced,0.037140479087829595
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,4,2,balanced,0.09172991752624512
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,4,2,balanced,0.09292287826538086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,4,2,balanced,0.04019455909729004
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,4,2,balanced,0.09417216300964355
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,4,2,balanced,0.09510272026062012
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,4,2,balanced,0.09777664184570313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,4,2,balanced,0.048256001472473144
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,4,2,balanced,0.0998259162902832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,4,2,balanced,0.10328319549560547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,4,2,balanced,0.10869503974914552
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,4,2,balanced,0.11715328216552734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,4,2,balanced,0.1266419219970703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,4,2,balanced,0.06337279796600341
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,4,2,balanced,0.13995264053344728
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,4,2,balanced,0.09087360382080079
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,4,2,balanced,0.09140992164611816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,4,2,power_law_1.01,0.04495744228363037
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,4,2,balanced,0.09110527992248535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,4,2,power_law_1.01,0.028268799781799313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,4,2,power_law_1.01,0.05484543800354004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,4,2,power_law_1.01,0.06649727821350097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,4,2,power_law_1.01,0.11305343627929687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,4,2,power_law_1.01,0.15666175842285157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,4,2,power_law_1.01,0.16516992568969727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,4,2,power_law_1.01,0.16544384002685547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,4,2,power_law_1.01,0.16995967864990233
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,4,2,power_law_1.01,0.09422207832336425
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,4,2,power_law_1.01,0.16986495971679688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,4,2,power_law_1.01,0.1922393608093262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,4,2,power_law_1.01,0.19594751358032228
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,4,2,power_law_1.01,0.2006028747558594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,4,2,power_law_1.01,0.20526079177856443
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,4,2,power_law_1.01,0.21279359817504884
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,4,2,power_law_1.01,0.2186227226257324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,4,2,power_law_1.01,0.23924480438232423
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,4,2,power_law_1.01,0.2700428771972656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,4,2,power_law_1.01,0.32678272247314455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,4,2,power_law_1.01,0.3705356979370117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,4,2,power_law_1.01,0.4781273651123047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,4,2,power_law_1.01,0.5844812774658203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,4,2,power_law_1.01,0.8140902709960937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,4,2,power_law_1.01,1.0408755493164061
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,4,2,power_law_1.01,0.029245440959930417
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,4,2,power_law_1.01,1.4553152465820314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,4,2,power_law_1.01,0.03250175952911377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,4,2,power_law_1.01,0.04517375946044922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,4,2,power_law_1.01,0.058197760581970216
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,4,2,power_law_1.01,0.0769920015335083
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,4,2,power_law_1.01,1.8760089111328124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,4,2,power_law_1.01,0.09719039916992187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,4,2,power_law_1.01,0.1088755226135254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,4,2,power_law_1.01,0.11746175765991211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,4,2,power_law_1.01,0.1166643238067627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,4,2,power_law_1.01,0.11799679756164552
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,4,2,power_law_1.01,0.12348031997680664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,4,2,power_law_1.01,0.141909761428833
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,4,2,power_law_1.01,0.14377856254577637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,4,2,power_law_1.01,0.1457536029815674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,4,2,power_law_1.01,0.1486233615875244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,4,2,power_law_1.01,0.15188608169555665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,4,2,power_law_1.01,0.15525504112243654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,4,2,power_law_1.01,0.16978559494018555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,4,2,power_law_1.01,0.18178176879882812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,4,2,power_law_1.01,0.21000831604003906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,4,2,power_law_1.01,0.24483455657958983
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,4,2,power_law_1.01,0.29998464584350587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,4,2,power_law_1.01,0.3604441452026367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,4,2,power_law_1.01,0.4827635192871094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,4,2,power_law_1.01,0.6034457778930664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,4,2,power_law_1.01,0.038314239978790285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,4,2,power_law_1.01,0.8558771514892578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,4,2,power_law_1.01,1.090154266357422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,4,2,power_law_1.01,0.040341758728027345
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,4,2,power_law_1.01,0.046012158393859866
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,4,2,power_law_1.01,0.05228928089141845
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,4,2,power_law_1.01,0.05867519855499268
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,4,2,power_law_1.01,0.06346367835998536
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,4,2,balanced,0.20445568084716798
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,4,2,power_law_1.01,0.07557248115539551
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,4,2,power_law_1.01,0.07982592105865478
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,4,2,power_law_1.01,0.08293375968933106
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,4,2,power_law_1.01,0.08117376327514648
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,4,2,power_law_1.01,0.08509056091308594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,4,2,power_law_1.01,0.08733056068420411
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,4,2,power_law_1.01,0.09003904342651367
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,4,2,balanced,0.3221286392211914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,4,2,power_law_1.01,0.0922265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,4,2,power_law_1.01,0.09569408416748046
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,4,2,power_law_1.01,0.08080384254455567
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,4,2,balanced,0.2297932815551758
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,4,2,balanced,0.4174835205078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,4,2,power_law_1.01,0.09931903839111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,4,2,power_law_1.01,0.28461183547973634
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,4,2,power_law_1.01,0.4027391815185547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,4,2,power_law_1.01,0.15545215606689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,4,2,power_law_1.01,0.12708095550537107
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,4,2,power_law_1.01,0.2384934425354004
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,4,2,power_law_1.01,0.1819852828979492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,4,2,power_law_1.2,0.02716928005218506
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,4,2,power_law_1.2,0.050677762031555176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,4,2,power_law_1.2,0.06812160015106201
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,4,2,power_law_1.2,0.09130623817443848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,4,2,power_law_1.2,0.10465151786804198
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,4,2,power_law_1.2,0.15889151573181154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,4,2,power_law_1.2,0.1664780807495117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,4,2,power_law_1.2,0.16236543655395508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,4,2,power_law_1.2,0.16561279296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,4,2,power_law_1.2,0.1669081687927246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,4,2,power_law_1.2,0.19267072677612304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,4,2,power_law_1.2,0.1951296043395996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,4,2,power_law_1.2,0.20173952102661136
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,4,2,power_law_1.2,0.207324161529541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,4,2,power_law_1.2,0.2118988800048828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,4,2,power_law_1.2,0.22003711700439452
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,4,2,power_law_1.2,0.04570496082305908
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,4,2,power_law_1.2,0.23809535980224608
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,4,2,power_law_1.2,0.2824076843261719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,4,2,power_law_1.2,0.32866943359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,4,2,power_law_1.2,0.3771392059326172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,4,2,power_law_1.2,0.4878617477416992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,4,2,power_law_1.2,0.59976318359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,4,2,power_law_1.2,0.8386700439453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,4,2,power_law_1.2,1.0516121673583985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,4,2,power_law_1.2,0.02880768060684204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,4,2,power_law_1.2,1.4859341430664064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,4,2,power_law_1.2,0.03290112018585205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,4,2,power_law_1.2,0.04288383960723877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,4,2,power_law_1.2,1.89578369140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,4,2,power_law_1.2,0.05700096130371094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,4,2,power_law_1.2,0.0775219202041626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,4,2,power_law_1.2,0.09193984031677246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,4,2,power_law_1.2,0.10776703834533692
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,4,2,power_law_1.2,0.11168512344360351
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,4,2,power_law_1.2,0.11307392120361329
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,4,2,power_law_1.2,0.11708543777465821
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,4,2,power_law_1.2,0.12345600128173828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,4,2,power_law_1.2,0.14108415603637697
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,4,2,power_law_1.2,0.14385536193847656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,4,2,power_law_1.2,0.1459008026123047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,4,2,power_law_1.2,0.14829824447631837
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,4,2,power_law_1.2,0.15273856163024901
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,4,2,power_law_1.2,0.1584768009185791
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,4,2,power_law_1.2,0.17159679412841797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,4,2,power_law_1.2,0.18850175857543944
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,4,2,power_law_1.2,0.2231385612487793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,4,2,power_law_1.2,0.25057024002075196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,4,2,power_law_1.2,0.31198848724365236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,4,2,power_law_1.2,0.37212032318115235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,4,2,power_law_1.2,0.5045375823974609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,4,2,power_law_1.2,0.6180377578735351
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,4,2,power_law_1.2,0.8645465850830079
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,4,2,power_law_1.2,0.03830399990081787
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,4,2,power_law_1.2,0.039946238994598385
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,4,2,power_law_1.2,1.1079180908203123
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,4,2,power_law_1.2,0.046045441627502445
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,4,2,power_law_1.2,0.05147007942199707
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,4,2,power_law_1.2,0.05761792182922363
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,4,2,power_law_1.2,0.060995841026306154
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,4,2,power_law_1.2,0.07535232067108154
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,4,2,power_law_1.2,0.07867263793945313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,4,2,power_law_1.2,0.08063743591308595
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,4,2,power_law_1.2,0.08173184394836426
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,4,2,power_law_1.2,0.08427647590637208
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,4,2,power_law_1.2,0.08609791755676269
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,4,2,power_law_1.2,0.08976639747619629
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,4,2,power_law_1.2,0.09271807670593261
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,4,2,power_law_1.2,0.0967027187347412
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,4,2,power_law_1.2,0.10138367652893066
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,4,2,power_law_1.2,0.1092505645751953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,4,2,power_law_1.01,0.5787392044067383
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,4,2,power_law_1.01,0.7251289367675782
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,4,2,power_law_1.01,1.0904422760009767
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,4,2,power_law_1.2,0.12902912139892578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,4,2,power_law_1.2,0.1862054443359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,4,2,power_law_1.2,0.14669055938720704
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,4,2,power_law_1.2,0.22753536224365237
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,4,2,power_law_1.2,0.3002521514892578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,8,1,balanced,0.030379519462585446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,8,1,balanced,0.044408960342407225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,8,1,balanced,0.05358528137207032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,8,1,balanced,0.072707200050354
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,8,1,balanced,0.11173760414123535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,8,1,balanced,0.19959871292114256
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,4,2,power_law_1.2,0.4040537643432617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,8,1,balanced,0.19831680297851562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,8,1,balanced,0.1991526412963867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,8,1,balanced,0.200185604095459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,8,1,balanced,0.20003391265869142
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,8,1,balanced,0.20171072006225588
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,8,1,balanced,0.20381120681762693
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,8,1,balanced,0.20528064727783205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,8,1,balanced,0.2099679946899414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,8,1,balanced,0.2577414321899414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,8,1,balanced,0.26191743850708005
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,8,1,balanced,0.2689139175415039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,8,1,balanced,0.2826483154296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,8,1,balanced,0.2972780799865723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,8,1,balanced,0.3267251205444336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,8,1,balanced,0.35607486724853515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,8,1,balanced,0.47046463012695316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,8,1,balanced,0.5426009750366211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,8,1,balanced,0.7477913665771484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,8,1,balanced,0.9447014617919922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,8,1,balanced,1.3488595581054688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,8,1,balanced,1.7462413024902343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,8,1,balanced,0.040828800201416014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,8,1,balanced,0.04632832050323486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,8,1,balanced,0.03529664039611816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,8,1,balanced,0.0703001594543457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,8,1,balanced,0.11333503723144531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,8,1,balanced,0.18660608291625977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,8,1,balanced,0.1867692756652832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,8,1,balanced,0.18840383529663085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,8,1,balanced,0.1887718391418457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,8,1,balanced,0.18886720657348632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,8,1,balanced,0.1913862419128418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,8,1,balanced,0.19286079406738282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,8,1,balanced,0.19463487625122072
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,8,1,balanced,0.23430591583251953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,4,2,power_law_1.2,0.5726911926269531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,8,1,balanced,0.2361881637573242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,4,2,power_law_1.2,0.7822092437744141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,8,1,balanced,0.2452544021606445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,8,1,balanced,0.19867328643798826
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,8,1,balanced,0.25577215194702146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,8,1,balanced,0.27023744583129883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,8,1,balanced,0.2882387161254883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,8,1,balanced,0.3257196807861328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,8,1,balanced,0.36487808227539065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,8,1,balanced,0.5515814590454101
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,8,1,balanced,0.6293068695068359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,8,1,balanced,0.9055712127685547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,8,1,balanced,1.1831903839111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,4,2,power_law_1.2,1.291068115234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,4,2,power_law_1.2,1.8026573181152343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,8,1,balanced,0.22730432510375978
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,8,1,balanced,0.06067840099334717
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,8,1,balanced,0.061223678588867195
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,8,1,balanced,0.06444992065429687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,8,1,balanced,0.07493887901306152
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,8,1,balanced,0.1308191967010498
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,8,1,balanced,0.09344511985778808
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,8,1,balanced,0.13166208267211915
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,8,1,balanced,0.13054335594177247
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,8,1,balanced,0.13284992218017577
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,8,1,balanced,0.1296345615386963
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,8,1,balanced,0.13275456428527832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,8,1,balanced,0.13516672134399416
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,8,1,balanced,0.1335923194885254
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,8,1,balanced,0.1371712017059326
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,8,1,balanced,0.13745152473449707
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,8,1,balanced,0.14059967994689943
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,8,1,balanced,0.15110272407531739
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,8,1,balanced,0.16125312805175782
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,8,1,balanced,0.1720038414001465
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,8,1,balanced,0.14531264305114747
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,8,1,balanced,0.19560895919799806
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,8,1,balanced,0.2208332824707031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,8,1,power_law_1.01,0.032829439640045165
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,8,1,balanced,0.12966591835021973
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,8,1,power_law_1.01,0.03582272052764893
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,8,1,power_law_1.01,0.05671872138977051
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,8,1,power_law_1.01,0.09758591651916504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,8,1,power_law_1.01,0.12311296463012696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,8,1,power_law_1.01,0.17037439346313477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,8,1,power_law_1.01,0.17816640853881835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,8,1,power_law_1.01,0.18275711059570313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,8,1,power_law_1.01,0.18264192581176758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,8,1,power_law_1.01,0.1887334442138672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,8,1,power_law_1.01,0.19129152297973634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,8,1,power_law_1.01,0.1941516876220703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,8,1,power_law_1.01,0.20427391052246097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,8,1,power_law_1.01,0.24922367095947268
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,8,1,power_law_1.01,0.25280319213867186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,8,1,power_law_1.01,0.26196352005004886
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,8,1,power_law_1.01,0.27783103942871096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,8,1,power_law_1.01,0.30358015060424803
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,8,1,power_law_1.01,0.3617139053344727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,8,1,power_law_1.01,0.4011801528930664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,8,1,power_law_1.01,0.4915161514282227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,8,1,power_law_1.01,0.5950976181030273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,8,1,power_law_1.01,0.7960371398925782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,8,1,power_law_1.01,1.002270736694336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,8,1,power_law_1.01,1.4057664489746093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,8,1,power_law_1.01,0.03654400110244751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,8,1,power_law_1.01,0.04101439952850342
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,8,1,power_law_1.01,1.8163699340820312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,8,1,power_law_1.01,0.04968319892883301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,8,1,power_law_1.01,0.07137216091156005
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,8,1,power_law_1.01,0.09808256149291991
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,8,1,power_law_1.01,0.13407232284545897
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,8,1,power_law_1.01,0.15166784286499024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,8,1,power_law_1.01,0.15645695686340333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,8,1,power_law_1.01,0.1621049690246582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,8,1,power_law_1.01,0.163832950592041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,8,1,power_law_1.01,0.17364288330078126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,8,1,power_law_1.01,0.1768275260925293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,8,1,power_law_1.01,0.17968191146850584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,8,1,power_law_1.01,0.06973440170288085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,8,1,power_law_1.01,0.18319103240966797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,8,1,power_law_1.01,0.21388416290283202
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,8,1,power_law_1.01,0.2266860771179199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,8,1,power_law_1.01,0.21151296615600587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,8,1,power_law_1.01,0.23648576736450194
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,8,1,power_law_1.01,0.2511987113952637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,8,1,power_law_1.01,0.2780556869506836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,8,1,power_law_1.01,0.3155948829650879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,8,1,power_law_1.01,0.3796793746948242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,8,1,power_law_1.01,0.44660671234130855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,8,1,power_law_1.01,0.5816569519042969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,8,1,power_law_1.01,0.7306444549560547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,8,1,power_law_1.01,1.0010054779052733
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,8,1,balanced,0.3746611022949219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,8,1,balanced,0.3261196899414062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,8,1,power_law_1.01,1.2789491271972655
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,8,1,balanced,0.685591049194336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,8,1,power_law_1.01,0.059936637878417964
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,8,1,power_law_1.01,0.06762368202209472
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,8,1,power_law_1.01,0.07584191799163818
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,8,1,power_law_1.01,0.08509568214416505
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,8,1,balanced,0.5293849563598633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,8,1,power_law_1.01,0.09443008422851562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,8,1,power_law_1.01,0.11082176208496093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,8,1,power_law_1.01,0.11661184310913086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,8,1,power_law_1.01,0.11394816398620607
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,8,1,power_law_1.01,0.11805952072143555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,8,1,power_law_1.01,0.12216511726379395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,8,1,power_law_1.01,0.12561408042907715
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,8,1,power_law_1.01,0.12826751708984374
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,8,1,power_law_1.01,0.13261823654174804
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,8,1,power_law_1.01,0.13785856246948242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,8,1,power_law_1.01,0.14251968383789063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,8,1,power_law_1.01,0.15205247879028322
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,8,1,power_law_1.01,0.1720863914489746
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,8,1,power_law_1.01,0.1950124740600586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,8,1,power_law_1.01,0.23948671340942385
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,8,1,power_law_1.01,0.2858195114135742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,8,1,power_law_1.01,0.38742977142333984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,8,1,power_law_1.2,0.029615359306335447
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,8,1,power_law_1.01,0.4959193420410156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,8,1,power_law_1.01,0.06041024208068848
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,8,1,power_law_1.01,0.7004640197753906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,8,1,power_law_1.2,0.055739521980285645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,8,1,power_law_1.2,0.07141056060791015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,8,1,power_law_1.01,0.9252422332763672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,8,1,power_law_1.2,0.11452351570129396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,8,1,power_law_1.2,0.16861248016357422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,8,1,power_law_1.2,0.17731840133666993
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,8,1,power_law_1.2,0.03406464099884034
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,8,1,power_law_1.2,0.17926271438598634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,8,1,power_law_1.2,0.18537792205810547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,8,1,power_law_1.2,0.1880614471435547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,8,1,power_law_1.2,0.1898067283630371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,8,1,power_law_1.2,0.1953228759765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,8,1,power_law_1.2,0.20938623428344724
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,8,1,power_law_1.2,0.2501427268981934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,8,1,power_law_1.2,0.2541145515441895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,8,1,power_law_1.2,0.2635935974121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,8,1,power_law_1.2,0.2829542350769043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,8,1,power_law_1.2,0.3114387130737305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,8,1,power_law_1.2,0.3666681671142578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,8,1,power_law_1.2,0.404692497253418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,8,1,power_law_1.2,0.5056326293945312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,8,1,power_law_1.2,0.6037740707397461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,8,1,power_law_1.2,0.8402041625976562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,8,1,power_law_1.2,0.03450815916061402
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,8,1,power_law_1.2,0.040655999183654784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,8,1,power_law_1.2,1.0124339294433593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,8,1,power_law_1.2,0.05078911781311035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,8,1,power_law_1.2,0.06698880195617676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,8,1,power_law_1.2,0.08834624290466309
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,8,1,power_law_1.2,0.1263468837738037
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,8,1,power_law_1.2,1.4192402648925782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,8,1,power_law_1.2,0.14903167724609373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,8,1,power_law_1.2,0.15493632316589356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,8,1,power_law_1.2,0.15800000190734864
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,8,1,power_law_1.2,0.16535680770874023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,8,1,power_law_1.2,1.8294618225097659
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,8,1,power_law_1.2,0.17079744338989258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,8,1,power_law_1.2,0.17561344146728514
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,8,1,power_law_1.2,0.1793721580505371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,8,1,power_law_1.2,0.18565759658813477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,8,1,power_law_1.2,0.2183238410949707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,8,1,power_law_1.2,0.22357503890991212
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,8,1,power_law_1.2,0.22802879333496096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,8,1,power_law_1.2,0.2436479949951172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,8,1,power_law_1.2,0.25863616943359374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,8,1,power_law_1.2,0.2940313529968262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,8,1,power_law_1.2,0.32391551971435545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,8,1,power_law_1.2,0.39634368896484373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,8,1,power_law_1.2,0.4674873733520508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,8,1,power_law_1.2,0.5900147247314453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,8,1,power_law_1.2,0.7415513610839843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,8,1,power_law_1.2,1.0162963104248046
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,8,1,power_law_1.2,0.0599187183380127
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,8,1,power_law_1.2,0.06113408088684082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,8,1,power_law_1.2,1.2949375915527344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,8,1,power_law_1.2,0.06765503883361816
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,8,1,power_law_1.2,0.08261823654174805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,8,1,power_law_1.2,0.0748524808883667
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,8,1,power_law_1.2,0.09292736053466796
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,8,1,power_law_1.2,0.11003456115722657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,8,1,power_law_1.2,0.11295104026794434
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,8,1,power_law_1.2,0.11541631698608398
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,8,1,power_law_1.2,0.12282239913940429
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,8,1,power_law_1.2,0.12716480255126955
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,8,1,power_law_1.2,0.13042752265930174
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,8,1,power_law_1.2,0.11951168060302735
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,8,1,power_law_1.2,0.13765503883361815
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,8,1,power_law_1.2,0.14529408454895018
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,8,1,power_law_1.2,0.15175935745239258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,8,1,power_law_1.2,0.0847494411468506
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,8,1,power_law_1.2,0.23453247070312502
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,8,1,power_law_1.01,1.3963244628906248
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,8,1,power_law_1.01,1.8216685485839843
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,8,1,power_law_1.2,0.1667251205444336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,8,1,power_law_1.2,0.1991187286376953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,8,1,power_law_1.2,0.30793664932250975
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,8,1,power_law_1.2,0.38646846771240234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,8,1,power_law_1.2,0.5563596725463867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,8,1,power_law_1.2,0.7470387268066406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,8,balanced,0.03619328022003174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,8,balanced,0.02576128005981445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,8,balanced,0.03354752063751221
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,8,balanced,0.06081791877746582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,8,balanced,0.10028927803039552
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,8,1,power_law_1.2,1.1374342346191406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,8,1,power_law_1.2,1.5778067016601562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,8,balanced,0.08130559921264649
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,8,balanced,0.10373120307922364
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,8,balanced,0.1096012783050537
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,8,balanced,0.12402688026428224
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,8,balanced,0.1256563186645508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,8,balanced,0.13317888259887695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,8,balanced,0.13040639877319335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,8,balanced,0.13505791664123534
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,8,balanced,0.1803366470336914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,8,balanced,0.18874496459960938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,8,balanced,0.16965375900268556
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,8,balanced,0.2190015983581543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,8,balanced,0.24614656448364255
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,8,balanced,0.11077759742736817
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,8,balanced,0.270963191986084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,8,balanced,0.4020108795166015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,8,balanced,0.11846272468566894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,8,balanced,0.45551361083984376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,8,balanced,0.6041279983520508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,8,balanced,0.13225600242614746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,8,balanced,0.7032294464111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,8,balanced,0.1351679992675781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,8,1,power_law_1.2,2.4904237365722657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,8,1,power_law_1.2,3.497628173828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,1,8,balanced,0.02191231966018677
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,1,8,balanced,0.021512959003448486
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,1,8,balanced,0.04849408149719238
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,1,8,balanced,0.04847871780395508
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,1,8,balanced,0.021073920726776125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,1,8,balanced,0.04855936050415039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,1,8,balanced,0.04902400016784668
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,1,8,balanced,0.04890751838684082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,1,8,balanced,0.04925183773040771
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,1,8,balanced,0.04924160003662109
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,1,8,balanced,0.049697279930114746
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,1,8,balanced,0.04948863983154297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,1,8,balanced,0.050472960472106934
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,1,8,balanced,0.05062784194946289
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,1,8,balanced,0.052122879028320315
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,1,8,balanced,0.05532800197601319
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,1,8,balanced,0.02641024112701416
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,1,8,balanced,0.03942528009414673
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,1,8,balanced,0.048071680068969724
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,1,8,balanced,0.06051328182220459
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,1,8,balanced,0.06043392181396484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,8,power_law_1.01,0.046498560905456544
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,8,power_law_1.01,0.07331071853637695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,8,power_law_1.01,0.05877247810363769
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,8,power_law_1.01,0.07789440155029297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,1,8,balanced,0.07415423870086671
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,1,8,balanced,0.10414079666137695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,1,8,balanced,0.07028351783752441
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,1,8,balanced,0.15251711845397947
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,1,8,balanced,0.09612544059753418
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,1,8,balanced,0.18500736236572266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,8,power_law_1.01,0.0880947208404541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,8,power_law_1.01,0.09542143821716309
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,8,power_law_1.01,0.12812159538269044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,8,power_law_1.01,0.12703871726989746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,8,power_law_1.01,0.12603263854980468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,8,power_law_1.01,0.12785280227661133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,8,power_law_1.01,0.08143744468688965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,8,power_law_1.01,0.17140863418579103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,8,power_law_1.01,0.10634623527526857
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,8,power_law_1.01,0.18347904205322266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,8,power_law_1.01,0.21482240676879885
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,8,power_law_1.01,0.29401727676391604
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,8,power_law_1.01,0.3563379287719727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,8,power_law_1.01,0.4666278457641601
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,8,power_law_1.01,0.11348095893859864
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,8,power_law_1.01,0.5538278579711914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,1,8,power_law_1.01,0.0258406400680542
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,1,8,power_law_1.01,0.03877887964248657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,8,power_law_1.01,0.8270706939697267
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,1,8,power_law_1.01,0.03275264024734497
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,1,8,power_law_1.01,0.03988607883453369
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,8,power_law_1.01,1.0228339385986327
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,1,8,power_law_1.01,0.03948031902313233
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,1,8,power_law_1.01,0.040952320098876956
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,1,8,power_law_1.01,0.04369664192199707
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,1,8,power_law_1.01,0.042987518310546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,1,8,power_law_1.01,0.04278656005859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,1,8,power_law_1.01,0.04283904075622559
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,1,8,power_law_1.01,0.04555903911590577
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,1,8,power_law_1.01,0.045447678565979005
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,1,8,power_law_1.01,0.04505472183227539
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,1,8,power_law_1.01,0.04587903976440429
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,1,8,power_law_1.01,0.046913280487060546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,1,8,power_law_1.01,0.04318336009979248
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,1,8,power_law_1.01,0.053109760284423824
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,1,8,power_law_1.01,0.05714431762695312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,8,power_law_1.01,0.10112383842468262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,8,power_law_1.01,0.10237183570861816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,8,power_law_1.01,0.1227302360534668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,8,power_law_1.01,0.15245183944702148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,8,power_law_1.01,0.2513484764099121
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,1,8,power_law_1.01,0.04892032146453858
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,1,8,power_law_1.01,0.0665011215209961
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,1,8,power_law_1.01,0.10075008392333984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,1,8,power_law_1.01,0.08226816177368164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,1,8,power_law_1.01,0.12826751708984374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,8,power_law_1.2,0.054711041450500486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,8,power_law_1.2,0.046581759452819824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,8,power_law_1.2,0.06533376216888429
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,8,power_law_1.2,0.08446847915649414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,1,8,power_law_1.01,0.18204032897949218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,8,power_law_1.2,0.09541760444641113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,8,power_law_1.2,0.1014681625366211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,8,power_law_1.2,0.11651071548461914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,8,power_law_1.2,0.1219328022003174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,8,power_law_1.2,0.11789440155029296
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,1,8,power_law_1.01,0.22351615905761718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,8,power_law_1.2,0.12967552185058592
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,8,power_law_1.2,0.07481599807739257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,8,power_law_1.2,0.13344639778137207
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,1,8,power_law_1.01,0.36033153533935547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,8,power_law_1.2,0.1656755256652832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,8,power_law_1.2,0.21461248397827148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,8,power_law_1.2,0.12910079956054688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,8,power_law_1.2,0.2553766441345215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,8,power_law_1.2,0.09331199645996094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,8,power_law_1.2,0.31235456466674805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,8,power_law_1.2,0.3734643173217774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,8,power_law_1.2,0.08440192222595215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,8,power_law_1.2,0.49693054199218756
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,8,power_law_1.2,0.9368819427490234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,1,8,power_law_1.2,0.025992960929870607
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,1,8,power_law_1.2,0.031111679077148437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,1,8,power_law_1.2,0.03536639928817749
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,1,8,power_law_1.2,0.03880064010620117
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,1,8,power_law_1.2,0.03993855953216553
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,1,8,power_law_1.2,0.0415718412399292
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,1,8,power_law_1.2,0.03999871969223022
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,1,8,power_law_1.2,0.04263552188873291
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,1,8,power_law_1.2,0.04307328224182129
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,1,8,power_law_1.2,0.0429798412322998
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,1,8,power_law_1.2,0.04468095779418945
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,1,8,power_law_1.2,0.046175999641418455
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,1,8,power_law_1.2,0.046647038459777836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,8,power_law_1.2,1.1607180786132811
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,8,power_law_1.2,0.6942809295654297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,1,8,power_law_1.2,0.04811647891998291
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,1,8,power_law_1.2,0.04953343868255615
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,1,8,power_law_1.2,0.05237247943878174
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,1,8,power_law_1.2,0.05679103851318359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,8,power_law_1.2,0.10785535812377929
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,1,8,power_law_1.2,0.046419200897216795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,8,power_law_1.2,0.17314048767089846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,8,power_law_1.2,0.18122751235961915
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,1,8,power_law_1.01,0.42303104400634767
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,1,8,power_law_1.2,0.06415616035461426
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,1,8,power_law_1.2,0.07573887825012207
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,1,8,power_law_1.2,0.09664511680603027
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,1,8,power_law_1.2,0.12710911750793458
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,1,8,power_law_1.2,0.16631935119628907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,4,balanced,0.023787519931793212
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,4,balanced,0.03376255989074707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,4,balanced,0.05454463958740234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,4,balanced,0.024800000190734865
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,1,8,power_law_1.2,0.2435980796813965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,4,balanced,0.07785600185394287
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,1,8,power_law_1.2,0.3581017684936524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,4,balanced,0.12078207969665526
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,4,balanced,0.1358937644958496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,4,balanced,0.12015744209289551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,4,balanced,0.1232102394104004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,4,balanced,0.09980544090270996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,4,balanced,0.10403327941894532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,4,balanced,0.12059904098510743
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,4,balanced,0.13280511856079102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,4,balanced,0.14378879547119142
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,4,balanced,0.18555776596069334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,4,balanced,0.17199871063232422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,4,balanced,0.19034879684448242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,1,8,power_law_1.2,0.5308902359008789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,4,balanced,0.27068031311035157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,4,balanced,0.10698240280151368
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,4,balanced,0.27984128952026366
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,4,balanced,0.10913151741027831
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,4,balanced,0.4253235244750977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,4,balanced,0.4840422439575195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,4,balanced,0.6947647857666015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,4,balanced,0.8092454528808594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,4,balanced,0.12250368118286133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,4,balanced,0.12895999908447267
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,1,8,power_law_1.2,0.6878758239746093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,2,4,balanced,0.023114240169525145
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,2,4,balanced,0.02337791919708252
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,2,4,balanced,0.05170559883117676
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,2,4,balanced,0.0512883186340332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,2,4,balanced,0.051637759208679204
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,2,4,balanced,0.023973119258880616
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,2,4,balanced,0.05162879943847656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,2,4,balanced,0.05212160110473633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,2,4,balanced,0.05236224174499512
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,2,4,balanced,0.05251200199127197
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,2,4,balanced,0.052607998847961426
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,2,4,balanced,0.05371391773223877
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,2,4,balanced,0.054737920761108394
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,2,4,balanced,0.030105600357055663
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,2,4,balanced,0.043056640624999996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,2,4,balanced,0.05674496173858643
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,2,4,balanced,0.059412479400634766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,2,4,balanced,0.062256641387939446
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,2,4,balanced,0.05134848117828369
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,2,4,balanced,0.05113215923309326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,4,power_law_1.01,0.03047424077987671
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,4,power_law_1.01,0.045608959197998046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,4,power_law_1.01,0.045134081840515136
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,4,power_law_1.01,0.060568318367004395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,4,power_law_1.01,0.06401408195495606
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,2,4,balanced,0.05159679889678955
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,2,4,balanced,0.0925004768371582
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,2,4,balanced,0.12324607849121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,2,4,balanced,0.08717951774597169
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,2,4,balanced,0.20105600357055664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,2,4,balanced,0.1331340789794922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,2,4,balanced,0.24712192535400393
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,4,power_law_1.01,0.11437184333801269
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,4,power_law_1.01,0.12155136108398437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,4,power_law_1.01,0.11633664131164552
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,4,power_law_1.01,0.10006655693054198
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,4,power_law_1.01,0.10898176193237305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,4,power_law_1.01,0.12080256462097168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,4,power_law_1.01,0.11385472297668457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,4,power_law_1.01,0.1838003158569336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,4,power_law_1.01,0.09800448417663574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,4,power_law_1.01,0.08712960243225097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,4,power_law_1.01,0.09370368003845216
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,4,power_law_1.01,0.12999936103820803
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,4,power_law_1.01,0.4525145721435547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,4,power_law_1.01,0.7664192199707032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,4,power_law_1.01,0.18847103118896485
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,2,4,power_law_1.01,0.02400383949279785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,4,power_law_1.01,0.27422847747802737
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,2,4,power_law_1.01,0.02964992046356201
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,2,4,power_law_1.01,0.031174399852752686
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,2,4,power_law_1.01,0.037794559001922606
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,2,4,power_law_1.01,0.04371327877044678
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,4,power_law_1.01,1.0086540985107422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,2,4,power_law_1.01,0.048890881538391114
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,2,4,power_law_1.01,0.04633344173431396
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,2,4,power_law_1.01,0.046661119461059566
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,2,4,power_law_1.01,0.04866559982299805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,2,4,power_law_1.01,0.04814208030700683
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,4,power_law_1.01,0.22021760940551757
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,2,4,power_law_1.01,0.049834241867065435
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,2,4,power_law_1.01,0.050462718009948726
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,2,4,power_law_1.01,0.05077760219573975
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,2,4,power_law_1.01,0.050969600677490234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,2,4,power_law_1.01,0.053013758659362795
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,2,4,power_law_1.01,0.0536678409576416
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,2,4,power_law_1.01,0.0570636796951294
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,2,4,power_law_1.01,0.07700479984283447
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,4,power_law_1.01,0.11606399536132812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,4,power_law_1.01,0.1404428768157959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,4,power_law_1.01,0.3510988616943359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,4,power_law_1.01,0.5836479949951172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,2,4,power_law_1.01,0.060266242027282715
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,2,4,power_law_1.01,0.06939263820648194
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,2,4,power_law_1.01,0.12441599845886231
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,4,power_law_1.2,0.029726719856262206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,4,power_law_1.2,0.045505280494689944
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,2,4,power_law_1.01,0.16412160873413087
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,2,4,power_law_1.01,0.22218368530273436
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,2,4,power_law_1.01,0.31864063262939457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,4,power_law_1.2,0.09124863624572754
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,2,4,power_law_1.01,0.09317376136779784
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,2,4,power_law_1.01,0.4271321487426758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,4,power_law_1.2,0.1130726432800293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,4,power_law_1.2,0.11369471549987793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,4,power_law_1.2,0.12828031539916993
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,4,power_law_1.2,0.11571840286254882
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,4,power_law_1.2,0.039184639453887934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,4,power_law_1.2,0.11608448028564453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,4,power_law_1.2,0.12334976196289063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,4,power_law_1.2,0.13140735626220704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,4,power_law_1.2,0.15053440093994142
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,4,power_law_1.2,0.18372224807739257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,4,power_law_1.2,0.09098752021789551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,4,power_law_1.2,0.19497472763061524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,4,power_law_1.2,0.23024383544921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,4,power_law_1.2,0.28544384002685547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,4,power_law_1.2,0.05911935806274414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,4,power_law_1.2,0.09595647811889649
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,4,power_law_1.2,0.06436863899230957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,4,power_law_1.2,0.4853849411010742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,2,4,power_law_1.2,0.024348158836364747
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,4,power_law_1.2,0.6059199905395507
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,2,4,power_law_1.2,0.02999808073043823
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,2,4,power_law_1.2,0.0313753604888916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,4,power_law_1.2,0.8311628723144532
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,2,4,power_law_1.2,0.03662591934204101
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,2,4,power_law_1.2,0.041845760345458984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,2,4,power_law_1.2,0.04561535835266113
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,2,4,power_law_1.2,0.04729216098785401
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,2,4,power_law_1.2,0.04611328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,4,power_law_1.2,0.09917056083679199
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,2,4,power_law_1.2,0.05018879890441894
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,2,4,power_law_1.2,0.050186238288879394
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,2,4,power_law_1.2,0.04960512161254883
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,2,4,power_law_1.2,0.05148928165435791
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,2,4,power_law_1.2,0.05243008136749268
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,2,4,power_law_1.2,0.05475840091705322
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,4,power_law_1.2,0.11812352180480956
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,2,4,power_law_1.2,0.048417282104492185
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,4,power_law_1.2,0.3735193634033203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,4,power_law_1.2,1.0004761505126953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,2,4,power_law_1.01,0.49079551696777346
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,2,4,power_law_1.2,0.06230400085449219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,2,4,power_law_1.2,0.05999231815338135
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,2,4,power_law_1.2,0.055870718955993645
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,2,4,power_law_1.2,0.07276159763336182
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,2,4,power_law_1.2,0.08724736213684083
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,4,2,balanced,0.021968638896942137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,4,2,balanced,0.02849663972854614
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,4,2,balanced,0.03212287902832031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,4,2,balanced,0.05276927947998047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,4,2,balanced,0.0751423978805542
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,4,2,balanced,0.09799039840698243
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,2,4,power_law_1.2,0.11657855987548829
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,2,4,power_law_1.2,0.14072192192077637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,4,2,balanced,0.09938816070556641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,4,2,balanced,0.10128512382507324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,4,2,balanced,0.103121919631958
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,4,2,balanced,0.10610431671142577
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,4,2,balanced,0.10715007781982422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,4,2,balanced,0.11827456474304199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,4,2,balanced,0.11587712287902832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,4,2,balanced,0.10009728431701662
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,4,2,balanced,0.11747072219848634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,4,2,balanced,0.12510080337524415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,4,2,balanced,0.1349286365509033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,4,2,balanced,0.15032575607299806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,4,2,balanced,0.20240127563476565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,4,2,balanced,0.10187904357910156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,4,2,balanced,0.21639936447143554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,4,2,balanced,0.27379711151123043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,4,2,balanced,0.26868736267089843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,4,2,balanced,0.43285503387451174
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,2,4,power_law_1.2,0.19422592163085936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,4,2,balanced,0.4861056137084961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,4,2,balanced,0.6970956420898438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,4,2,power_law_1.01,0.029203200340270997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,4,2,power_law_1.01,0.03976703882217407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,4,2,balanced,0.8026496124267577
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,4,2,power_law_1.01,0.06204671859741211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,4,2,power_law_1.01,0.07744639873504638
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,4,2,power_law_1.01,0.022566399574279784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,4,2,power_law_1.01,0.08674559593200684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,4,2,power_law_1.01,0.09212287902832031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,4,2,power_law_1.01,0.0972441577911377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,4,2,power_law_1.01,0.10181119918823242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,4,2,power_law_1.01,0.11251328468322755
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,4,2,power_law_1.01,0.11793408393859864
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,4,2,power_law_1.01,0.1276595211029053
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,4,2,power_law_1.01,0.13837823867797852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,4,2,power_law_1.01,0.13049599647521973
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,4,2,power_law_1.01,0.08755840301513672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,4,2,power_law_1.01,0.053396477699279785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,4,2,power_law_1.01,0.1508505630493164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,4,2,power_law_1.01,0.21145984649658206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,4,2,power_law_1.01,0.25798400878906247
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,4,2,power_law_1.01,0.2768460845947266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,4,2,power_law_1.01,0.34153472900390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,4,2,power_law_1.01,0.44861568450927736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,4,2,power_law_1.01,0.5441164779663086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,4,2,power_law_1.01,0.7375897979736328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,4,2,power_law_1.01,0.9437081909179688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,4,2,power_law_1.2,0.03081984043121338
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,4,2,power_law_1.2,0.04009215831756592
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,4,2,power_law_1.2,0.021909759044647217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,4,2,power_law_1.01,0.08995455741882324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,4,2,power_law_1.2,0.05022463798522949
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,4,2,power_law_1.2,0.08173824310302734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,4,2,power_law_1.2,0.08560768127441407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,4,2,power_law_1.2,0.08849023818969727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,4,2,power_law_1.2,0.09168255805969239
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,4,2,power_law_1.2,0.09224575996398926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,4,2,power_law_1.2,0.09884544372558593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,4,2,power_law_1.2,0.1158348846435547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,4,2,power_law_1.2,0.1074073600769043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,4,2,power_law_1.2,0.1277888011932373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,4,2,power_law_1.2,0.13446271896362305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,4,2,power_law_1.2,0.14293248176574708
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,4,2,power_law_1.2,0.14164352416992188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,4,2,power_law_1.2,0.15523712158203123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,4,2,power_law_1.2,0.17549055099487304
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,2,4,power_law_1.2,0.8320422363281249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,4,2,power_law_1.2,0.23713407516479493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,4,2,power_law_1.2,0.2644377517700195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,4,2,power_law_1.2,0.2913254356384277
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,4,2,power_law_1.2,0.35049728393554686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,4,2,power_law_1.2,0.4615347290039063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,4,2,power_law_1.2,0.05822080135345459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,4,2,power_law_1.2,0.564142074584961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,4,2,power_law_1.2,0.7459878540039062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,4,2,power_law_1.01,0.1608639907836914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,8,1,balanced,0.0674451208114624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,8,1,balanced,0.02959359884262085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,8,1,balanced,0.03960383892059326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,8,1,balanced,0.11820799827575683
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,2,4,power_law_1.2,0.5853043365478515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,8,1,balanced,0.1973734474182129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,8,1,balanced,0.2587871932983398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,8,1,balanced,0.26213760375976564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,8,1,balanced,0.27301631927490233
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,4,2,power_law_1.2,0.9602060699462891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,8,1,balanced,0.27536319732666015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,8,1,balanced,0.2652729606628418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,8,1,balanced,0.2650419235229492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,8,1,balanced,0.2663667106628418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,8,1,balanced,0.27014272689819335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,8,1,balanced,0.2809382438659668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,8,1,balanced,0.27581439971923827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,8,1,balanced,0.432674560546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,8,1,balanced,0.2417228889465332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,8,1,balanced,0.4286336135864258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,8,1,balanced,0.32576961517333985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,8,1,balanced,0.38598270416259767
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,8,1,balanced,0.3715769577026367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,8,1,balanced,0.6647430419921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,8,1,balanced,0.6905817413330079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,8,1,balanced,0.9913152313232422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,8,1,power_law_1.01,0.07245567798614502
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,8,1,power_law_1.01,0.038614399433135986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,8,1,power_law_1.01,0.030460801124572758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,8,1,power_law_1.01,0.2032512092590332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,8,1,power_law_1.01,0.1059539222717285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,8,1,balanced,1.0686630249023437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,8,1,power_law_1.01,0.21283008575439455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,8,1,balanced,1.662408905029297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,8,1,power_law_1.01,0.14054400444030762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,8,1,power_law_1.01,0.22241920471191406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,8,1,balanced,2.0297715759277346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,8,1,power_law_1.01,0.24060224533081054
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,8,1,power_law_1.01,0.25747840881347656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,8,1,power_law_1.01,0.3513177490234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,8,1,power_law_1.01,0.23341503143310546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,8,1,power_law_1.01,0.29495744705200194
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,8,1,power_law_1.01,0.3800275039672852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,8,1,power_law_1.01,0.4644204711914062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,8,1,power_law_1.01,0.27487167358398434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,8,1,power_law_1.01,0.5164972686767578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,8,1,power_law_1.01,0.6277920150756836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,8,1,power_law_1.01,0.7433849334716797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,8,1,power_law_1.01,0.32281471252441407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,8,1,power_law_1.01,0.9903129577636719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,8,1,power_law_1.01,0.31913984298706055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,8,1,power_law_1.01,0.39104766845703126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,8,1,power_law_1.2,0.030823678970336915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,8,1,power_law_1.01,1.2250016021728516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,8,1,power_law_1.2,0.04224512100219727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,8,1,power_law_1.01,1.6957221984863282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,8,1,power_law_1.01,2.1881619262695313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,8,1,power_law_1.2,0.2018284797668457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,8,1,power_law_1.2,0.21571392059326172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,8,1,power_law_1.01,0.4192716979980469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,8,1,power_law_1.2,0.23724992752075194
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,8,1,power_law_1.2,0.2270528030395508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,8,1,power_law_1.2,0.24744064331054688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,8,1,power_law_1.2,0.07238336086273194
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,8,1,power_law_1.2,0.26498367309570314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,8,1,power_law_1.2,0.2876044845581055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,8,1,power_law_1.2,0.12823424339294434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,8,1,power_law_1.2,0.09854911804199219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,8,1,power_law_1.2,0.3027084732055664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,8,1,power_law_1.2,0.3340281677246094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,8,1,power_law_1.2,0.32728511810302735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,8,1,power_law_1.2,0.38978431701660154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,8,1,power_law_1.2,0.42469566345214843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,8,1,power_law_1.2,0.528862075805664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,8,1,power_law_1.2,0.6476416015625001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,8,1,power_law_1.2,0.7531263732910156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,8,1,power_law_1.2,0.35911041259765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,8,1,power_law_1.2,1.000872344970703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,8,1,power_law_1.2,1.2363410949707032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,8,balanced,0.09339391708374023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,8,1,power_law_1.2,0.3967782211303711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,8,balanced,0.07056128025054932
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,8,1,power_law_1.2,0.47482559204101565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,8,1,power_law_1.2,1.7167456054687498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,8,balanced,0.1506713581085205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,8,balanced,0.28618879318237306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,8,balanced,0.6902681732177735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,8,balanced,0.7127065277099609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,8,balanced,0.5041740798950196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,8,balanced,0.6485401916503906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,8,1,power_law_1.2,2.2017114257812502
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,8,balanced,0.6905484771728515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,8,balanced,0.7002444458007813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,8,balanced,0.8278540802001952
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,8,balanced,0.7440230560302734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,8,balanced,0.8221695709228516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,8,balanced,0.8378278350830077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,8,balanced,1.069543685913086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,8,balanced,0.9595967864990234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,8,balanced,1.1198079681396484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,8,balanced,1.29486083984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,8,balanced,0.7659763336181641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,8,balanced,1.3212249755859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,8,balanced,0.7732428741455079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,8,balanced,1.4862860107421876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,8,balanced,1.773218536376953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,8,balanced,0.046913280487060546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,8,balanced,0.07490816116333007
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,8,balanced,0.1091648006439209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,8,balanced,0.182740478515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,8,balanced,0.29193088531494144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,8,balanced,2.6190362548828126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,8,balanced,0.3718361663818359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,8,balanced,0.3864614486694336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,8,balanced,2.9249688720703126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,8,balanced,0.3941721725463867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,8,balanced,0.41985534667968755
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,8,balanced,0.41860095977783207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,8,balanced,0.42351745605468755
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,8,balanced,0.4422732925415039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,8,balanced,0.45777919769287106
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,8,balanced,0.49291519165039066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,8,balanced,0.46082176208496095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,8,balanced,0.4737612915039063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,8,balanced,4.289388732910156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,8,balanced,0.5120486450195313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,8,balanced,0.5646579360961914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,8,balanced,0.6085887908935547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,8,balanced,0.7513433837890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,8,balanced,4.935566101074219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,8,balanced,0.860126724243164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,8,balanced,1.2277452850341797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,8,balanced,1.444427490234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,8,balanced,2.0626853942871093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,8,balanced,2.4946726989746093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,8,balanced,3.7702822875976567
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,8,balanced,4.826169738769531
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,1,8,balanced,0.032175359725952146
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,1,8,balanced,0.22877056121826173
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,1,8,balanced,0.04807040214538574
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,1,8,balanced,0.2288435173034668
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,1,8,balanced,0.06607999801635742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,1,8,balanced,0.22868736267089843
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,1,8,balanced,0.2295129585266113
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,1,8,balanced,0.2302720069885254
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,1,8,balanced,0.23110912322998045
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,1,8,balanced,0.23344640731811522
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,1,8,balanced,0.23565952301025392
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,1,8,balanced,0.2374553680419922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,1,8,balanced,0.24941055297851564
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,1,8,balanced,0.25917184829711915
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,1,8,balanced,0.2696102333068848
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,1,8,balanced,0.17599231719970704
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,1,8,balanced,0.11053440093994141
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,1,8,balanced,0.24318336486816405
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,1,8,balanced,0.2282969665527344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,1,8,balanced,0.22666751861572268
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,1,8,balanced,0.41549057006835943
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,1,8,balanced,0.39052288055419926
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,1,8,balanced,0.5735475158691405
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,1,8,balanced,0.6187571334838867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,1,8,balanced,1.0140697479248046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,8,power_law_1.01,0.5964223861694335
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,1,8,balanced,1.2237619018554686
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,1,8,balanced,0.22957056045532226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,8,power_law_1.01,0.37395969390869144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,8,power_law_1.01,0.4345830535888672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,8,power_law_1.01,0.49405441284179685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,8,power_law_1.01,0.5569228744506836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,8,power_law_1.01,0.612619514465332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,8,power_law_1.01,0.5952384185791015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,8,power_law_1.01,0.6898751831054687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,8,power_law_1.01,0.2526028823852539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,8,power_law_1.01,0.7174208068847656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,8,power_law_1.01,1.053470687866211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,8,power_law_1.01,0.60967041015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,8,power_law_1.01,0.7682176208496093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,8,power_law_1.01,0.6410380554199219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,8,power_law_1.01,0.7379622650146485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,8,power_law_1.01,0.7016512298583985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,8,power_law_1.01,0.7536716461181641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,8,power_law_1.01,1.841144256591797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,8,power_law_1.01,0.9688972473144531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,8,power_law_1.01,2.7185073852539063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,8,power_law_1.01,0.15858304023742675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,8,power_law_1.01,0.2640473556518555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,8,power_law_1.01,0.9750272369384765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,8,power_law_1.01,0.23923967361450194
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,8,power_law_1.01,0.3351884841918945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,8,power_law_1.01,0.3565004730224609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,8,power_law_1.01,1.3088064575195313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,8,power_law_1.01,0.3324236679077148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,8,power_law_1.01,0.3513600158691406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,8,power_law_1.01,0.3754406356811523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,8,power_law_1.01,0.3689228820800781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,8,power_law_1.01,0.36327041625976564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,8,power_law_1.01,0.38898433685302736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,8,power_law_1.01,0.4074995040893555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,8,power_law_1.01,0.28140031814575195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,8,power_law_1.01,5.8601611328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,8,power_law_1.01,0.4544063949584961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,8,power_law_1.01,0.4291648101806641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,8,power_law_1.01,2.123252410888672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,8,power_law_1.01,0.5688780975341797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,8,power_law_1.01,0.6269120025634766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,8,power_law_1.01,0.9242431640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,8,power_law_1.01,1.250814743041992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,8,power_law_1.01,3.513370971679687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,8,power_law_1.01,1.5412800598144532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,8,power_law_1.01,2.169550018310547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,8,power_law_1.01,4.849688415527344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,1,8,power_law_1.01,0.10003968238830567
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,8,power_law_1.01,1.4213848876953126
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,1,8,power_law_1.01,0.16687231063842772
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,1,8,power_law_1.01,0.14640895843505858
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,1,8,power_law_1.01,0.16336000442504883
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,1,8,power_law_1.01,0.20640384674072268
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,1,8,power_law_1.01,0.22750080108642576
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,1,8,power_law_1.01,0.1936332893371582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,8,power_law_1.01,4.027765808105469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,1,8,power_law_1.01,0.19893503189086914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,1,8,power_law_1.01,0.19481983184814453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,8,power_law_1.01,0.44089473724365236
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,1,8,power_law_1.01,0.19888639450073242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,1,8,power_law_1.01,0.20603519439697265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,8,power_law_1.01,0.4995699310302735
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,1,8,power_law_1.01,0.20691968917846681
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,1,8,power_law_1.01,0.21737087249755857
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,1,8,power_law_1.01,0.2220672035217285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,1,8,power_law_1.01,0.22263679504394532
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,1,8,power_law_1.01,0.23197952270507813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,8,power_law_1.01,5.196124267578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,8,power_law_1.01,0.7815309143066406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,8,power_law_1.01,2.8856536865234377
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,1,8,power_law_1.01,0.29274879455566405
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,1,8,power_law_1.01,0.35936000823974606
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,1,8,power_law_1.01,0.2722777557373047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,1,8,power_law_1.01,0.24481023788452147
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,1,8,power_law_1.01,0.41794815063476565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,8,power_law_1.2,0.2565171241760254
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,1,8,power_law_1.01,0.5612083053588867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,1,8,power_law_1.01,0.6902579498291016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,8,power_law_1.2,0.39159168243408204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,8,power_law_1.2,0.46330623626708983
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,1,8,power_law_1.01,0.9515750122070312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,8,power_law_1.2,0.586187515258789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,8,power_law_1.2,0.568279037475586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,8,power_law_1.2,0.6509696197509766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,8,power_law_1.2,0.3034111976623535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,8,power_law_1.2,0.6236991882324219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,8,power_law_1.2,0.597154541015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,8,power_law_1.2,0.7809420776367187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,8,power_law_1.2,0.7826624298095703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,8,power_law_1.2,0.62279296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,8,power_law_1.2,0.6932851409912109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,8,power_law_1.2,0.8352102661132813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,8,power_law_1.2,0.7354534149169922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,8,power_law_1.2,1.0427507019042968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,8,power_law_1.2,1.0313817596435546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,8,power_law_1.2,1.146654739379883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,8,power_law_1.2,1.4275392150878905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,8,power_law_1.2,1.513623046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,8,power_law_1.2,1.9735577392578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,8,power_law_1.2,0.7179865264892579
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,8,power_law_1.2,2.2761689758300783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,8,power_law_1.2,2.9032626342773438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,8,power_law_1.2,0.15690496444702148
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,1,8,power_law_1.01,1.137624282836914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,1,8,power_law_1.01,1.8513395690917966
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,1,8,power_law_1.01,2.2899737548828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,8,power_law_1.2,0.29532159805297853
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,8,power_law_1.2,0.33187198638916016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,8,power_law_1.2,3.802534484863281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,8,power_law_1.2,0.32276607513427735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,8,power_law_1.2,0.38159999847412107
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,8,power_law_1.2,0.3800934219360351
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,8,power_law_1.2,0.36491008758544924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,8,power_law_1.2,6.2964672851562495
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,8,power_law_1.2,0.4006399917602539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,8,power_law_1.2,0.4158822250366211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,8,power_law_1.2,0.4540531158447266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,8,power_law_1.2,0.47822208404541017
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,8,power_law_1.2,0.47096958160400393
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,8,power_law_1.2,0.5291174316406251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,8,power_law_1.2,0.5924492645263671
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,8,power_law_1.2,0.19591808319091797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,8,power_law_1.2,0.656478729248047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,8,power_law_1.2,0.8950950622558593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,8,power_law_1.2,0.9636914825439453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,8,power_law_1.2,5.21994384765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,8,power_law_1.2,1.2979315185546876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,8,power_law_1.2,0.24370176315307618
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,1,8,power_law_1.2,0.09899264335632324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,8,power_law_1.2,1.6247167968749998
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,1,8,power_law_1.2,0.1274227237701416
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,1,8,power_law_1.2,0.15065088272094726
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,1,8,power_law_1.2,0.17084287643432616
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,8,power_law_1.2,2.235764465332031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,1,8,power_law_1.2,0.21254655838012698
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,8,power_law_1.2,0.3433023834228516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,1,8,power_law_1.2,0.2280166435241699
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,1,8,power_law_1.2,0.191014404296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,1,8,power_law_1.2,0.1892620849609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,8,power_law_1.2,2.9259136962890624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,1,8,power_law_1.2,0.2066431999206543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,1,8,power_law_1.2,0.19327999114990235
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,1,8,power_law_1.2,0.20868608474731448
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,1,8,power_law_1.2,0.21329792022705077
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,1,8,power_law_1.2,0.2183091163635254
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,1,8,power_law_1.2,0.23156864166259766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,1,8,power_law_1.2,0.2585318374633789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,1,8,power_law_1.2,0.24001024246215819
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,8,power_law_1.2,0.32953983306884765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,8,power_law_1.2,4.396502990722657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,8,power_law_1.2,5.771981811523437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,1,8,power_law_1.2,0.30168447494506834
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,1,8,power_law_1.2,0.24935808181762695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,1,8,power_law_1.2,0.3295180892944336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,1,8,power_law_1.2,0.42718593597412113
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,1,8,power_law_1.2,0.5323302459716797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,1,8,power_law_1.2,0.7067609405517578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,4,balanced,0.0639296007156372
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,4,balanced,0.08964480400085449
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,4,balanced,0.272300796508789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,4,balanced,0.1505894374847412
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,4,balanced,0.46696319580078127
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,1,8,power_law_1.2,0.8639654541015626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,4,balanced,0.650376968383789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,4,balanced,0.6554611206054688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,4,balanced,0.662339859008789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,4,balanced,0.6735321807861328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,4,balanced,0.7402867126464844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,4,balanced,0.7389055633544922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,1,8,power_law_1.2,1.204422378540039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,4,balanced,0.7523391723632813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,4,balanced,0.7601881408691407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,4,balanced,0.8389145660400391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,4,balanced,0.7753651428222657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,4,balanced,0.8814015960693359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,4,balanced,0.996864013671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,4,balanced,1.1631475067138672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,4,balanced,1.0954598236083986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,4,balanced,0.7910527801513672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,4,balanced,1.4426138305664062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,4,balanced,1.607841339111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,4,balanced,0.7830131530761719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,4,balanced,2.6102822875976566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,4,balanced,0.06475647926330566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,4,balanced,0.10452992439270019
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,4,balanced,2.859239807128906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,4,balanced,0.1793382453918457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,4,balanced,0.29821823120117186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,4,balanced,0.38293632507324216
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,4,balanced,0.3947916793823242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,4,balanced,0.4040089416503906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,4,balanced,4.151288452148437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,4,balanced,0.41515647888183593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,4,balanced,0.4200396728515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,4,balanced,0.4306777572631836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,4,balanced,4.622494812011719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,4,balanced,0.46657150268554687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,4,balanced,0.03982079982757568
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,4,balanced,0.4735756683349609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,4,balanced,0.48768894195556645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,4,balanced,0.5176025772094727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,4,balanced,0.5355763244628906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,4,balanced,0.5793344116210937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,1,8,power_law_1.2,1.7097637939453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,4,balanced,0.6174591827392578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,4,balanced,0.6932466888427735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,4,balanced,0.7770278167724609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,4,balanced,0.4826163101196289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,4,balanced,1.154287338256836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,4,balanced,1.3085440063476563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,1,8,power_law_1.2,2.673950805664062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,4,balanced,1.907519989013672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,4,balanced,2.2514073181152345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,4,balanced,3.402086486816406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,4,balanced,4.300657958984376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,1,8,power_law_1.2,3.547484130859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,2,4,balanced,0.04310656070709228
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,2,4,balanced,0.05404672145843505
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,2,4,balanced,0.24456703186035159
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,2,4,balanced,0.24458112716674804
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,2,4,balanced,0.0734003210067749
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,2,4,balanced,0.24672256469726564
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,2,4,balanced,0.24717056274414065
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,2,4,balanced,0.24749183654785156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,2,4,balanced,0.25033472061157225
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,2,4,balanced,0.11042176246643068
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,2,4,balanced,0.2508441543579102
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,2,4,balanced,0.25464319229125976
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,2,4,balanced,0.2578201675415039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,2,4,balanced,0.2655526351928711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,2,4,balanced,0.1797145652770996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,2,4,balanced,0.2740339279174805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,2,4,balanced,0.29060352325439454
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,2,4,balanced,0.30419967651367186
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,2,4,balanced,0.2426304054260254
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,2,4,balanced,0.24398719787597659
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,2,4,balanced,0.24310783386230467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,4,power_law_1.01,0.14815360069274902
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,4,power_law_1.01,0.24912895202636717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,4,power_law_1.01,0.331383056640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,4,power_law_1.01,0.2541900825500488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,4,power_law_1.01,0.38637184143066405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,4,power_law_1.01,0.5535859298706055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,4,power_law_1.01,0.5393215942382812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,4,power_law_1.01,0.5781836700439453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,4,power_law_1.01,0.5899251174926758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,4,power_law_1.01,0.6677887725830078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,4,power_law_1.01,0.6818739318847656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,4,power_law_1.01,0.6952524566650391
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,2,4,balanced,0.6911270141601562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,2,4,balanced,0.4858240127563477
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,2,4,balanced,0.44995582580566407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,2,4,balanced,0.7470054626464844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,4,power_law_1.01,0.7742195129394531
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,2,4,balanced,1.2210009765624998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,4,power_law_1.01,0.8265433502197265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,4,power_law_1.01,0.8817984008789063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,4,power_law_1.01,1.153264617919922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,4,power_law_1.01,1.1118374633789063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,2,4,balanced,1.581958465576172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,4,power_law_1.01,1.2759820556640624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,4,power_law_1.01,0.679156494140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,4,power_law_1.01,0.7203890991210937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,4,power_law_1.01,1.7083584594726564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,4,power_law_1.01,1.9886912536621093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,4,power_law_1.01,0.09630592346191406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,4,power_law_1.01,0.15920384407043459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,4,power_law_1.01,2.5068800354003904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,4,power_law_1.01,0.16508928298950196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,4,power_law_1.01,0.2110655975341797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,4,power_law_1.01,0.2630054473876953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,4,power_law_1.01,0.7293452453613282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,4,power_law_1.01,3.2112423706054685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,4,power_law_1.01,0.3435558319091797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,4,power_law_1.01,0.3642406463623047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,4,power_law_1.01,0.38019073486328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,4,power_law_1.01,0.4328895950317383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,4,power_law_1.01,4.223411254882812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,4,power_law_1.01,0.480871696472168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,4,power_law_1.01,0.4432038497924805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,4,power_law_1.01,0.4611366271972656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,4,power_law_1.01,0.568798713684082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,4,power_law_1.01,5.29525390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,4,power_law_1.01,0.43928192138671873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,4,power_law_1.01,0.6166169738769531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,4,power_law_1.01,0.7569049835205078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,4,power_law_1.01,0.876153564453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,4,power_law_1.01,0.3316864013671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,4,power_law_1.01,1.1515865325927734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,4,power_law_1.01,0.38948993682861327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,4,power_law_1.01,1.3863885498046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,2,4,power_law_1.01,0.07594624042510986
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,2,4,power_law_1.01,0.11102592468261718
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,2,4,power_law_1.01,0.10992511749267578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,4,power_law_1.01,1.9859890747070312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,2,4,power_law_1.01,0.13230719566345214
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,2,4,power_law_1.01,0.1642252731323242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,2,4,power_law_1.01,0.2286732864379883
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,2,4,power_law_1.01,0.21795711517333985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,4,power_law_1.01,2.492991943359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,2,4,power_law_1.01,0.21367679595947267
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,2,4,power_law_1.01,0.2130892753601074
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,2,4,power_law_1.01,0.21556735992431642
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,4,power_law_1.01,0.4974476623535156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,2,4,power_law_1.01,0.22105600357055666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,2,4,power_law_1.01,0.22672767639160157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,2,4,power_law_1.01,0.23412096023559573
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,2,4,power_law_1.01,0.2389299201965332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,2,4,power_law_1.01,0.24519168853759768
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,2,4,power_law_1.01,0.25064064025878907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,4,power_law_1.01,0.3444518280029297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,2,4,power_law_1.01,0.2665894317626953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,4,power_law_1.01,3.616834411621094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,2,4,power_law_1.01,0.30161920547485355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,4,power_law_1.01,4.746768493652344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,2,4,power_law_1.01,0.4157056045532227
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,2,4,power_law_1.01,0.49441280364990237
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,2,4,power_law_1.01,0.3372159957885742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,2,4,power_law_1.01,0.6484979248046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,4,power_law_1.2,0.14630784034729002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,4,power_law_1.2,0.198472957611084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,4,power_law_1.2,0.3181119918823242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,2,4,power_law_1.01,1.1627852630615234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,2,4,power_law_1.01,1.5847756958007813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,4,power_law_1.2,0.39252864837646484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,4,power_law_1.2,0.5156991958618165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,4,power_law_1.2,0.5887104034423828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,4,power_law_1.2,0.6700991821289063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,4,power_law_1.2,0.6589376068115234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,4,power_law_1.2,0.7015846252441407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,4,power_law_1.2,0.5438822555541992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,4,power_law_1.2,0.24650239944458008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,4,power_law_1.2,0.7364710235595704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,4,power_law_1.2,0.5821836853027345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,4,power_law_1.2,0.83583740234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,4,power_law_1.2,0.7285619354248046
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,2,4,power_law_1.01,2.4135090637207033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,4,power_law_1.2,0.9183692932128906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,4,power_law_1.2,1.1919872283935546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,4,power_law_1.2,1.1566655731201172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,4,power_law_1.2,0.7225075531005859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,4,power_law_1.2,1.3204185485839843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,4,power_law_1.2,1.8346829223632812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,4,power_law_1.2,2.1110809326171873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,4,power_law_1.2,0.09524991989135742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,4,power_law_1.2,0.12304767608642578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,4,power_law_1.2,2.704919128417969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,4,power_law_1.2,0.16466943740844725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,4,power_law_1.2,3.4082611083984373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,4,power_law_1.2,0.21093376159667968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,4,power_law_1.2,0.2605273628234863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,4,power_law_1.2,0.34134143829345703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,4,power_law_1.2,0.31736576080322265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,4,power_law_1.2,0.3407219314575195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,4,power_law_1.2,0.36433151245117185
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,4,power_law_1.2,0.3722150421142578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,4,power_law_1.2,0.39104766845703126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,4,power_law_1.2,0.43240703582763673
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,4,power_law_1.2,0.4395072174072266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,4,power_law_1.2,0.4920281600952149
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,4,power_law_1.2,5.764835815429687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,4,power_law_1.2,0.46175487518310543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,4,power_law_1.2,0.752195816040039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,4,power_law_1.2,0.5273561477661133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,4,power_law_1.2,0.5811289596557617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,4,power_law_1.2,0.6450150299072266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,4,power_law_1.2,0.7660006713867188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,4,power_law_1.2,0.9023974609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,4,power_law_1.2,1.175077133178711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,4,power_law_1.2,1.4325978088378908
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,4,power_law_1.2,2.0163877868652342
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,4,power_law_1.2,4.630320739746094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,2,4,power_law_1.2,0.07696767807006835
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,2,4,power_law_1.2,0.09052543640136719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,4,power_law_1.2,2.6082366943359374
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,2,4,power_law_1.2,0.11196800231933594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,2,4,power_law_1.2,0.12762240409851072
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,2,4,power_law_1.2,0.1643596839904785
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,2,4,power_law_1.2,0.21996543884277342
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,2,4,power_law_1.2,0.2214259147644043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,2,4,power_law_1.2,0.21220352172851564
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,2,4,power_law_1.2,0.21981695175170896
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,2,4,power_law_1.2,0.2176857566833496
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,2,4,power_law_1.2,0.22718975067138675
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,2,4,power_law_1.2,0.23252735137939454
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,2,4,power_law_1.2,0.2381235122680664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,2,4,power_law_1.2,0.25401727676391606
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,2,4,power_law_1.2,0.25992063522338865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,4,power_law_1.2,0.47255294799804687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,2,4,power_law_1.01,2.954437255859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,4,power_law_1.2,3.7844415283203126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,4,power_law_1.2,4.983558349609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,2,4,power_law_1.2,0.3338572692871094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,2,4,power_law_1.2,0.2743040084838867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,2,4,power_law_1.2,0.29894912719726563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,2,4,power_law_1.2,0.49408512115478515
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,2,4,power_law_1.2,0.3851596832275391
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,2,4,power_law_1.2,0.5964083099365235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,4,2,balanced,0.05732863903045654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,4,2,balanced,0.09131391525268555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,2,4,power_law_1.2,0.8383168029785157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,4,2,balanced,0.16994688034057617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,4,2,balanced,0.24673152923583985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,4,2,balanced,0.46518783569335936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,4,2,balanced,0.6544793701171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,4,2,balanced,0.6495500946044922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,4,2,balanced,0.6621990203857422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,4,2,balanced,0.6649190521240234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,4,2,balanced,0.6905433654785156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,2,4,power_law_1.2,1.1127462768554688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,4,2,balanced,0.6762368011474609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,4,2,balanced,0.687509765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,2,4,power_law_1.2,1.8215052795410156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,4,2,balanced,0.695367660522461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,2,4,power_law_1.2,2.106394805908203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,4,2,balanced,0.7956928253173828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,4,2,balanced,0.8465958404541016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,2,4,power_law_1.2,3.45846923828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,4,2,balanced,0.7850994873046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,4,2,balanced,0.7640422058105469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,4,2,balanced,0.772938232421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,4,2,balanced,1.382617645263672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,4,2,balanced,1.740807647705078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,4,2,balanced,0.998888931274414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,2,4,power_law_1.2,4.80094970703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,4,2,balanced,1.082117156982422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,4,2,balanced,0.06687232017517089
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,4,2,balanced,1.9017485046386717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,4,2,balanced,0.10620032310485841
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,4,2,balanced,0.19714303970336916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,4,2,balanced,0.32096382141113283
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,4,2,balanced,0.4386943817138672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,4,2,balanced,0.4451443099975586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,4,2,balanced,0.0451251220703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,4,2,balanced,0.4509043121337891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,4,2,balanced,2.7153060913085936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,4,2,balanced,0.45885951995849605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,4,2,balanced,0.46146686553955074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,4,2,balanced,0.46979328155517575
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,4,2,balanced,0.490967025756836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,4,2,balanced,0.4904447937011719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,4,2,balanced,0.5090214538574218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,4,2,balanced,2.9971301269531248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,4,2,balanced,0.5538880157470703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,4,2,balanced,0.5771353530883789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,4,2,balanced,0.5775078582763672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,4,2,balanced,0.6329433441162109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,4,2,balanced,0.6498188781738281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,4,2,balanced,0.7318271636962891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,4,2,balanced,0.7980735778808594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,4,2,balanced,4.558796691894531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,4,2,balanced,1.243088607788086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,4,2,balanced,1.3850445556640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,4,2,balanced,1.9773516845703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,4,2,balanced,2.2707212829589842
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,4,2,balanced,5.623567504882812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,4,2,balanced,3.468628540039062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,4,2,balanced,4.328688659667969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,4,2,balanced,0.06380032062530518
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,4,2,balanced,0.07407872200012207
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,4,2,balanced,0.29231103897094723
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,4,2,balanced,0.29130111694335936
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,4,2,balanced,0.2956185531616211
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,4,2,balanced,0.09180800437927246
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,4,2,balanced,0.2939289665222168
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,4,2,balanced,0.2953996849060059
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,4,2,balanced,0.29915008544921873
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,4,2,balanced,0.30162687301635743
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,4,2,balanced,0.12889087677001954
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,4,2,balanced,0.3055398368835449
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,4,2,balanced,0.3121958351135254
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,4,2,balanced,0.3244044876098633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,4,2,balanced,0.3363404846191406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,4,2,balanced,0.3590118408203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,4,2,balanced,0.3822515106201172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,4,2,balanced,0.2108723258972168
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,4,2,balanced,0.2879052734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,4,2,balanced,0.28986368179321287
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,4,2,power_law_1.01,0.0824294376373291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,4,2,power_law_1.01,0.136494083404541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,4,2,power_law_1.01,0.1958732795715332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,4,2,power_law_1.01,0.25443328857421876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,4,2,balanced,0.6653094482421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,4,2,balanced,0.6114905548095704
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,4,2,balanced,0.9899980926513672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,4,2,power_law_1.01,0.49893630981445314
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,4,2,balanced,1.0961293029785157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,4,2,power_law_1.01,0.567083511352539
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,4,2,balanced,1.7702297973632812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,4,2,power_law_1.01,0.3508377456665039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,4,2,power_law_1.01,0.6211084747314454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,4,2,power_law_1.01,0.645643539428711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,4,2,power_law_1.01,0.6771481323242188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,4,2,balanced,2.168695068359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,4,2,power_law_1.01,0.7511949157714843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,4,2,power_law_1.01,0.7604441833496094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,4,2,power_law_1.01,0.8016140747070313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,4,2,power_law_1.01,0.5381990432739258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,4,2,power_law_1.01,1.1913330841064453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,4,2,power_law_1.01,0.5468979263305664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,4,2,power_law_1.01,0.6124415969848632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,4,2,power_law_1.01,1.4905381774902344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,4,2,power_law_1.01,0.770982437133789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,4,2,power_law_1.01,1.6751091003417968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,4,2,power_law_1.01,0.858387222290039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,4,2,power_law_1.01,0.06490367889404297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,4,2,power_law_1.01,2.6220800781249998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,4,2,power_law_1.01,0.10006015777587891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,4,2,power_law_1.01,1.9918310546874998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,4,2,power_law_1.01,0.18975872039794922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,4,2,power_law_1.01,0.25273088455200193
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,4,2,power_law_1.01,0.3660019302368164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,4,2,power_law_1.01,4.635083618164062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,4,2,power_law_1.01,0.379736328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,4,2,power_law_1.01,0.3873625564575195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,4,2,power_law_1.01,1.0340838623046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,4,2,power_law_1.01,0.4072499084472656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,4,2,power_law_1.01,0.4156441497802734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,4,2,power_law_1.01,0.42303615570068354
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,4,2,power_law_1.01,5.847171630859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,4,2,power_law_1.01,0.4375449752807617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,4,2,power_law_1.01,0.44869758605957033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,4,2,power_law_1.01,0.4704435348510742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,4,2,power_law_1.01,0.5133644866943359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,4,2,power_law_1.01,0.5364966583251953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,4,2,power_law_1.01,0.5475328063964844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,4,2,power_law_1.01,0.6210457611083984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,4,2,power_law_1.01,0.6673369598388672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,4,2,power_law_1.01,0.8014297485351562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,4,2,power_law_1.01,0.9179647827148438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,4,2,power_law_1.01,0.13130623817443848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,4,2,power_law_1.01,1.1783372497558593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,4,2,power_law_1.01,1.4256947326660157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,4,2,power_law_1.01,0.07988736152648926
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,4,2,power_law_1.01,0.1023423957824707
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,4,2,power_law_1.01,0.10912128448486329
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,4,2,power_law_1.01,1.92574462890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,4,2,power_law_1.01,0.1344601631164551
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,4,2,power_law_1.01,0.16737024307250975
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,4,2,power_law_1.01,0.24014463424682617
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,4,2,power_law_1.01,0.24639616012573243
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,4,2,power_law_1.01,0.2546700859069824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,4,2,power_law_1.01,2.4349554443359374
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,4,2,power_law_1.01,0.2643456077575684
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,4,2,power_law_1.01,0.27015424728393556
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,4,2,power_law_1.01,0.27918848037719723
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,4,2,power_law_1.01,0.28074752807617187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,4,2,power_law_1.01,0.293176326751709
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,4,2,power_law_1.01,0.3040755271911621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,4,2,power_law_1.01,3.543997497558594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,4,2,power_law_1.01,0.31351680755615235
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,4,2,power_law_1.01,0.3852326583862305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,4,2,power_law_1.01,3.258069763183594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,4,2,power_law_1.01,4.500769348144532
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,4,2,power_law_1.01,0.2587276840209961
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,4,2,power_law_1.01,0.3393049621582031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,4,2,power_law_1.01,0.43215103149414064
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,4,2,power_law_1.01,0.5437977600097657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,4,2,power_law_1.01,0.935357437133789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,4,2,power_law_1.01,1.141333770751953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,4,2,power_law_1.01,0.700285415649414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,4,2,power_law_1.2,0.11624064445495605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,4,2,power_law_1.2,0.08354304313659669
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,4,2,power_law_1.01,1.6499110412597655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,4,2,power_law_1.2,0.19779199600219727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,4,2,power_law_1.2,0.244267520904541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,4,2,power_law_1.2,0.5281740951538085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,4,2,power_law_1.2,0.5416742324829101
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,4,2,power_law_1.2,0.5728947067260742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,4,2,power_law_1.2,0.6089036941528321
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,4,2,power_law_1.2,0.3335808181762695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,4,2,power_law_1.2,0.6601113891601562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,4,2,power_law_1.2,0.7008806610107422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,4,2,power_law_1.2,0.6283622360229492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,4,2,power_law_1.2,0.7682508850097657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,4,2,power_law_1.2,0.5156108856201171
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,4,2,power_law_1.2,0.7885247802734374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,4,2,power_law_1.2,0.8368307495117188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,4,2,power_law_1.2,0.9114649963378906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,4,2,power_law_1.2,1.2658099365234374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,4,2,power_law_1.2,0.8317849731445313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,4,2,power_law_1.2,1.6846438598632811
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,4,2,power_law_1.2,2.0280435180664065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,4,2,power_law_1.2,2.7585650634765626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,4,2,power_law_1.2,1.0961023712158204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,4,2,power_law_1.2,3.505135498046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,4,2,power_law_1.2,1.5625050354003907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,4,2,power_law_1.2,0.09389183998107911
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,4,2,power_law_1.2,0.06320000171661377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,4,2,power_law_1.2,0.13237119674682618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,4,2,power_law_1.2,0.18451328277587892
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,4,2,power_law_1.2,0.24239103317260743
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,4,2,power_law_1.2,0.3655487823486328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,4,2,power_law_1.2,0.3755545425415039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,4,2,power_law_1.2,0.38932735443115235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,4,2,power_law_1.2,0.40835582733154296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,4,2,power_law_1.2,0.41513599395751954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,4,2,power_law_1.2,0.4227468872070313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,4,2,power_law_1.2,6.06740478515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,4,2,power_law_1.01,2.066754608154297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,4,2,power_law_1.2,0.4455680084228516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,4,2,power_law_1.2,0.45441665649414065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,4,2,power_law_1.2,0.4932608032226563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,4,2,power_law_1.01,3.061829223632812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,4,2,power_law_1.2,0.5343475341796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,4,2,power_law_1.2,0.5566336059570313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,4,2,power_law_1.2,0.583185920715332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,4,2,power_law_1.2,0.6501350402832031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,4,2,power_law_1.01,4.224936828613282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,4,2,power_law_1.2,0.6865715026855469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,4,2,power_law_1.2,0.8139968109130858
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,4,2,power_law_1.2,0.9369229125976564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,4,2,power_law_1.2,1.2183795166015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,4,2,power_law_1.2,1.4789234924316408
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,4,2,power_law_1.2,0.07974783897399902
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,4,2,power_law_1.2,0.09093888282775879
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,4,2,power_law_1.2,0.10919551849365235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,4,2,power_law_1.2,4.85309814453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,4,2,power_law_1.2,0.13255807876586914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,4,2,power_law_1.2,1.9720831298828123
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,4,2,power_law_1.2,0.15975680351257324
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,4,2,power_law_1.2,0.22946559906005862
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,4,2,power_law_1.2,0.24768640518188478
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,4,2,power_law_1.2,2.494263000488281
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,4,2,power_law_1.2,0.2505983924865723
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,4,2,power_law_1.2,0.2606732749938965
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,4,2,power_law_1.2,0.2738188743591309
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,4,2,power_law_1.2,0.2903615951538086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,4,2,power_law_1.2,0.3092428779602051
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,4,2,power_law_1.2,0.3242675018310547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,4,2,power_law_1.2,0.26427263259887696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,4,2,power_law_1.2,3.646757202148437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,4,2,power_law_1.2,4.614517822265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,4,2,power_law_1.2,0.33815807342529297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,4,2,power_law_1.2,0.36590206146240234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,4,2,power_law_1.2,0.44196735382080077
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,4,2,power_law_1.2,0.5061312103271485
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,4,2,power_law_1.2,0.6584819030761719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,4,2,power_law_1.2,0.7910489654541015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,8,1,balanced,0.07432960033416748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,8,1,balanced,0.10208127975463868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,8,1,balanced,0.15826111793518066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,8,1,balanced,0.2694495964050293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,4,2,power_law_1.2,1.5577740478515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,4,2,power_law_1.2,1.190414047241211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,8,1,balanced,0.5160204696655273
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,4,2,power_law_1.2,2.384241943359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,8,1,balanced,0.7227327728271484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,4,2,power_law_1.2,3.088057556152344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,8,1,balanced,0.7169427490234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,8,1,balanced,0.6997081756591796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,8,1,balanced,0.7054950714111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,4,2,power_law_1.2,5.085180053710937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,8,1,balanced,0.7052178955078124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,8,1,balanced,0.7337862396240235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,8,1,balanced,0.8667430114746093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,8,1,balanced,0.7342636871337891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,8,1,balanced,0.7316134643554688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,8,1,balanced,0.9575635528564452
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,4,2,power_law_1.2,6.76912109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,8,1,balanced,0.7555436706542968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,8,1,balanced,0.740414047241211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,8,1,balanced,0.8621638488769531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,8,1,balanced,1.3004934692382812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,8,1,balanced,1.4542930603027344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,8,1,balanced,1.2503302764892577
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,8,1,balanced,0.06226816177368164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,8,1,balanced,1.9641081237792968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,8,1,balanced,0.08732352256774903
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,8,1,balanced,0.13693440437316895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,8,1,balanced,0.2264384078979492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,8,1,balanced,0.5715264129638672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,8,1,balanced,0.5799641418457031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,8,1,balanced,0.5837030410766602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,8,1,balanced,0.5963219070434571
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,8,1,balanced,1.8631602478027343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,8,1,balanced,0.5980460739135742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,8,1,balanced,0.6070713424682618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,8,1,balanced,0.618487663269043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,8,1,balanced,3.032695007324219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,8,1,balanced,0.6248051071166992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,8,1,balanced,0.6417894744873047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,8,1,balanced,0.6256038284301757
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,8,1,balanced,0.6337932968139649
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,8,1,balanced,0.6530879974365235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,8,1,balanced,3.369816284179687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,8,1,balanced,0.7895391845703126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,8,1,balanced,0.39970558166503906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,8,1,balanced,0.8676934051513673
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,8,1,balanced,0.761994857788086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,8,1,balanced,0.938884506225586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,8,1,balanced,1.4853926086425782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,8,1,balanced,1.6233351135253904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,8,1,balanced,4.880631713867187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,8,1,balanced,2.3426054382324217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,8,1,balanced,5.54221923828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,8,1,balanced,2.629974365234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,8,1,power_law_1.01,0.10467776298522949
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,8,1,power_law_1.01,0.3544454574584961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,8,1,balanced,4.118761596679688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,8,1,power_law_1.01,0.07459712028503418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,8,1,power_law_1.01,0.6115929412841796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,8,1,power_law_1.01,0.1590559959411621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,8,1,balanced,5.075662231445312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,8,1,power_law_1.01,0.5822137451171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,8,1,power_law_1.01,0.6786156463623046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,8,1,power_law_1.01,0.5632179260253907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,8,1,power_law_1.01,0.5946694564819336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,8,1,power_law_1.01,0.7157881927490235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,8,1,power_law_1.01,0.7849951934814453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,8,1,power_law_1.01,0.7968863677978516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,8,1,power_law_1.01,0.8849407958984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,8,1,power_law_1.01,1.012257308959961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,8,1,power_law_1.01,1.0713235473632812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,8,1,power_law_1.01,0.6304127883911133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,8,1,power_law_1.01,0.6462854766845704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,8,1,power_law_1.01,1.4657919311523437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,8,1,power_law_1.01,1.804505615234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,8,1,power_law_1.01,1.2477069091796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,8,1,power_law_1.01,0.256615047454834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,8,1,power_law_1.01,1.2734118652343749
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,8,1,power_law_1.01,3.461453552246094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,8,1,power_law_1.01,0.13546624183654785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,8,1,power_law_1.01,0.08822848320007323
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,8,1,power_law_1.01,0.0598476791381836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,8,1,power_law_1.01,0.3017753601074219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,8,1,power_law_1.01,4.6294183349609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,8,1,power_law_1.01,0.4616902542114258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,8,1,power_law_1.01,0.48733760833740236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,8,1,power_law_1.01,0.5050592041015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,8,1,power_law_1.01,0.5229164886474609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,8,1,power_law_1.01,0.5299129486083984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,8,1,power_law_1.01,5.8583398437500005
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,8,1,power_law_1.01,0.5714144134521485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,8,1,power_law_1.01,0.5994854354858399
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,8,1,power_law_1.01,0.5412799835205078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,8,1,power_law_1.01,0.5930841445922852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,8,1,power_law_1.01,0.6163910293579102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,8,1,power_law_1.01,0.6544876861572265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,8,1,power_law_1.01,0.7570022583007813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,8,1,power_law_1.01,0.8660934448242188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,8,1,power_law_1.01,0.9522541046142579
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,8,1,power_law_1.01,1.0863148498535158
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,8,1,power_law_1.01,0.21460544586181643
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,8,1,power_law_1.01,2.2483897399902344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,8,1,power_law_1.01,1.3786636352539063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,8,1,power_law_1.01,2.8344384765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,8,1,power_law_1.2,0.07182528018951416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,8,1,power_law_1.01,1.6521856689453123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,8,1,power_law_1.01,2.227815704345703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,8,1,power_law_1.01,2.789411315917969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,8,1,power_law_1.2,0.1577791976928711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,8,1,power_law_1.2,0.10215680122375488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,8,1,power_law_1.01,3.957028503417969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,8,1,power_law_1.01,0.5585599899291992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,8,1,power_law_1.2,0.5560633468627929
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,8,1,power_law_1.2,0.5939564895629883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,8,1,power_law_1.2,0.2380441665649414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,8,1,power_law_1.2,0.6589266967773437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,8,1,power_law_1.2,0.3149164772033691
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,8,1,power_law_1.2,0.703614730834961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,8,1,power_law_1.2,0.7254112243652344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,8,1,power_law_1.2,0.6130444717407226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,8,1,power_law_1.2,0.8737651062011718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,8,1,power_law_1.2,0.9203449249267578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,8,1,power_law_1.01,5.105239562988281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,8,1,power_law_1.2,1.0237721252441407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,8,1,power_law_1.2,0.573193588256836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,8,1,power_law_1.2,0.635156478881836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,8,1,power_law_1.2,1.5041651916503906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,8,1,power_law_1.2,1.1194931030273438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,8,1,power_law_1.2,0.8033299255371095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,8,1,power_law_1.2,1.8515782165527344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,8,1,power_law_1.2,2.2962835693359374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,8,1,power_law_1.2,0.0609222412109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,8,1,power_law_1.2,1.3457139587402343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,8,1,power_law_1.2,0.08656895637512207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,8,1,power_law_1.2,0.13431360244750976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,8,1,power_law_1.2,1.2705741119384766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,8,1,power_law_1.2,0.1986252784729004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,8,1,power_law_1.2,0.27702783584594726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,8,1,power_law_1.2,2.8811474609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,8,1,power_law_1.2,0.4512255859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,8,1,power_law_1.2,0.4895987319946289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,8,1,power_law_1.2,0.5009683227539062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,8,1,power_law_1.2,0.5221990585327149
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,8,1,power_law_1.2,0.5353132629394531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,8,1,power_law_1.2,0.5510067367553712
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,8,1,power_law_1.2,0.5709791946411132
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,8,1,power_law_1.2,0.5830227279663086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,8,1,power_law_1.2,0.6177113723754883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,8,1,power_law_1.2,0.6169823837280274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,8,1,power_law_1.2,0.635395851135254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,8,1,power_law_1.2,0.6803526306152344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,8,1,power_law_1.2,0.7935398101806641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,8,1,power_law_1.2,3.479806213378906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,8,1,power_law_1.2,0.850730209350586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,8,1,power_law_1.2,0.983235855102539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,8,1,power_law_1.2,1.1222124481201172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,8,1,power_law_1.2,1.4020396423339843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,8,1,power_law_1.2,1.6849542236328126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,2,8,balanced,0.09495936393737793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,2,8,balanced,0.09501567840576172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,2,8,balanced,0.1706675148010254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,2,8,balanced,0.17035776138305664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,2,8,balanced,0.09483391761779784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,2,8,balanced,0.17033599853515624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,8,1,power_law_1.2,2.257016296386719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,2,8,balanced,0.17052415847778318
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,2,8,balanced,0.1709529685974121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,2,8,balanced,0.17291135787963868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,2,8,balanced,0.17258623123168945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,2,8,balanced,0.17551231384277344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,2,8,balanced,0.17617664337158204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,2,8,balanced,0.17951616287231445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,2,8,balanced,0.18116607666015624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,2,8,balanced,0.18213760375976562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,2,8,balanced,0.19026176452636717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,2,8,balanced,0.21410943984985353
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,2,8,balanced,0.23146879196166995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,2,8,balanced,0.17364351272583006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,2,8,balanced,0.2618099212646484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,2,8,balanced,0.27156864166259764
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,2,8,balanced,0.36927486419677735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,2,8,balanced,0.46512897491455074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,8,1,power_law_1.2,2.832678527832031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,2,8,balanced,0.6573798370361328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,2,8,balanced,0.0529369592666626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,2,8,balanced,0.05334015846252441
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,2,8,balanced,0.8646873474121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,2,8,balanced,0.05275392055511474
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,2,8,balanced,0.0629094409942627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,2,8,balanced,0.060833277702331545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,2,8,balanced,0.06214272022247315
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,2,8,balanced,0.06181119918823243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,2,8,balanced,1.3030975341796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,2,8,balanced,0.06361472129821777
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,2,8,balanced,0.06400383949279785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,2,8,balanced,0.06370431900024413
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,2,8,balanced,0.06330624103546142
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,2,8,balanced,0.0667136001586914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,2,8,balanced,0.06351871967315674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,2,8,balanced,0.06585343837738036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,2,8,balanced,1.7163737487792967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,2,8,balanced,0.06907264232635499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,2,8,balanced,0.07288191795349122
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,8,1,power_law_1.2,4.6555456542968745
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,2,8,balanced,0.08414976119995117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,2,8,balanced,0.08524160385131836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,2,8,balanced,0.10609536170959473
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,2,8,balanced,0.13364992141723633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,2,8,balanced,0.10640512466430666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,2,8,balanced,0.1566220760345459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,2,8,balanced,0.21823488235473634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,2,8,balanced,0.26595455169677734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,2,8,balanced,0.39181438446044925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,2,8,balanced,0.5087807846069337
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,2,8,balanced,0.06857600212097167
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,8,1,power_law_1.2,3.9698962402343754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,8,1,power_law_1.2,5.896130981445312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,8,1,power_law_1.2,5.146951293945312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,2,8,balanced,0.021096959114074706
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,2,8,balanced,0.020787200927734374
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,2,8,balanced,0.021533439159393312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,2,8,balanced,0.021176319122314453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,2,8,balanced,0.021707520484924317
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,2,8,balanced,0.02143872022628784
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,2,8,balanced,0.02187648057937622
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,2,8,balanced,0.021495039463043215
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,2,8,balanced,0.021760001182556152
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,2,8,balanced,0.02215424060821533
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,2,8,balanced,0.022342400550842283
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,2,8,balanced,0.020709118843078612
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,2,8,balanced,0.021105918884277344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,2,8,balanced,0.02103935956954956
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,2,8,balanced,0.021122560501098633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,2,8,balanced,0.021269760131835937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,2,8,balanced,0.021290240287780763
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,2,8,balanced,0.028582398891448972
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,2,8,balanced,0.030488319396972656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,2,8,balanced,0.04286975860595703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,2,8,balanced,0.03707135915756225
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,2,8,balanced,0.058119678497314455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,2,8,power_law_1.01,0.09519359588623047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,2,8,power_law_1.01,0.17093631744384766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,2,8,power_law_1.01,0.17172735214233398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,2,8,power_law_1.01,0.17092607498168946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,2,8,power_law_1.01,0.17373952865600586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,2,8,power_law_1.01,0.17708927154541015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,2,8,power_law_1.01,0.1811903953552246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,2,8,power_law_1.01,0.1824985694885254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,2,8,power_law_1.01,0.18779136657714843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,2,8,power_law_1.01,0.20019584655761719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,2,8,power_law_1.01,0.21519487380981445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,2,8,power_law_1.01,0.22690048217773437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,2,8,power_law_1.01,0.24320640563964843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,2,8,power_law_1.01,0.2610214424133301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,2,8,power_law_1.01,0.167956485748291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,2,8,power_law_1.01,0.2700441551208496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,2,8,power_law_1.01,0.1733964729309082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,2,8,power_law_1.01,0.3488025665283203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,2,8,balanced,0.07084288120269774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,2,8,power_law_1.01,0.4541017532348633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,2,8,power_law_1.01,0.7183257293701172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,2,8,power_law_1.01,0.5976153564453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,2,8,balanced,0.09441791534423828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,2,8,power_law_1.01,1.7044236755371096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,2,8,power_law_1.01,3.097395324707031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,2,8,power_law_1.01,0.05543551921844483
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,2,8,power_law_1.01,2.5673611450195315
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,2,8,power_law_1.01,0.06372608184814453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,2,8,power_law_1.01,0.06072319984436035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,2,8,power_law_1.01,1.1300160217285158
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,2,8,balanced,0.1183743953704834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,2,8,power_law_1.01,0.06342656135559081
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,2,8,power_law_1.01,0.06410367965698242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,2,8,power_law_1.01,0.06595967769622803
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,2,8,power_law_1.01,0.06710015773773194
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,2,8,power_law_1.01,0.06939775943756103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,2,8,power_law_1.01,0.0744051218032837
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,2,8,power_law_1.01,0.07850111961364746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,2,8,power_law_1.01,0.08284928321838378
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,2,8,power_law_1.01,0.05948800086975098
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,2,8,power_law_1.01,0.09363967895507812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,2,8,power_law_1.01,0.10350720405578613
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,2,8,power_law_1.01,0.1049292755126953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,2,8,power_law_1.01,0.14108799934387206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,2,8,power_law_1.01,0.12077952384948731
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,2,8,power_law_1.01,0.25656320571899416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,2,8,power_law_1.01,0.19698047637939453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,2,8,power_law_1.01,0.4287910461425781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,2,8,power_law_1.01,0.06043776035308838
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,2,8,balanced,0.17098112106323243
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,2,8,balanced,0.2162944030761719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,2,8,power_law_1.01,0.35303550720214844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,2,8,power_law_1.01,0.02068095922470093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,2,8,power_law_1.01,0.021086719036102295
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,2,8,power_law_1.01,0.02108288049697876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,2,8,power_law_1.01,0.021012480258941653
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,2,8,power_law_1.01,0.021155838966369626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,2,8,power_law_1.01,0.021111040115356444
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,2,8,power_law_1.01,0.021539840698242187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,2,8,power_law_1.01,0.021262080669403077
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,2,8,power_law_1.01,0.02147968053817749
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,2,8,power_law_1.01,0.021510400772094727
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,2,8,power_law_1.01,0.021960959434509278
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,2,8,power_law_1.01,0.7232358551025391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,2,8,power_law_1.01,0.8907315063476563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,2,8,power_law_1.01,1.828468475341797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,2,8,power_law_1.01,0.06081791877746582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,2,8,power_law_1.01,7.142691650390626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,2,8,power_law_1.01,5.3523095703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,2,8,power_law_1.01,1.4780288696289063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,2,8,power_law_1.01,0.024779520034790038
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,2,8,power_law_1.01,0.028414719104766846
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,2,8,power_law_1.01,0.11386624336242676
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,2,8,power_law_1.01,0.030283520221710204
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,2,8,power_law_1.01,0.034204161167144774
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,2,8,power_law_1.01,0.04229504108428955
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,2,8,power_law_1.01,0.03783423900604248
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,2,8,power_law_1.01,0.05547391891479492
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,2,8,power_law_1.01,0.06618239879608154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,2,8,power_law_1.2,0.09416319847106933
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,2,8,power_law_1.2,0.1711052894592285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,2,8,power_law_1.2,0.1712396812438965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,2,8,power_law_1.2,0.17184255599975587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,2,8,power_law_1.2,0.17125631332397462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,2,8,power_law_1.2,0.17331584930419922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,2,8,power_law_1.2,0.17545343399047852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,2,8,power_law_1.2,0.17807872772216798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,2,8,power_law_1.2,0.18068607330322267
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,2,8,power_law_1.2,0.18525951385498046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,2,8,power_law_1.2,0.19206655502319336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,2,8,power_law_1.2,0.2132966423034668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,2,8,power_law_1.2,0.21287551879882813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,2,8,power_law_1.2,0.23149183273315427
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,2,8,power_law_1.2,0.2432793617248535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,2,8,power_law_1.2,0.2617356872558594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,2,8,power_law_1.2,0.27475839614868164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,2,8,power_law_1.2,0.35353473663330076
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,2,8,power_law_1.01,0.08712448120117187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,2,8,power_law_1.2,0.45905792236328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,2,8,power_law_1.2,0.6313049697875976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,2,8,power_law_1.2,0.7388082885742188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,2,8,power_law_1.2,0.9974566650390624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,2,8,power_law_1.01,0.1577868843078613
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,2,8,power_law_1.01,0.19325439453124998
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,2,8,power_law_1.01,0.29163776397705077
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,2,8,power_law_1.01,0.34796031951904294
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,2,8,power_law_1.2,0.05329919815063476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,2,8,power_law_1.2,0.0597376012802124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,2,8,power_law_1.2,0.06051199913024903
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,2,8,power_law_1.2,0.06077439785003662
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,2,8,power_law_1.2,0.06083072185516357
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,2,8,power_law_1.2,0.06190720081329346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,2,8,power_law_1.2,0.06231423854827881
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,2,8,power_law_1.2,0.06317567825317383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,2,8,power_law_1.2,0.06530687808990479
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,2,8,power_law_1.2,0.06693376064300537
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,2,8,power_law_1.2,1.700701446533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,2,8,power_law_1.2,0.06958208084106446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,2,8,power_law_1.2,0.07835648059844971
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,2,8,power_law_1.2,0.07779712200164794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,2,8,power_law_1.2,0.08425344467163086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,2,8,power_law_1.2,0.09290623664855957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,2,8,power_law_1.2,0.10395903587341308
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,2,8,power_law_1.2,0.10445568084716797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,2,8,power_law_1.2,0.14385536193847656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,2,8,power_law_1.2,0.12093824386596679
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,2,8,power_law_1.2,0.2576793670654297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,2,8,power_law_1.2,0.1954240036010742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,2,8,power_law_1.2,0.38691585540771484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,2,8,power_law_1.2,0.6293939208984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,2,8,power_law_1.2,2.513388824462891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,2,8,power_law_1.2,0.3648819351196289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,2,8,power_law_1.2,3.239517517089844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,2,8,power_law_1.2,0.02191999912261963
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,2,8,power_law_1.2,0.021492478847503663
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,2,8,power_law_1.2,0.021902079582214358
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,2,8,power_law_1.2,0.02152064085006714
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,2,8,power_law_1.2,0.02109312057495117
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,2,8,power_law_1.2,0.02149631977081299
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,2,8,power_law_1.2,0.021217279434204102
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,2,8,power_law_1.2,0.0214899206161499
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,2,8,power_law_1.2,0.021519360542297365
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,2,8,power_law_1.2,0.021928958892822266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,2,8,power_law_1.2,0.021885440349578858
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,2,8,power_law_1.2,0.026382079124450685
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,2,8,power_law_1.2,0.026508800983428955
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,2,8,power_law_1.2,0.02886528015136719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,2,8,power_law_1.2,0.8770777893066406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,2,8,power_law_1.01,0.5773030471801758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,2,8,power_law_1.2,5.663147583007812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,2,8,power_law_1.2,1.771942443847656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,2,8,power_law_1.2,0.06530432224273681
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,2,8,power_law_1.2,1.5241253662109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,2,8,power_law_1.01,0.8857266998291016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,2,8,power_law_1.2,0.03418751955032349
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,2,8,power_law_1.2,0.036686079502105715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,2,8,power_law_1.2,8.666597290039062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,2,8,power_law_1.2,0.04184447765350342
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,2,8,power_law_1.2,0.0555187177658081
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,2,8,power_law_1.2,0.08568448066711426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,4,4,balanced,0.05992832183837891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,4,4,balanced,0.05985663890838623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,4,4,balanced,0.07199488162994386
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,4,4,balanced,0.12074751853942871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,4,4,balanced,0.12018688201904297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,4,4,balanced,0.12126336097717286
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,4,4,balanced,0.12233728408813475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,4,4,balanced,0.12181632041931152
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,4,4,balanced,0.12436479568481444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,4,4,balanced,0.12451328277587889
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,4,4,balanced,0.12486911773681642
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,4,4,balanced,0.12791168212890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,4,4,balanced,0.1294924831390381
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,4,4,balanced,0.13278976440429688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,4,4,balanced,0.13624320030212403
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,4,4,balanced,0.13737600326538085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,4,4,balanced,0.14187904357910158
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,4,4,balanced,0.1773478317260742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,4,4,balanced,0.15934720039367675
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,2,8,power_law_1.2,0.6195622253417968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,4,4,balanced,0.2273753547668457
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,2,8,power_law_1.2,0.11485823631286621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,4,4,balanced,0.28652544021606446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,4,4,balanced,0.3603647994995117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,4,4,balanced,0.4503308868408203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,4,4,balanced,0.6670105743408203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,4,4,balanced,0.8530022430419922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,4,4,balanced,0.03584000110626221
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,4,4,balanced,0.03583103895187378
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,4,4,balanced,0.041880321502685544
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,4,4,balanced,1.2937472534179686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,4,4,balanced,0.04766592025756836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,4,4,balanced,0.05239808082580566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,4,4,balanced,0.048947200775146485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,4,4,balanced,0.05059328079223633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,4,4,balanced,0.04940032005310059
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,4,4,balanced,0.051121921539306635
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,4,4,balanced,0.05017471790313721
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,4,4,balanced,1.6987840270996095
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,2,8,power_law_1.2,0.15431551933288573
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,4,4,balanced,0.05345664024353027
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,4,4,balanced,0.0523967981338501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,4,4,balanced,0.054305281639099125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,4,4,balanced,0.05943295955657959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,4,4,balanced,0.06089600086212158
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,4,4,balanced,0.06372863769531249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,4,4,balanced,0.07604991912841796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,4,4,balanced,0.07017727851867675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,4,4,balanced,0.09138688087463379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,4,4,balanced,0.10338560104370118
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,4,4,balanced,0.14161151885986328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,4,4,balanced,0.17118080139160158
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,4,4,balanced,0.23982463836669923
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,4,4,balanced,0.3159091186523438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,4,4,balanced,0.4576742553710938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,4,4,balanced,0.5930969619750976
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,2,8,power_law_1.2,0.17067392349243166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,4,4,balanced,0.05633024215698242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,2,8,power_law_1.2,0.3113011169433594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,2,8,power_law_1.2,0.38750720977783204
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,4,4,balanced,0.018650879859924314
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,4,4,balanced,0.019144959449768066
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,4,4,balanced,0.019589120149612428
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,4,4,balanced,0.019408639669418335
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,4,4,balanced,0.020279040336608885
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,4,4,balanced,0.02062335968017578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,4,4,balanced,0.02026623964309692
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,4,4,balanced,0.02035072088241577
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,4,4,balanced,0.02065664052963257
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,4,4,balanced,0.02066431999206543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,4,4,balanced,0.02065279960632324
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,4,4,balanced,0.02065407991409302
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,4,4,balanced,0.019868160486221316
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,4,4,balanced,0.019326720237731933
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,4,4,balanced,0.019394559860229494
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,4,4,balanced,0.01950719952583313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,4,4,balanced,0.01976704001426697
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,4,4,balanced,0.02888319969177246
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,4,4,balanced,0.02928639888763428
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,4,4,balanced,0.04111487865447998
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,4,4,balanced,0.04762239933013916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,4,4,power_law_1.01,0.06449408054351806
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,4,4,balanced,0.06222464084625244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,4,4,power_law_1.01,0.06751488208770752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,4,4,power_law_1.01,0.11830143928527832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,4,4,power_law_1.01,0.12039936065673829
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,4,4,power_law_1.01,0.12124928474426269
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,4,4,power_law_1.01,0.12239744186401366
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,4,4,power_law_1.01,0.12402048110961914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,4,4,power_law_1.01,0.12552576065063475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,4,4,power_law_1.01,0.12753664016723634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,4,4,power_law_1.01,0.12918527603149416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,4,4,power_law_1.01,0.13239551544189454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,4,4,power_law_1.01,0.13649024009704588
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,4,4,power_law_1.01,0.13831168174743652
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,4,4,power_law_1.01,0.1518566417694092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,4,4,power_law_1.01,0.17825023651123045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,4,4,power_law_1.01,0.18847488403320312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,4,4,power_law_1.01,0.20416383743286132
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,4,4,power_law_1.01,0.2491763114929199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,4,4,power_law_1.01,0.28974208831787107
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,4,4,power_law_1.01,0.5141299057006836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,4,4,power_law_1.01,0.40828289031982423
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,4,4,power_law_1.01,0.7287667083740235
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,4,4,balanced,0.07631743907928466
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,4,4,power_law_1.01,1.025464324951172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,4,4,balanced,0.10011648178100585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,4,4,power_law_1.01,2.001141815185547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,4,4,power_law_1.01,0.039448320865631104
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,4,4,power_law_1.01,0.0434611177444458
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,4,4,power_law_1.01,0.04725632190704346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,4,4,power_law_1.01,1.492428741455078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,2,8,power_law_1.2,0.855068130493164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,4,4,power_law_1.01,0.04823296070098877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,4,4,power_law_1.01,0.04886528015136719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,4,4,power_law_1.01,0.05471744060516357
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,4,4,power_law_1.01,0.04951039791107178
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,4,4,power_law_1.01,0.057393918037414546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,4,4,power_law_1.01,0.05148928165435791
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,4,4,power_law_1.01,0.05313536167144776
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,4,4,balanced,0.12493439674377442
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,4,4,power_law_1.01,0.055866880416870116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,4,4,power_law_1.01,0.060016641616821284
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,4,4,power_law_1.01,0.058432002067565915
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,4,4,balanced,0.1780031967163086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,4,4,power_law_1.01,0.06536831855773925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,4,4,power_law_1.01,0.07067647933959961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,4,4,power_law_1.01,0.07213952064514159
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,4,4,power_law_1.01,0.07703936100006104
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,4,4,power_law_1.01,0.10510975837707519
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,4,4,power_law_1.01,0.08745087623596191
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,4,4,power_law_1.01,0.18247167587280272
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,4,4,balanced,0.23070463180541992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,4,4,power_law_1.01,0.13720959663391114
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,4,4,power_law_1.01,0.23835519790649412
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,4,4,power_law_1.01,0.3117913627624512
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,4,4,power_law_1.01,0.01904255986213684
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,4,4,power_law_1.01,0.019454720020294188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,4,4,power_law_1.01,0.019415040016174317
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,4,4,power_law_1.01,0.019619840383529662
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,4,4,power_law_1.01,0.019754240512847902
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,4,4,power_law_1.01,0.020209920406341553
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,4,4,power_law_1.01,0.020199680328369142
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,4,4,power_law_1.01,0.0204915189743042
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,4,4,power_law_1.01,0.020244479179382324
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,4,4,power_law_1.01,0.020665600299835205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,4,4,power_law_1.01,0.4959718322753906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,4,4,power_law_1.01,0.9143218994140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,4,4,power_law_1.01,3.607244873046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,4,4,power_law_1.01,0.6004403305053712
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,4,4,power_law_1.01,2.946397399902344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,4,4,power_law_1.01,0.03779072046279907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,4,4,power_law_1.01,1.1513804626464843
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,4,4,power_law_1.01,0.018981120586395263
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,4,4,power_law_1.01,0.023531520366668703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,4,4,power_law_1.01,0.024764161109924316
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,4,4,power_law_1.01,0.027980799674987795
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,4,4,power_law_1.01,0.03174015998840332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,4,4,power_law_1.01,0.030120959281921388
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,4,4,power_law_1.01,0.15362303733825683
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,4,4,power_law_1.01,0.04258560180664063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,4,4,power_law_1.01,0.051823358535766605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,4,4,power_law_1.2,0.064235520362854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,4,4,power_law_1.2,0.06507264137268066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,4,4,power_law_1.2,0.11788288116455078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,4,4,power_law_1.2,0.12062848091125489
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,4,4,power_law_1.2,0.12128767967224122
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,4,4,power_law_1.2,0.12172287940979003
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,4,4,power_law_1.2,0.12409088134765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,4,4,power_law_1.2,0.12626048088073732
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,4,4,power_law_1.2,0.1282521629333496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,4,4,power_law_1.2,0.12980992317199708
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,4,4,power_law_1.2,0.13204352378845213
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,4,4,power_law_1.01,0.06760704040527343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,4,4,power_law_1.2,0.13818623542785644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,4,4,power_law_1.2,0.13577856063842772
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,4,4,power_law_1.2,0.15432831764221192
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,4,4,power_law_1.2,0.17757568359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,4,4,power_law_1.2,0.19082496643066407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,4,4,power_law_1.2,0.20669183731079102
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,4,4,power_law_1.01,0.10346240043640136
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,4,4,power_law_1.2,0.24546432495117188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,4,4,power_law_1.2,0.29131776809692383
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,4,4,power_law_1.01,0.08558208465576171
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,4,4,power_law_1.2,0.40155265808105467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,4,4,power_law_1.2,0.4816678237915039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,4,4,power_law_1.2,0.988963851928711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,4,4,power_law_1.2,0.7830413055419922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,4,4,power_law_1.01,0.18757503509521484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,4,4,power_law_1.2,0.037943038940429684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,4,4,power_law_1.2,0.043298559188842775
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,4,4,power_law_1.2,0.04693247795104981
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,4,4,power_law_1.2,0.04770431995391845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,4,4,power_law_1.2,0.04824575901031494
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,4,4,power_law_1.2,0.04902527809143066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,4,4,power_law_1.2,0.04951168060302734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,4,4,power_law_1.2,0.04991231918334961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,4,4,power_law_1.2,0.05213183879852294
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,4,4,power_law_1.2,0.05266047954559326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,4,4,power_law_1.2,0.05502592086791992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,4,4,power_law_1.2,0.059572482109069826
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,4,4,power_law_1.2,0.061456642150878905
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,4,4,power_law_1.01,0.3863167953491211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,4,4,power_law_1.2,0.06531328201293946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,4,4,power_law_1.2,0.07139455795288085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,4,4,power_law_1.2,0.07229055881500243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,4,4,power_law_1.2,0.07974271774291992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,4,4,power_law_1.2,0.08884223937988281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,4,4,power_law_1.2,0.10579456329345702
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,4,4,power_law_1.2,1.411083526611328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,4,4,power_law_1.2,0.13506688117980956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,4,4,power_law_1.2,0.18203392028808593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,4,4,power_law_1.2,1.7861491394042968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,4,4,power_law_1.01,0.2773555183410644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,4,4,power_law_1.2,0.2628198432922363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,4,4,power_law_1.2,0.33148799896240233
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,4,4,power_law_1.2,0.018992639780044555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,4,4,power_law_1.2,0.018999040126800537
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,4,4,power_law_1.2,0.019459840059280396
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,4,4,power_law_1.2,0.01938048005104065
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,4,4,power_law_1.2,0.019792640209198
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,4,4,power_law_1.2,0.01989248037338257
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,4,4,power_law_1.2,0.02021631956100464
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,4,4,power_law_1.2,0.020268800258636473
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,4,4,power_law_1.2,0.02026623964309692
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,4,4,power_law_1.2,0.0202726411819458
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,4,4,power_law_1.2,0.020682239532470705
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,4,4,power_law_1.2,0.024312319755554198
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,4,4,power_law_1.2,0.5402316665649414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,4,4,power_law_1.2,0.025223679542541504
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,4,4,power_law_1.2,0.028116478919982913
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,4,4,power_law_1.2,0.029208319187164305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,4,4,power_law_1.2,0.03128704071044922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,4,4,power_law_1.2,2.90524658203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,4,4,power_law_1.2,0.47906944274902347
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,4,4,power_law_1.2,0.8971430206298828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,4,4,power_law_1.2,3.6509951782226566
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,4,4,power_law_1.01,0.5825279998779297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,4,4,power_law_1.2,0.03781888008117676
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,4,4,power_law_1.2,0.04462719917297363
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,4,4,power_law_1.2,0.051641597747802734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,4,4,power_law_1.2,1.250481948852539
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,4,4,power_law_1.2,0.06656511783599853
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,8,2,balanced,0.030520319938659668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,8,2,balanced,0.04984447956085205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,8,2,balanced,0.062174720764160155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,8,2,balanced,0.09903615951538086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,8,2,balanced,0.09847807884216309
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,8,2,balanced,0.09962240219116211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,8,2,balanced,0.1016486358642578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,8,2,balanced,0.10199423789978028
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,8,2,balanced,0.10434304237365724
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,8,2,balanced,0.10461312294006349
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,8,2,balanced,0.10598912239074706
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,8,2,balanced,0.10804608345031738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,8,2,balanced,0.10912511825561524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,8,2,balanced,0.1145471954345703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,8,2,balanced,0.12524543762207033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,8,2,balanced,0.12615424156188965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,8,2,balanced,0.11730815887451171
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,8,2,balanced,0.13530112266540528
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,8,2,balanced,0.165482234954834
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,4,4,power_law_1.2,0.0876364803314209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,8,2,balanced,0.23155071258544924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,8,2,balanced,0.2682022476196289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,8,2,balanced,0.3764467239379883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,8,2,balanced,0.48323841094970704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,8,2,balanced,0.7057561492919922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,8,2,balanced,0.9271692657470704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,8,2,balanced,1.3855462646484376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,8,2,balanced,0.03050112009048462
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,4,4,power_law_1.2,0.2560947227478027
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,8,2,balanced,1.812769317626953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,8,2,balanced,0.04188159942626953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,8,2,balanced,0.04157440185546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,8,2,balanced,0.04280704021453857
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,8,2,balanced,0.033263359069824215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,8,2,balanced,0.04241536140441894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,8,2,balanced,0.04492800235748291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,8,2,balanced,0.0460595178604126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,8,2,balanced,0.04592127799987793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,8,2,balanced,0.04509823799133301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,8,2,balanced,0.049664001464843746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,8,2,balanced,0.051827201843261714
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,8,2,balanced,0.055486721992492674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,8,2,balanced,0.05483391761779786
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,8,2,balanced,0.06356224060058593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,8,2,balanced,0.05753727912902832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,8,2,balanced,0.06844031810760498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,8,2,balanced,0.07709695816040038
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,8,2,balanced,0.10403455734252928
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,8,2,balanced,0.12010623931884765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,8,2,balanced,0.1669081687927246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,8,2,balanced,0.2222400093078613
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,8,2,balanced,0.30853248596191407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,8,2,balanced,0.4042854309082031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,8,2,balanced,0.598205451965332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,8,2,balanced,0.7758694458007812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,8,2,balanced,0.038926079273223876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,4,4,power_law_1.2,0.12449536323547364
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,4,4,power_law_1.2,0.162108154296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,4,4,power_law_1.2,0.2140620803833008
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,8,2,balanced,0.018211840391159057
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,8,2,balanced,0.019631359577178955
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,8,2,balanced,0.019586559534072876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,8,2,balanced,0.01986304044723511
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,8,2,balanced,0.018626559972763062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,8,2,balanced,0.01987328052520752
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,8,2,balanced,0.02035840034484863
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,8,2,balanced,0.020288000106811525
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,8,2,balanced,0.020678400993347168
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,8,2,balanced,0.020776960849761963
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,8,2,balanced,0.021081600189208984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,8,2,balanced,0.02154752016067505
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,8,2,balanced,0.02152575969696045
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,8,2,balanced,0.019423999786376954
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,4,4,power_law_1.2,0.42569217681884763
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,8,2,balanced,0.01943168044090271
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,8,2,balanced,0.019471360445022585
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,8,2,balanced,0.01941375970840454
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,8,2,balanced,0.032285439968109134
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,8,2,balanced,0.03342463970184326
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,8,2,balanced,0.045204482078552245
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,8,2,balanced,0.052330241203308106
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,4,4,power_law_1.2,0.5723020935058594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,8,2,power_law_1.01,0.05631743907928467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,8,2,power_law_1.01,0.056892161369323735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,8,2,power_law_1.01,0.09902848243713379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,8,2,power_law_1.01,0.098538236618042
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,8,2,power_law_1.01,0.10004608154296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,8,2,power_law_1.01,0.1013542366027832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,8,2,power_law_1.01,0.10251263618469238
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,8,2,balanced,0.06917503833770752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,8,2,power_law_1.01,0.10461440086364746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,8,2,power_law_1.01,0.10634752273559571
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,8,2,power_law_1.01,0.10431360244750976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,8,2,power_law_1.01,0.10816384315490724
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,8,2,power_law_1.01,0.1112844753265381
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,8,2,power_law_1.01,0.1198540782928467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,8,2,power_law_1.01,0.12905216217041016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,8,2,power_law_1.01,0.049354238510131834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,8,2,power_law_1.01,0.1338764762878418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,8,2,power_law_1.01,0.1534553623199463
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,8,2,power_law_1.01,0.19017728805541992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,8,2,power_law_1.01,0.21327743530273438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,8,2,power_law_1.01,0.2854348754882813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,8,2,power_law_1.01,0.33398784637451173
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,8,2,balanced,0.08121855735778809
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,8,2,power_law_1.01,0.504925422668457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,8,2,power_law_1.01,1.5036338806152343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,8,2,power_law_1.01,0.6677555084228516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,8,2,power_law_1.01,0.03056256055831909
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,8,2,power_law_1.01,0.03748608112335205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,8,2,power_law_1.01,0.037601280212402347
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,8,2,power_law_1.01,0.04165247917175293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,8,2,power_law_1.01,0.04246399879455566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,8,2,power_law_1.01,0.042803201675415035
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,8,2,balanced,0.11325056076049804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,8,2,power_law_1.01,0.04321407794952393
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,8,2,power_law_1.01,0.04362112045288086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,8,2,power_law_1.01,0.044948477745056156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,8,2,power_law_1.01,0.04610432147979736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,8,2,power_law_1.01,0.04687359809875488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,8,2,power_law_1.01,0.9800307464599609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,8,2,power_law_1.01,0.050671358108520505
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,8,2,power_law_1.01,0.053038082122802734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,8,2,power_law_1.01,0.054097919464111334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,8,2,power_law_1.01,0.05919360160827637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,8,2,power_law_1.01,0.06172927856445313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,8,2,power_law_1.01,0.0659891176223755
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,8,2,power_law_1.01,0.08195839881896973
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,8,2,power_law_1.01,0.09251968383789062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,8,2,power_law_1.01,0.11426943778991698
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,8,2,power_law_1.01,0.14594559669494628
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,8,2,power_law_1.01,0.36115585327148436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,8,2,power_law_1.01,0.4345446395874023
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,8,2,balanced,0.14254336357116698
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,8,2,power_law_1.01,0.20583679199218752
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,8,2,power_law_1.01,0.018581759929656983
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,8,2,balanced,0.20194431304931643
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,8,2,power_law_1.01,0.018558720350265505
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,8,2,power_law_1.01,0.2427916717529297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,8,2,power_law_1.01,0.01967743992805481
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,8,2,power_law_1.01,0.019453439712524414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,8,2,power_law_1.01,0.019521280527114868
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,8,2,power_law_1.01,0.019860479831695557
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,8,2,power_law_1.01,0.020320000648498534
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,8,2,power_law_1.01,0.02021503925323486
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,8,2,balanced,0.2644313621520996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,8,2,power_law_1.01,0.020252161026000977
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,8,2,power_law_1.01,0.020252161026000977
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,8,2,power_law_1.01,0.023788800239562986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,8,2,power_law_1.01,1.9076083374023436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,8,2,power_law_1.01,2.117121276855469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,8,2,power_law_1.01,0.7310015869140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,8,2,power_law_1.01,0.019454720020294188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,8,2,power_law_1.01,0.022725119590759277
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,8,2,power_law_1.01,0.030739200115203858
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,8,2,power_law_1.01,0.8344140624999999
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,8,2,power_law_1.01,0.024715518951416014
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,8,2,power_law_1.01,0.032486400604248046
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,8,2,power_law_1.01,0.03491199970245361
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,8,2,power_law_1.01,0.041994237899780275
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,8,2,power_law_1.01,0.048784642219543456
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,8,2,power_law_1.01,0.06075776100158692
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,8,2,power_law_1.2,0.04679296016693115
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,8,2,power_law_1.01,0.0680614423751831
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,8,2,power_law_1.2,0.049331197738647456
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,8,2,power_law_1.2,0.09820287704467774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,8,2,power_law_1.2,0.09862784385681153
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,8,2,power_law_1.2,0.09962240219116211
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,8,2,power_law_1.01,0.11854720115661621
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,8,2,power_law_1.01,0.10499456405639648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,8,2,power_law_1.2,0.056965122222900386
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,8,2,power_law_1.2,0.10096256256103515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,8,2,power_law_1.2,0.10358400344848633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,8,2,power_law_1.2,0.10410112380981444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,8,2,power_law_1.2,0.1045695972442627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,8,2,power_law_1.2,0.10751487731933593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,8,2,power_law_1.2,0.10869759559631346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,8,2,power_law_1.2,0.11302271842956543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,8,2,power_law_1.2,0.12239487648010254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,8,2,power_law_1.2,0.12679807662963866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,8,2,power_law_1.2,0.132542724609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,8,2,power_law_1.2,0.1513548755645752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,8,2,power_law_1.2,0.21014015197753907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,8,2,power_law_1.2,0.18536832809448242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,8,2,power_law_1.2,0.3255820846557617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,8,2,power_law_1.2,0.2819584083557129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,8,2,power_law_1.2,0.4706047821044922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,8,2,power_law_1.01,0.16780927658081055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,8,2,power_law_1.2,0.625665283203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,8,2,power_law_1.2,0.03086591958999634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,8,2,power_law_1.2,0.0375167989730835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,8,2,power_law_1.2,2.4633702087402343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,8,2,power_law_1.2,0.037662720680236815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,8,2,power_law_1.2,0.041982722282409665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,8,2,power_law_1.2,0.04223231792449951
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,8,2,power_law_1.2,0.0421619176864624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,8,2,power_law_1.2,0.04322048187255859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,8,2,power_law_1.2,0.043632640838623046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,8,2,power_law_1.2,0.045154561996459966
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,8,2,power_law_1.2,0.04488959789276123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,8,2,power_law_1.2,0.04899712085723877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,8,2,power_law_1.2,0.05108096122741699
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,8,2,power_law_1.2,0.0520691204071045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,8,2,power_law_1.2,0.05511040210723876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,8,2,power_law_1.2,0.05764351844787598
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,8,2,power_law_1.2,0.0620633602142334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,8,2,power_law_1.2,0.06476672172546387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,8,2,power_law_1.2,0.07951488018035888
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,8,2,power_law_1.2,0.09221119880676269
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,8,2,power_law_1.01,0.22118656158447267
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,8,2,power_law_1.2,1.0083814239501954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,8,2,power_law_1.2,0.11282688140869142
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,8,2,power_law_1.2,0.14443519592285156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,8,2,power_law_1.2,0.19518463134765623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,8,2,power_law_1.2,1.147150115966797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,8,2,power_law_1.2,0.35922176361083985
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,8,2,power_law_1.2,0.018247679471969605
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,8,2,power_law_1.2,0.01874559998512268
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,8,2,power_law_1.2,0.01863935947418213
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,8,2,power_law_1.2,0.019471360445022585
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,8,2,power_law_1.2,0.019389439821243287
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,8,2,power_law_1.2,0.01982591986656189
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,8,2,power_law_1.2,0.019814399480819703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,8,2,power_law_1.2,0.020062720775604247
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,8,2,power_law_1.2,0.020217599868774413
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,8,2,power_law_1.2,0.02010240077972412
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,8,2,power_law_1.2,0.020616960525512696
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,8,2,power_law_1.2,0.023141119480133057
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,8,2,power_law_1.2,0.023948800563812257
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,8,2,power_law_1.2,0.026338560581207277
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,8,2,power_law_1.2,1.7821580505371095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,8,2,power_law_1.2,0.2516057586669922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,8,2,power_law_1.2,0.4498432159423828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,8,2,power_law_1.01,0.4873676681518555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,8,2,power_law_1.2,0.045308160781860354
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,8,2,power_law_1.01,0.294279670715332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,8,2,power_law_1.2,0.6810956573486328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,8,2,power_law_1.2,0.030737919807434084
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,8,2,power_law_1.2,0.032249600887298585
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,8,2,power_law_1.2,0.0342195200920105
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,8,2,power_law_1.2,0.04108672142028809
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,8,2,power_law_1.2,0.06170368194580078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,8,2,power_law_1.2,0.11670144081115723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,16,1,balanced,0.03247936010360718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,16,1,balanced,0.050730881690979
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,16,1,balanced,0.06396160125732422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,16,1,balanced,0.06416384220123292
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,8,2,power_law_1.2,0.8067456054687501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,16,1,balanced,0.09461055755615234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,16,1,balanced,0.09670080184936523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,16,1,balanced,0.09603839874267578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,16,1,balanced,0.09827263832092285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,16,1,balanced,0.09779840469360351
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,16,1,balanced,0.09848256111145019
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,16,1,balanced,0.09928704261779785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,16,1,balanced,0.10160127639770508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,16,1,balanced,0.10186752319335937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,16,1,balanced,0.10656064033508302
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,8,2,power_law_1.2,0.06853248119354248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,16,1,balanced,0.11246975898742675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,16,1,balanced,0.11602432250976562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,16,1,balanced,0.11452095985412598
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,16,1,balanced,0.13050304412841798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,16,1,balanced,0.15482815742492675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,16,1,balanced,0.2195449638366699
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,16,1,balanced,0.2512620735168457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,16,1,balanced,0.35195327758789063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,16,1,balanced,0.45268798828124995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,16,1,balanced,0.6587328338623047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,16,1,balanced,0.8621689605712891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,16,1,balanced,1.3023942565917968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,16,1,balanced,0.034517760276794436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,16,1,balanced,0.03655424118041992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,16,1,balanced,0.03919935941696167
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,16,1,balanced,0.03998016119003296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,16,1,balanced,0.04515007972717285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,16,1,balanced,1.7251635742187499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,16,1,balanced,0.046384639739990234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,16,1,balanced,0.04674623966217041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,16,1,balanced,0.04875648021697998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,16,1,balanced,0.05009344100952149
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,16,1,balanced,0.04797952175140381
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,8,2,power_law_1.2,0.10534655570983886
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,16,1,balanced,0.05153215885162353
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,16,1,balanced,0.05083648204803467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,16,1,balanced,0.05925183773040772
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,16,1,balanced,0.06079360008239746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,16,1,balanced,0.062186880111694334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,16,1,balanced,0.0724396800994873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,16,1,balanced,0.0785536003112793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,16,1,balanced,0.10888575553894042
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,16,1,balanced,0.12555904388427735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,16,1,balanced,0.17418624877929687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,16,1,balanced,0.049687042236328124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,16,1,balanced,0.2176326370239258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,16,1,balanced,0.057079038619995116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,16,1,balanced,0.3086879920959473
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,16,1,balanced,0.40211456298828124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,16,1,balanced,0.5833043289184571
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,16,1,balanced,0.7637318420410156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,8,2,power_law_1.2,0.15805312156677248
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,8,2,power_law_1.2,0.23032703399658203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,16,1,balanced,0.02014528036117554
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,16,1,balanced,0.01849984049797058
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,16,1,balanced,0.02235136032104492
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,16,1,balanced,0.022579200267791748
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,16,1,balanced,0.022622718811035156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,16,1,balanced,0.022519679069519044
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,16,1,balanced,0.022591359615325927
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,16,1,balanced,0.022810881137847898
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,16,1,balanced,0.021436800956726072
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,16,1,balanced,0.022593278884887696
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,16,1,balanced,0.02281343936920166
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,16,1,balanced,0.023449599742889404
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,16,1,balanced,0.024718079566955566
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,16,1,balanced,0.022110719680786133
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,16,1,balanced,0.02550015926361084
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,8,2,power_law_1.2,0.4052659225463867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,8,2,power_law_1.2,0.29977216720581057
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,16,1,balanced,0.02151871919631958
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,16,1,balanced,0.022198400497436523
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,16,1,balanced,0.04180352210998535
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,16,1,balanced,0.03835968017578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,16,1,balanced,0.05551296234130859
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,16,1,balanced,0.06321216106414795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,16,1,power_law_1.01,0.0462553596496582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,16,1,power_law_1.01,0.05296256065368652
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,16,1,power_law_1.01,0.056848001480102536
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,16,1,power_law_1.01,0.0643225622177124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,16,1,power_law_1.01,0.09504256248474122
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,16,1,power_law_1.01,0.09493056297302246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,16,1,power_law_1.01,0.09796095848083496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,16,1,power_law_1.01,0.0967750358581543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,16,1,power_law_1.01,0.0992569637298584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,16,1,power_law_1.01,0.09911423683166504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,16,1,power_law_1.01,0.10345664024353027
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,16,1,power_law_1.01,0.10309247970581055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,16,1,power_law_1.01,0.10852288246154784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,16,1,power_law_1.01,0.11041919708251953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,16,1,power_law_1.01,0.11499711990356445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,16,1,power_law_1.01,0.12158783912658691
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,16,1,power_law_1.01,0.1265664005279541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,16,1,power_law_1.01,0.14996095657348635
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,16,1,power_law_1.01,0.18150911331176758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,16,1,power_law_1.01,0.2317184066772461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,16,1,power_law_1.01,0.28150463104248047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,16,1,power_law_1.01,0.38105728149414064
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,16,1,balanced,0.08486335754394532
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,16,1,balanced,0.10562368392944337
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,16,1,power_law_1.01,0.48248958587646484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,16,1,power_law_1.01,0.6876441955566406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,16,1,power_law_1.01,0.033656959533691404
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,16,1,power_law_1.01,0.03796159982681274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,16,1,power_law_1.01,0.041036162376403805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,16,1,power_law_1.01,0.03948031902313233
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,16,1,balanced,0.14590399742126464
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,16,1,power_law_1.01,0.04475264072418213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,16,1,power_law_1.01,0.04694655895233154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,16,1,power_law_1.01,0.046798081398010255
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,16,1,power_law_1.01,0.04883135795593262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,16,1,power_law_1.01,0.04643008232116699
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,16,1,power_law_1.01,0.050103678703308105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,16,1,power_law_1.01,0.04974207878112793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,16,1,power_law_1.01,0.051735038757324214
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,16,1,power_law_1.01,0.05371007919311523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,16,1,power_law_1.01,0.059976320266723636
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,16,1,power_law_1.01,0.061468801498413085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,16,1,power_law_1.01,0.06507455825805664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,16,1,power_law_1.01,0.06578688144683838
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,16,1,balanced,0.189117431640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,16,1,power_law_1.01,0.07313983917236329
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,16,1,power_law_1.01,0.8923763275146485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,16,1,power_law_1.01,0.09110912322998047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,16,1,balanced,0.27495359420776366
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,16,1,balanced,0.3630636978149414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,16,1,power_law_1.01,0.11544768333435058
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,16,1,power_law_1.01,0.1309152030944824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,16,1,power_law_1.01,0.17625280380249025
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,16,1,power_law_1.01,0.018935680389404297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,16,1,power_law_1.01,0.019742720127105713
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,16,1,power_law_1.01,0.0205785608291626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,16,1,power_law_1.01,0.021396479606628417
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,16,1,power_law_1.01,0.021413760185241697
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,16,1,power_law_1.01,0.021804161071777343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,16,1,power_law_1.01,0.22769664764404296
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,16,1,power_law_1.01,0.021822080612182618
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,16,1,power_law_1.01,0.022221438884735108
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,16,1,power_law_1.01,0.022207999229431154
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,16,1,power_law_1.01,0.022200961112976075
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,16,1,power_law_1.01,0.022206079959869385
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,16,1,power_law_1.01,0.023994879722595213
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,16,1,power_law_1.01,0.027538559436798095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,16,1,power_law_1.01,1.3164671325683595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,16,1,power_law_1.01,0.3112691116333008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,16,1,power_law_1.01,0.41263038635253907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,16,1,power_law_1.01,1.746958770751953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,16,1,power_law_1.01,0.5940038299560546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,16,1,power_law_1.01,0.06417215824127197
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,16,1,power_law_1.01,0.7739398193359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,16,1,power_law_1.01,0.025085439682006837
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,16,1,power_law_1.01,0.02975935935974121
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,16,1,power_law_1.01,0.03289151906967163
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,16,1,power_law_1.01,0.036551039218902584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,16,1,power_law_1.01,0.04326079845428467
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,16,1,power_law_1.01,0.05073279857635497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,16,1,power_law_1.2,0.04819007873535156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,16,1,power_law_1.2,0.05368127822875977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,16,1,power_law_1.2,0.055000319480895996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,16,1,power_law_1.2,0.06603839874267578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,16,1,power_law_1.2,0.09460607528686524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,16,1,power_law_1.2,0.09671423912048341
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,16,1,power_law_1.2,0.09824640274047851
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,16,1,power_law_1.2,0.09829119682312012
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,16,1,power_law_1.2,0.09980671882629395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,16,1,power_law_1.2,0.10049023628234863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,16,1,power_law_1.2,0.10365056037902831
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,16,1,power_law_1.2,0.10553983688354492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,16,1,power_law_1.2,0.10844223976135255
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,16,1,power_law_1.01,0.07700223922729492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,16,1,power_law_1.2,0.11282496452331543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,16,1,power_law_1.2,0.11479104042053223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,16,1,power_law_1.2,0.12536319732666015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,16,1,power_law_1.2,0.12636799812316896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,16,1,power_law_1.2,0.14860351562500002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,16,1,power_law_1.2,0.18176767349243164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,16,1,power_law_1.2,0.23399168014526367
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,16,1,power_law_1.01,0.10280639648437499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,16,1,power_law_1.2,0.28104639053344727
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,16,1,power_law_1.01,0.12892160415649415
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,16,1,power_law_1.01,0.18207168579101562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,16,1,power_law_1.2,0.3818163299560547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,16,1,power_law_1.2,0.48305534362792973
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,16,1,power_law_1.2,0.03422656059265137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,16,1,power_law_1.2,0.03799743890762329
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,16,1,power_law_1.2,0.03979712009429932
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,16,1,power_law_1.01,0.2355596733093262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,16,1,power_law_1.2,0.04485951900482178
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,16,1,power_law_1.2,0.04661695957183838
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,16,1,power_law_1.2,0.047607679367065434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,16,1,power_law_1.2,0.04836607933044433
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,16,1,power_law_1.2,0.04868031978607178
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,16,1,power_law_1.2,0.049271039962768555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,16,1,power_law_1.2,0.05104576110839844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,16,1,power_law_1.2,0.0513702392578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,16,1,power_law_1.2,0.0541164779663086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,16,1,power_law_1.2,0.0588595199584961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,16,1,power_law_1.2,0.06442944049835206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,16,1,power_law_1.2,0.06758848190307618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,16,1,power_law_1.2,0.06578368186950684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,16,1,power_law_1.2,0.07300352096557618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,16,1,power_law_1.2,0.03947648048400879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,16,1,power_law_1.2,0.6873824310302734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,16,1,power_law_1.01,0.34526336669921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,16,1,power_law_1.2,0.08955519676208497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,16,1,power_law_1.2,0.11427392005920409
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,16,1,power_law_1.2,0.12999168395996094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,16,1,power_law_1.2,0.17576511383056642
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,16,1,power_law_1.2,0.8905849456787109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,16,1,power_law_1.2,0.2280838394165039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,16,1,power_law_1.01,0.45297088623046877
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,16,1,power_law_1.2,0.019567359685897828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,16,1,power_law_1.2,0.020176639556884764
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,16,1,power_law_1.2,0.021776640415191652
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,16,1,power_law_1.2,0.021772799491882325
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,16,1,power_law_1.2,0.021811840534210206
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,16,1,power_law_1.2,0.02196991920471191
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,16,1,power_law_1.2,0.022271358966827394
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,16,1,power_law_1.2,0.022554240226745605
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,16,1,power_law_1.2,0.02220031976699829
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,16,1,power_law_1.2,0.022593278884887696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,16,1,power_law_1.2,0.3117657661437988
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,16,1,power_law_1.2,1.3182418823242188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,16,1,power_law_1.2,0.02829632043838501
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,16,1,power_law_1.2,0.029836800098419193
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,16,1,power_law_1.2,0.034235520362854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,16,1,power_law_1.2,0.4125113677978516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,16,1,power_law_1.2,0.018910720348358154
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,16,1,power_law_1.2,0.05063360214233399
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,16,1,power_law_1.2,0.023836801052093508
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,16,1,power_law_1.2,0.02578943967819214
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,16,1,power_law_1.2,1.7601702880859373
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,16,1,power_law_1.2,0.106693115234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,16,1,power_law_1.2,0.5932089614868163
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,16,1,power_law_1.2,0.0366598391532898
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,16,1,power_law_1.2,0.04364031791687012
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,16,1,power_law_1.2,0.06507264137268066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,2,8,balanced,0.12033280372619629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,2,8,balanced,0.12058624267578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,2,8,balanced,0.12059136390686036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,2,8,balanced,0.21914239883422854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,2,8,balanced,0.2211673545837402
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,2,8,balanced,0.22112768173217773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,2,8,balanced,0.22113536834716796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,2,8,balanced,0.22033920288085937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,2,8,balanced,0.22179712295532225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,2,8,balanced,0.2218342399597168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,16,1,power_law_1.2,0.7743199920654297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,2,8,balanced,0.22125183105468751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,2,8,balanced,0.22399360656738282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,2,8,balanced,0.22535167694091798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,2,8,balanced,0.22783103942871094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,2,8,balanced,0.24491519927978517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,2,8,balanced,0.24333696365356444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,2,8,balanced,0.2602560043334961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,2,8,balanced,0.30087295532226566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,2,8,balanced,0.3233638381958008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,2,8,balanced,0.3969919967651367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,2,8,balanced,0.3879667282104492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,2,8,balanced,0.5968460845947265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,2,8,balanced,0.7509683227539062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,2,8,balanced,1.0908812713623046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,2,8,balanced,1.425041961669922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,16,1,power_law_1.2,0.07879744052886964
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,2,8,balanced,0.07115263938903808
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,2,8,balanced,0.07122432231903077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,2,8,balanced,0.07114880084991455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,2,8,balanced,0.07794943809509278
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,2,8,balanced,0.07851263999938965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,2,8,balanced,0.07876863956451416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,2,8,balanced,0.0794316816329956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,2,8,balanced,2.143996124267578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,2,8,balanced,0.07982719898223876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,2,8,balanced,0.08413567543029785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,2,8,balanced,0.08051072120666504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,2,8,balanced,0.08239104270935058
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,2,8,balanced,0.08123264312744141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,2,8,balanced,0.08285951614379883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,2,8,balanced,0.08287487983703613
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,2,8,balanced,0.08560895919799805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,2,8,balanced,0.08583552360534667
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,2,8,balanced,0.09008255958557129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,2,8,balanced,2.9422183227539063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,2,8,balanced,0.09689599990844726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,2,8,balanced,0.1028940773010254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,2,8,balanced,0.14079744338989258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,2,8,balanced,0.13206399917602538
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,2,8,balanced,0.17330047607421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,2,8,balanced,0.23699199676513674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,2,8,balanced,0.29736192703247066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,2,8,balanced,0.3794188690185547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,2,8,balanced,0.5646195220947265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,2,8,balanced,0.7423436737060547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,16,1,power_law_1.2,0.1326630401611328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,16,1,power_law_1.2,0.18806207656860352
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,16,1,power_law_1.2,0.24407167434692384
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,2,8,balanced,0.027601919174194335
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,2,8,balanced,0.026109440326690676
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,2,8,balanced,0.02761856079101562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,2,8,balanced,0.02681727886199951
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,2,8,balanced,0.027710719108581545
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,2,8,balanced,0.02665855884552002
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,2,8,balanced,0.027294719219207765
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,2,8,balanced,0.026645760536193847
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,2,8,balanced,0.027180800437927245
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,2,8,balanced,0.027918078899383546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,2,8,balanced,0.027019519805908204
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,2,8,balanced,0.02896384000778198
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,2,8,balanced,0.029319679737091063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,2,8,balanced,0.02944256067276001
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,16,1,power_law_1.2,0.35626174926757814
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,2,8,balanced,0.027617280483245847
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,2,8,balanced,0.02787967920303345
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,16,1,power_law_1.2,0.4707001495361328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,2,8,balanced,0.02724992036819458
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,2,8,balanced,0.051303682327270506
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,2,8,balanced,0.04269311904907226
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,2,8,balanced,0.040284161567687986
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,2,8,balanced,0.06631936073303223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,2,8,power_law_1.01,0.12083328247070313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,2,8,power_law_1.01,0.21830528259277343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,2,8,power_law_1.01,0.21993087768554687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,2,8,power_law_1.01,0.2196710395812988
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,2,8,power_law_1.01,0.21925632476806642
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,2,8,power_law_1.01,0.22067840576171874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,2,8,power_law_1.01,0.22477951049804687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,2,8,power_law_1.01,0.22591743469238282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,2,8,power_law_1.01,0.24033151626586916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,2,8,power_law_1.01,0.2417958450317383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,2,8,power_law_1.01,0.2525811195373535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,2,8,power_law_1.01,0.2790169525146484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,2,8,power_law_1.01,0.28922624588012696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,2,8,power_law_1.01,0.30909952163696286
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,2,8,power_law_1.01,0.348460807800293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,2,8,power_law_1.01,0.38212223052978517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,2,8,power_law_1.01,0.3802739334106445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,2,8,power_law_1.01,0.5500326538085938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,2,8,balanced,0.08310527801513672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,2,8,power_law_1.01,0.6804096221923828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,2,8,power_law_1.01,0.9867622375488281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,2,8,power_law_1.01,1.1939238739013671
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,2,8,power_law_1.01,1.8545689392089844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,2,8,power_law_1.01,2.350661163330078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,2,8,balanced,0.10168831825256348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,2,8,power_law_1.01,4.779051513671876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,2,8,power_law_1.01,0.07613183975219726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,2,8,power_law_1.01,0.08058879852294923
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,2,8,power_law_1.01,0.07990272045135498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,2,8,power_law_1.01,0.07983744144439697
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,2,8,power_law_1.01,0.07990784168243407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,2,8,power_law_1.01,0.08088959693908691
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,2,8,power_law_1.01,0.08138879776000976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,2,8,power_law_1.01,0.0820633602142334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,2,8,power_law_1.01,0.08334976196289062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,2,8,power_law_1.01,0.08413951873779298
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,2,8,power_law_1.01,3.403036193847656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,2,8,power_law_1.01,0.08679807662963868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,2,8,power_law_1.01,0.09585408210754395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,2,8,power_law_1.01,0.1012339210510254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,2,8,power_law_1.01,0.11346816062927245
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,2,8,power_law_1.01,0.12049792289733888
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,2,8,power_law_1.01,0.09365887641906738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,2,8,power_law_1.01,0.1335167980194092
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,2,8,balanced,0.1429708766937256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,2,8,power_law_1.01,0.15727999687194824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,2,8,power_law_1.01,0.21047679901123048
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,2,8,power_law_1.01,0.25617151260375975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,2,8,power_law_1.01,0.5207104110717774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,2,8,power_law_1.01,0.328851203918457
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,2,8,balanced,0.18485376358032227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,2,8,power_law_1.01,0.6377369689941406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,2,8,power_law_1.01,9.241802368164063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,2,8,power_law_1.01,0.02829952001571655
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,2,8,power_law_1.01,0.028558080196380613
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,2,8,power_law_1.01,0.027669761180877682
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,2,8,power_law_1.01,0.02747904062271118
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,2,8,power_law_1.01,0.027472639083862306
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,2,8,power_law_1.01,0.027971839904785155
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,2,8,power_law_1.01,0.028044800758361816
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,2,8,power_law_1.01,0.02781183958053589
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,2,8,power_law_1.01,0.028367359638214112
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,2,8,power_law_1.01,0.02764928102493286
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,2,8,power_law_1.01,0.028385279178619383
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,2,8,power_law_1.01,0.03462016105651856
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,2,8,power_law_1.01,0.036847360134124756
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,2,8,power_law_1.01,0.0405350399017334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,2,8,power_law_1.01,1.0688678741455078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,2,8,power_law_1.01,2.0395468139648436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,2,8,power_law_1.01,1.3785958862304688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,2,8,balanced,0.27190399169921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,2,8,power_law_1.01,2.7819903564453123
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,2,8,power_law_1.01,0.12849408149719238
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,2,8,balanced,0.3543961715698242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,2,8,power_law_1.01,13.416420898437499
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,2,8,power_law_1.01,0.045181441307067874
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,2,8,power_law_1.01,0.04855040073394775
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,2,8,power_law_1.01,0.05891456127166748
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,2,8,power_law_1.01,0.07443712234497071
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,2,8,power_law_1.01,0.0948031997680664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,2,8,power_law_1.2,0.12141951560974121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,2,8,power_law_1.2,0.217761287689209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,2,8,power_law_1.2,0.22009471893310545
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,2,8,power_law_1.01,0.23857280731201175
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,2,8,power_law_1.2,0.221157112121582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,2,8,power_law_1.2,0.22162303924560547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,2,8,power_law_1.2,0.2203276824951172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,2,8,power_law_1.2,0.22474624633789064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,2,8,power_law_1.2,0.22585344314575195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,2,8,power_law_1.2,0.24039680480957032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,2,8,power_law_1.2,0.23871999740600586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,2,8,power_law_1.2,0.2529523277282715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,2,8,power_law_1.2,0.2803865623474121
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,2,8,power_law_1.01,0.16934528350830078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,2,8,power_law_1.2,0.2891212844848633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,2,8,power_law_1.2,0.30903039932250975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,2,8,power_law_1.2,0.34197887420654294
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,2,8,power_law_1.2,0.3780441665649414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,2,8,power_law_1.2,0.5591475296020507
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,2,8,power_law_1.2,0.6922879791259765
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,2,8,power_law_1.01,0.29813631057739254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,2,8,power_law_1.2,0.3947993469238281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,2,8,power_law_1.2,0.992642593383789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,2,8,power_law_1.2,1.128259811401367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,2,8,power_law_1.2,1.9688755798339845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,2,8,power_law_1.2,2.5322752380371094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,2,8,power_law_1.2,3.7529638671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,2,8,power_law_1.2,0.0736025619506836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,2,8,power_law_1.2,0.0799616003036499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,2,8,power_law_1.2,0.07909632205963135
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,2,8,power_law_1.2,0.07858816146850586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,2,8,power_law_1.2,0.07932032108306884
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,2,8,power_law_1.2,0.08020607948303222
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,2,8,power_law_1.2,0.08121855735778809
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,2,8,power_law_1.2,0.08219136238098144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,2,8,power_law_1.2,0.08360575675964356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,2,8,power_law_1.2,0.08419839859008789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,2,8,power_law_1.2,0.08747903823852539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,2,8,power_law_1.2,0.0918489646911621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,2,8,power_law_1.2,0.09446016311645508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,2,8,power_law_1.2,0.09937024116516113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,2,8,power_law_1.2,5.311727294921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,2,8,power_law_1.2,0.1150937557220459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,2,8,power_law_1.2,0.13385600090026856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,2,8,power_law_1.2,0.12354687690734863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,2,8,power_law_1.2,0.15975168228149414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,2,8,power_law_1.2,0.2105855941772461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,2,8,power_law_1.2,0.346321907043457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,2,8,power_law_1.2,0.27467775344848633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,2,8,power_law_1.2,9.425282592773438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,2,8,power_law_1.2,0.5317990493774414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,2,8,power_law_1.01,0.44599552154541017
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,2,8,power_law_1.2,0.668689956665039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,2,8,power_law_1.2,0.9764288330078126
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,2,8,power_law_1.2,0.02986112117767334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,2,8,power_law_1.2,0.02738944053649902
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,2,8,power_law_1.2,0.02758527994155884
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,2,8,power_law_1.2,0.027215359210968015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,2,8,power_law_1.2,0.02758143901824951
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,2,8,power_law_1.2,0.027334399223327636
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,2,8,power_law_1.2,0.027630081176757814
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,2,8,power_law_1.2,0.027553279399871827
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,2,8,power_law_1.2,0.02806272029876709
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,2,8,power_law_1.2,0.02846719980239868
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,2,8,power_law_1.2,0.030013439655303953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,2,8,power_law_1.01,0.5802726364135742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,2,8,power_law_1.2,2.0547340393066404
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,2,8,power_law_1.2,0.0455731201171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,2,8,power_law_1.2,0.04794623851776123
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,2,8,power_law_1.2,0.049616641998291015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,2,8,power_law_1.01,1.3709709167480468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,2,8,power_law_1.2,13.526862792968751
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,2,8,power_law_1.2,0.08026880264282227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,2,8,power_law_1.2,1.4281369018554688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,2,8,power_law_1.2,2.8658407592773436
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,2,8,power_law_1.01,1.1263577270507814
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,2,8,power_law_1.2,0.0348467206954956
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,2,8,power_law_1.2,0.03909888029098511
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,2,8,power_law_1.2,0.06036352157592774
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,2,8,power_law_1.2,0.13101696014404296
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,2,8,power_law_1.2,0.09655679702758789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,2,8,power_law_1.2,0.17111040115356446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,4,4,balanced,0.0837388801574707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,4,4,balanced,0.09871487617492676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,4,4,balanced,0.1605824089050293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,4,4,balanced,0.1608064079284668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,4,4,balanced,0.1613260841369629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,4,4,balanced,0.1623961639404297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,4,4,balanced,0.16254207611083984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,4,4,balanced,0.16539520263671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,4,4,balanced,0.1650611114501953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,4,4,balanced,0.16585472106933594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,4,4,balanced,0.08400768280029297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,4,4,balanced,0.16899967193603516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,4,4,balanced,0.17154687881469727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,4,4,balanced,0.17592575073242187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,4,4,balanced,0.1859929656982422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,4,4,balanced,0.18518016815185548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,4,4,balanced,0.1983603286743164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,4,4,balanced,0.2660544013977051
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,4,4,balanced,0.24220800399780273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,4,4,balanced,0.35026176452636715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,4,4,balanced,0.4001152038574219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,4,4,balanced,0.5716902542114257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,4,4,balanced,0.7276338958740235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,4,4,balanced,1.0753164672851563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,4,4,balanced,1.4467494201660156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,2,8,power_law_1.2,0.6689228820800781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,4,4,balanced,0.049432320594787596
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,4,4,balanced,2.1661888122558595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,4,4,balanced,0.06371840000152587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,4,4,balanced,0.06578559875488281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,4,4,balanced,0.06526976108551026
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,4,4,balanced,0.06521471977233886
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,4,4,balanced,0.06575104236602783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,4,4,balanced,0.06592127799987793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,4,4,balanced,2.930888977050781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,4,4,balanced,0.05005440235137939
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,4,4,balanced,0.0669593620300293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,4,4,balanced,0.06904191970825195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,4,4,balanced,0.0679641580581665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,4,4,balanced,0.06829823970794677
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,4,4,balanced,0.07004415988922119
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,4,4,balanced,0.07170688152313233
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,4,4,balanced,0.07561855792999268
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,4,4,balanced,0.07673215866088867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,4,4,balanced,0.08256383895874023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,4,4,balanced,0.094136323928833
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,4,4,balanced,0.09050880432128906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,4,4,balanced,0.12255104064941406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,4,4,balanced,0.14450176239013673
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,4,4,balanced,0.19169151306152343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,4,4,balanced,0.24047615051269533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,4,4,balanced,0.34789119720458983
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,4,4,balanced,0.4431679916381836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,4,4,balanced,0.6510208129882813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,4,4,balanced,0.8505996704101563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,2,8,power_law_1.2,0.2554726409912109
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,2,8,power_law_1.2,0.3452492904663086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,2,8,power_law_1.2,0.4974937438964844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,4,4,balanced,0.021233921051025388
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,4,4,balanced,0.025310719013214113
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,4,4,balanced,0.025116159915924075
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,4,4,balanced,0.025176320075988766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,4,4,balanced,0.025612800121307372
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,4,4,balanced,0.02109951972961426
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,4,4,balanced,0.025532159805297848
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,4,4,balanced,0.02682111978530884
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,4,4,balanced,0.02695552110671997
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,4,4,balanced,0.02666111946105957
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,4,4,balanced,0.02763263940811157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,4,4,balanced,0.025029120445251463
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,4,4,balanced,0.028266239166259765
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,2,8,power_law_1.2,1.5438490295410157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,4,4,balanced,0.02846719980239868
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,4,4,balanced,0.025331199169158936
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,4,4,balanced,0.02603264093399048
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,4,4,balanced,0.02528127908706665
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,4,4,balanced,0.04340864181518554
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,4,4,balanced,0.045576958656311034
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,4,4,balanced,0.05435647964477539
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,4,4,balanced,0.06515967845916748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,4,4,power_law_1.01,0.08364928245544434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,4,4,power_law_1.01,0.08666879653930663
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,4,4,power_law_1.01,0.1535961627960205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,4,4,power_law_1.01,0.16057472229003905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,4,4,power_law_1.01,0.16107519149780275
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,4,4,power_law_1.01,0.16247552871704102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,4,4,power_law_1.01,0.16485759735107422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,4,4,power_law_1.01,0.16720256805419922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,4,4,power_law_1.01,0.17249408721923828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,4,4,power_law_1.01,0.17288576126098634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,4,4,power_law_1.01,0.17691648483276368
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,4,4,power_law_1.01,0.18843008041381837
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,4,4,power_law_1.01,0.1906777572631836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,4,4,power_law_1.01,0.21532800674438474
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,4,4,power_law_1.01,0.2517081642150879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,4,4,power_law_1.01,0.2676915168762207
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,2,8,power_law_1.2,1.149680633544922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,4,4,power_law_1.01,0.3004761505126953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,4,4,power_law_1.01,0.38023551940917966
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,4,4,power_law_1.01,0.43378177642822263
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,4,4,balanced,0.08536959648132324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,4,4,power_law_1.01,0.6168819046020507
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,4,4,power_law_1.01,0.7885734558105468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,4,4,power_law_1.01,1.2257344055175783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,4,4,power_law_1.01,1.5460748291015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,4,4,balanced,0.10482175827026367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,4,4,power_law_1.01,3.137511596679688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,4,4,power_law_1.01,0.048682241439819335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,4,4,power_law_1.01,2.222205505371094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,4,4,power_law_1.01,0.06149119853973388
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,4,4,power_law_1.01,0.06620543956756592
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,4,4,power_law_1.01,0.06539904117584229
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,4,4,power_law_1.01,0.0658188819885254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,4,4,power_law_1.01,0.06646783828735352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,4,4,power_law_1.01,0.06739071846008302
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,4,4,power_law_1.01,0.06876416206359863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,4,4,power_law_1.01,0.06934016227722169
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,4,4,power_law_1.01,0.07089536190032959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,4,4,power_law_1.01,0.07402368068695069
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,4,4,power_law_1.01,0.07533055782318115
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,4,4,power_law_1.01,0.08006912231445312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,4,4,balanced,0.14843008041381836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,4,4,power_law_1.01,0.08569087982177734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,4,4,power_law_1.01,0.09120256423950195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,4,4,power_law_1.01,0.11239423751831054
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,4,4,power_law_1.01,0.12801407814025878
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,4,4,power_law_1.01,0.05870336055755615
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,4,4,power_law_1.01,0.2554035186767578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,4,4,power_law_1.01,0.1496908760070801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,4,4,power_law_1.01,0.21790592193603517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,4,4,power_law_1.01,0.34895233154296873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,4,4,power_law_1.01,5.26434814453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,4,4,power_law_1.01,0.48645759582519527
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,4,4,balanced,0.1869900894165039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,4,4,power_law_1.01,0.02350719928741455
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,4,4,balanced,0.2740928077697754
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,4,4,power_law_1.01,0.02508928060531616
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,4,4,power_law_1.01,0.023613440990447997
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,4,4,power_law_1.01,0.025445120334625242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,4,4,power_law_1.01,0.6915596771240234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,4,4,power_law_1.01,0.025626881122589113
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,4,4,balanced,0.36154624938964847
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,4,4,power_law_1.01,0.027886080741882324
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,4,4,power_law_1.01,0.025890560150146485
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,4,4,power_law_1.01,0.028034560680389405
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,4,4,power_law_1.01,0.02696063995361328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,4,4,power_law_1.01,0.9146790313720704
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,4,4,power_law_1.01,0.0276364803314209
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,4,4,power_law_1.01,0.030495998859405515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,4,4,power_law_1.01,1.8654054260253907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,4,4,power_law_1.01,1.488116455078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,4,4,power_law_1.01,0.026451199054718016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,4,4,power_law_1.01,0.038293759822845455
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,4,4,power_law_1.01,0.033351678848266605
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,4,4,power_law_1.01,0.04335487842559814
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,4,4,power_law_1.01,0.04525184154510498
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,4,4,power_law_1.01,0.09792511940002442
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,4,4,power_law_1.01,0.05065983772277832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,4,4,power_law_1.01,6.862457885742188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,4,4,power_law_1.01,0.06348031997680664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,4,4,power_law_1.01,0.07008128166198731
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,4,4,power_law_1.2,0.08424960136413574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,4,4,power_law_1.2,0.0875443172454834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,4,4,power_law_1.2,0.15576959609985352
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,4,4,power_law_1.01,0.16273792266845705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,4,4,power_law_1.2,0.161397762298584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,4,4,power_law_1.2,0.16052480697631838
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,4,4,power_law_1.2,0.16278528213500976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,4,4,power_law_1.2,0.16530431747436522
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,4,4,power_law_1.2,0.16873472213745117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,4,4,power_law_1.2,0.17185152053833008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,4,4,power_law_1.2,0.17337215423583985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,4,4,power_law_1.2,0.17754112243652345
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,4,4,power_law_1.01,0.11649920463562011
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,4,4,power_law_1.2,0.1852787208557129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,4,4,power_law_1.2,0.22329727172851563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,4,4,power_law_1.2,0.19185663223266602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,4,4,power_law_1.2,0.2511372756958008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,4,4,power_law_1.2,0.27025407791137696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,4,4,power_law_1.2,0.3061004829406738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,4,4,power_law_1.2,0.38977664947509766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,4,4,power_law_1.2,0.44996990203857423
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,4,4,power_law_1.01,0.22565631866455077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,4,4,power_law_1.2,0.7403852844238281
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,4,4,power_law_1.01,0.31785600662231445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,4,4,power_law_1.2,0.6170406341552734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,4,4,power_law_1.2,1.4999859619140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,4,4,power_law_1.2,1.1429888153076173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,4,4,power_law_1.2,0.048671998977661134
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,4,4,power_law_1.2,0.05772672176361084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,4,4,power_law_1.2,0.06162816047668457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,4,4,power_law_1.2,0.06467199802398682
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,4,4,power_law_1.2,0.06530432224273681
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,4,4,power_law_1.2,0.06616064071655273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,4,4,power_law_1.2,0.06687615871429443
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,4,4,power_law_1.2,0.0681113576889038
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,4,4,power_law_1.2,0.06973440170288085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,4,4,power_law_1.2,0.07027071952819824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,4,4,power_law_1.2,0.07206272125244141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,4,4,power_law_1.2,0.07499263763427735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,4,4,power_law_1.2,0.07729919910430907
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,4,4,power_law_1.01,0.4021747207641601
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,4,4,power_law_1.2,0.08064895629882812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,4,4,power_law_1.2,0.08468095779418945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,4,4,power_law_1.2,0.09128704071044921
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,4,4,power_law_1.2,0.10621567726135255
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,4,4,power_law_1.2,0.13009663581848147
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,4,4,power_law_1.2,0.20535423278808596
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,4,4,power_law_1.2,2.1200230407714846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,4,4,power_law_1.2,0.1536012840270996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,4,4,power_law_1.2,0.25639551162719726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,4,4,power_law_1.2,4.483002929687499
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,4,4,power_law_1.01,1.0627289581298829
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,4,4,power_law_1.2,0.3496908950805664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,4,4,power_law_1.01,0.6831641387939453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,4,4,power_law_1.2,0.02607487916946411
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,4,4,power_law_1.2,0.4670809555053711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,4,4,power_law_1.2,0.026525440216064455
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,4,4,power_law_1.2,0.023425281047821045
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,4,4,power_law_1.2,0.02576128005981445
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,4,4,power_law_1.2,0.025566720962524415
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,4,4,power_law_1.2,0.025835518836975095
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,4,4,power_law_1.2,0.026177279949188233
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,4,4,power_law_1.2,0.025743360519409182
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,4,4,power_law_1.2,0.02645247936248779
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,4,4,power_law_1.2,0.026849279403686522
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,4,4,power_law_1.2,0.02749311923980713
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,4,4,power_law_1.2,0.03624959945678711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,4,4,power_law_1.2,0.6920563507080078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,4,4,power_law_1.2,0.8917926025390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,4,4,power_law_1.2,0.038835198879241944
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,4,4,power_law_1.2,0.04626304149627686
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,4,4,power_law_1.2,0.052101120948791504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,4,4,power_law_1.2,3.381688232421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,4,4,power_law_1.2,0.09619968414306641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,4,4,power_law_1.2,7.586767578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,4,4,power_law_1.2,1.8028941345214844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,4,4,power_law_1.2,1.4200845336914063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,4,4,power_law_1.2,0.21649919509887697
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,4,4,power_law_1.2,0.032206079959869384
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,4,4,power_law_1.2,0.0435481595993042
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,4,4,power_law_1.2,0.06573823928833009
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,4,4,power_law_1.2,0.07421567916870117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,8,2,balanced,0.037447679042816165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,8,2,balanced,0.08647808074951172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,8,2,balanced,0.135032320022583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,8,2,balanced,0.13587967872619627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,8,2,balanced,0.13697919845581055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,8,2,balanced,0.14030207633972167
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,8,2,balanced,0.1407923221588135
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,8,2,balanced,0.07234560012817383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,8,2,balanced,0.14285568237304686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,8,2,balanced,0.14416640281677245
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,8,2,balanced,0.1447539234161377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,8,2,balanced,0.14726143836975097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,8,2,balanced,0.14948863983154298
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,8,2,balanced,0.15484543800354006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,8,2,balanced,0.17244159698486328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,8,2,balanced,0.17796607971191408
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,8,2,balanced,0.1626688003540039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,8,2,balanced,0.2151513671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,8,2,balanced,0.25735424041748045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,8,2,balanced,0.34589439392089844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,8,2,balanced,0.4143027114868164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,8,2,balanced,0.5948812866210937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,8,2,balanced,0.761349105834961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,8,2,balanced,1.1350323486328127
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,4,4,power_law_1.2,0.11660032272338867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,8,2,balanced,1.5007936096191405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,8,2,balanced,2.251667175292969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,8,2,balanced,0.041587200164794925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,8,2,balanced,0.05563263893127442
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,8,2,balanced,0.05884160041809082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,8,2,balanced,0.05941887855529785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,8,2,balanced,0.06015103816986084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,8,2,balanced,3.012651672363281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,8,2,balanced,0.06097919940948486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,8,2,balanced,0.06456192016601563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,8,2,balanced,0.06284160137176513
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,8,2,balanced,0.06581120014190674
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,4,4,power_law_1.2,0.1672230339050293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,8,2,balanced,0.06626431941986084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,8,2,balanced,0.0704972791671753
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,8,2,balanced,0.0713100814819336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,8,2,balanced,0.07410943984985352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,8,2,balanced,0.07886975765228271
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,8,2,balanced,0.08279680252075196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,8,2,balanced,0.07686912059783936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,8,2,balanced,0.09603584289550782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,8,2,balanced,0.10439167976379396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,8,2,balanced,0.1437388801574707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,8,2,balanced,0.1670796775817871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,8,2,balanced,0.23671680450439453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,8,2,balanced,0.31343616485595704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,8,2,balanced,0.442314224243164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,8,2,balanced,0.5675904083251953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,8,2,balanced,0.8477209472656251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,8,2,balanced,1.1027903747558594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,8,2,balanced,0.03545344114303589
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,4,4,power_law_1.2,0.29815935134887694
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,4,4,power_law_1.2,0.41418113708496096
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,8,2,balanced,0.01986304044723511
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,8,2,balanced,0.029694719314575197
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,8,2,balanced,0.02960383892059326
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,8,2,balanced,0.030090239048004154
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,8,2,balanced,0.029268479347229003
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,8,2,balanced,0.029962239265441892
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,8,2,balanced,0.02070528030395508
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,8,2,balanced,0.027788798809051513
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,8,2,balanced,0.029273600578308107
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,8,2,balanced,0.03017215967178345
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,8,2,balanced,0.03091200113296509
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,8,2,balanced,0.031083519458770754
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,8,2,balanced,0.031306240558624264
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,8,2,balanced,0.03386751890182495
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,8,2,balanced,0.02842623949050903
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,4,4,power_law_1.2,0.5995558547973633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,4,4,power_law_1.2,0.8382015991210938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,8,2,balanced,0.029038081169128416
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,8,2,balanced,0.029474558830261233
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,8,2,balanced,0.04484992027282715
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,8,2,balanced,0.058408961296081544
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,8,2,balanced,0.0470527982711792
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,8,2,power_law_1.01,0.06162687778472901
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,8,2,power_law_1.01,0.06915711879730224
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,8,2,power_law_1.01,0.07578752040863038
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,8,2,power_law_1.01,0.13386240005493164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,8,2,power_law_1.01,0.1346623992919922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,8,2,balanced,0.07273856163024903
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,8,2,power_law_1.01,0.1355776023864746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,8,2,power_law_1.01,0.13965439796447754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,8,2,power_law_1.01,0.14098048210144043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,8,2,power_law_1.01,0.14268671989440918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,8,2,power_law_1.01,0.1441536045074463
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,8,2,power_law_1.01,0.14585472106933595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,8,2,power_law_1.01,0.1481049633026123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,8,2,power_law_1.01,0.15029760360717775
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,8,2,power_law_1.01,0.17001728057861326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,8,2,power_law_1.01,0.18450431823730468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,8,2,power_law_1.01,0.19750400543212893
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,8,2,power_law_1.01,0.22421888351440428
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,8,2,power_law_1.01,0.2930892753601074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,8,2,power_law_1.01,0.33152511596679685
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,8,2,balanced,0.09482624053955077
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,8,2,balanced,0.11644543647766112
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,8,2,power_law_1.01,0.45540737152099614
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,8,2,power_law_1.01,0.53631103515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,8,2,balanced,0.16345216751098632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,8,2,power_law_1.01,0.9627238464355468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,8,2,power_law_1.01,0.8389209747314453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,8,2,power_law_1.01,0.04418303966522217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,8,2,power_law_1.01,0.04702335834503174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,8,2,power_law_1.01,0.04844927787780762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,8,2,power_law_1.01,0.05964288234710693
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,8,2,power_law_1.01,0.0598796796798706
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,8,2,power_law_1.01,0.06023039817810059
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,8,2,power_law_1.01,0.06087679862976074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,8,2,power_law_1.01,0.06255104064941405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,8,2,power_law_1.01,0.06301568031311036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,8,2,power_law_1.01,0.06362751960754395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,8,2,power_law_1.01,0.06596735954284669
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,8,2,power_law_1.01,0.06882431983947754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,8,2,power_law_1.01,0.07093503952026367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,8,2,power_law_1.01,0.07473408222198487
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,8,2,power_law_1.01,0.08319487571716308
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,8,2,power_law_1.01,0.08766464233398438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,8,2,power_law_1.01,0.09035008430480958
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,8,2,power_law_1.01,1.4579443359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,8,2,power_law_1.01,0.11207679748535156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,8,2,power_law_1.01,0.1266419219970703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,8,2,power_law_1.01,0.151342077255249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,8,2,power_law_1.01,0.21587455749511716
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,8,2,power_law_1.01,0.2964531135559082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,8,2,power_law_1.01,3.5258203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,8,2,balanced,0.21093887329101563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,8,2,power_law_1.01,0.4806835174560547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,8,2,power_law_1.01,0.3447539138793945
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,8,2,power_law_1.01,0.022750720977783204
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,8,2,power_law_1.01,0.020677120685577394
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,8,2,power_law_1.01,0.02436608076095581
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,8,2,power_law_1.01,1.7646194458007813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,8,2,power_law_1.01,0.029407999515533446
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,8,2,power_law_1.01,0.026851840019226074
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,8,2,power_law_1.01,0.028412160873413084
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,8,2,power_law_1.01,0.028760321140289306
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,8,2,power_law_1.01,0.03014656066894531
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,8,2,power_law_1.01,0.029509119987487793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,8,2,power_law_1.01,0.02989952087402344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,8,2,power_law_1.01,0.030572800636291503
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,8,2,power_law_1.01,0.038926079273223876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,8,2,power_law_1.01,0.040052480697631836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,8,2,balanced,0.30942079544067386
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,8,2,balanced,0.41057792663574216
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,8,2,power_law_1.01,0.607083511352539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,8,2,power_law_1.01,3.214364013671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,8,2,power_law_1.01,0.9807833862304687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,8,2,power_law_1.01,1.338223419189453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,8,2,power_law_1.01,0.03854847908020019
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,8,2,power_law_1.01,0.04236800193786621
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,8,2,power_law_1.01,0.04261631965637207
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,8,2,power_law_1.01,0.04568064212799072
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,8,2,power_law_1.01,0.058178558349609374
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,8,2,power_law_1.01,0.06780799865722656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,8,2,power_law_1.01,0.07782015800476075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,8,2,power_law_1.2,0.0661030387878418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,8,2,power_law_1.2,0.06984831809997558
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,8,2,power_law_1.2,0.07532288074493408
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,8,2,power_law_1.2,0.13497599601745605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,8,2,power_law_1.2,0.1352012825012207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,8,2,power_law_1.2,0.13757311820983886
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,8,2,power_law_1.2,0.13920000076293945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,8,2,power_law_1.2,0.14116352081298827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,8,2,power_law_1.2,0.14388480186462402
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,8,2,power_law_1.2,0.1441152000427246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,8,2,power_law_1.2,0.14576000213623047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,8,2,power_law_1.2,0.1490982437133789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,8,2,power_law_1.2,0.15333632469177244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,8,2,power_law_1.2,0.16869119644165037
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,8,2,power_law_1.01,0.10553728103637697
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,8,2,power_law_1.2,0.190435848236084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,8,2,power_law_1.2,0.22952703475952146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,8,2,power_law_1.2,0.20648191452026365
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,8,2,power_law_1.2,0.3005209541320801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,8,2,power_law_1.2,0.3158182334899903
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,8,2,power_law_1.2,0.5826086425781251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,8,2,power_law_1.2,0.4555046463012696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,8,2,power_law_1.2,0.7647590637207031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,8,2,power_law_1.2,0.9028262329101562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,8,2,power_law_1.01,0.12351360321044921
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,8,2,power_law_1.01,0.1757030487060547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,8,2,power_law_1.01,0.2233638381958008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,8,2,power_law_1.2,0.04165247917175293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,8,2,power_law_1.2,0.045331201553344726
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,8,2,power_law_1.01,0.3135206413269043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,8,2,power_law_1.2,0.04845568180084229
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,8,2,power_law_1.2,0.0596608018875122
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,8,2,power_law_1.2,0.059022078514099116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,8,2,power_law_1.2,0.06134143829345703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,8,2,power_law_1.2,0.06229375839233399
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,8,2,power_law_1.2,0.06246655941009521
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,8,2,power_law_1.2,0.06354944229125976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,8,2,power_law_1.2,0.06486783981323242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,8,2,power_law_1.2,0.06630400180816651
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,8,2,power_law_1.2,0.06917119979858398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,8,2,power_law_1.2,0.07057407855987549
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,8,2,power_law_1.2,0.07448063850402832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,8,2,power_law_1.2,0.08105600357055665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,8,2,power_law_1.2,0.08824831962585448
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,8,2,power_law_1.2,0.09241855621337891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,8,2,power_law_1.2,1.4239897155761718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,8,2,power_law_1.2,0.11165056228637696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,8,2,power_law_1.2,0.12891648292541505
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,8,2,power_law_1.2,0.21601408004760742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,8,2,power_law_1.2,0.301977596282959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,8,2,power_law_1.2,0.16186752319335937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,8,2,power_law_1.01,0.6681049346923829
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,8,2,power_law_1.2,0.3312192153930664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,8,2,power_law_1.2,0.5017548751831055
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,8,2,power_law_1.2,0.022302720546722412
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,8,2,power_law_1.2,0.6965529632568359
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,8,2,power_law_1.2,0.021480960845947267
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,8,2,power_law_1.2,0.02844288110733032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,8,2,power_law_1.2,3.890936279296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,8,2,power_law_1.2,0.02716032028198242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,8,2,power_law_1.2,0.029528319835662842
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,8,2,power_law_1.2,0.028387839794158938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,8,2,power_law_1.2,0.02949631929397583
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,8,2,power_law_1.2,0.029341440200805664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,8,2,power_law_1.2,0.029845759868621823
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,8,2,power_law_1.2,0.030526719093322753
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,8,2,power_law_1.2,0.03885567903518677
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,8,2,power_law_1.2,0.039717121124267576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,8,2,power_law_1.2,1.5965811157226564
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,8,2,power_law_1.2,0.043022079467773436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,8,2,power_law_1.2,3.1479437255859373
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,8,2,power_law_1.01,0.5193920135498047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,8,2,power_law_1.2,0.06433792114257812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,8,2,power_law_1.2,1.4184512329101562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,8,2,power_law_1.2,1.0286566162109376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,8,2,power_law_1.2,0.024200959205627443
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,8,2,power_law_1.2,0.03843712091445923
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,8,2,power_law_1.2,0.04622719764709472
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,8,2,power_law_1.2,0.0431987190246582
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,8,2,power_law_1.2,0.05685247898101806
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,8,2,power_law_1.2,0.08566911697387695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,16,1,balanced,0.0386080002784729
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,16,1,balanced,0.07281919956207275
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,16,1,balanced,0.08928447723388672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,16,1,balanced,0.09029184341430664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,16,1,balanced,0.12896575927734374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,16,1,balanced,0.1312351989746094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,16,1,balanced,0.13285887718200684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,16,1,balanced,0.13389504432678223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,16,1,balanced,0.13442367553710938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,16,1,balanced,0.13382399559020997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,16,1,balanced,0.13607487678527833
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,16,1,balanced,0.13758272171020508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,16,1,balanced,0.13910079956054688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,16,1,balanced,0.14434752464294434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,16,1,balanced,0.15938624382019043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,16,1,balanced,0.1622060775756836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,16,1,balanced,0.1624064064025879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,16,1,balanced,0.18610559463500978
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,8,2,power_law_1.2,0.10592639923095704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,16,1,balanced,0.22372608184814452
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,16,1,balanced,0.3261676788330078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,16,1,balanced,0.39572414398193356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,16,1,balanced,0.5601363372802735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,16,1,balanced,0.7250323486328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,16,1,balanced,1.0709951782226563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,16,1,balanced,1.446588134765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,16,1,balanced,2.187804870605469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,16,1,balanced,0.04065152168273926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,16,1,balanced,0.04472064018249512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,16,1,balanced,0.06407616138458253
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,16,1,balanced,0.06386752128601074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,16,1,balanced,0.0689900779724121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,16,1,balanced,0.07001920223236084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,16,1,balanced,0.07127423763275147
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,16,1,balanced,0.07130943775177002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,16,1,balanced,2.993870849609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,16,1,balanced,0.07176576137542726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,16,1,balanced,0.07708415985107422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,16,1,balanced,0.07401343822479248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,16,1,balanced,0.07737855911254883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,16,1,balanced,0.0770636796951294
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,16,1,balanced,0.08073599815368653
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,16,1,balanced,0.08195712089538573
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,16,1,balanced,0.08557824134826661
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,16,1,balanced,0.08371711730957031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,16,1,balanced,0.09935040473937988
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,16,1,balanced,0.10569600105285644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,16,1,balanced,0.1478271961212158
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,16,1,balanced,0.1724723243713379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,16,1,balanced,0.2422310447692871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,16,1,balanced,0.3135219192504883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,16,1,balanced,0.45348545074462887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,16,1,balanced,0.5708883285522461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,8,2,power_law_1.2,0.15501952171325684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,16,1,balanced,0.8506893157958985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,16,1,balanced,1.105944290161133
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,8,2,power_law_1.2,0.1856051254272461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,8,2,power_law_1.2,0.21495552062988282
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,8,2,power_law_1.2,0.4793446350097656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,8,2,power_law_1.2,0.3485184097290039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,16,1,balanced,0.01976191997528076
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,16,1,balanced,0.032439680099487306
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,16,1,balanced,0.03225919961929321
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,16,1,balanced,0.03253504037857056
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,16,1,balanced,0.032447359561920165
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,16,1,balanced,0.03278527975082397
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,16,1,balanced,0.032707200050354004
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,16,1,balanced,0.022193920612335206
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,16,1,balanced,0.03510848045349121
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,16,1,balanced,0.035490560531616214
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,16,1,balanced,0.0356275200843811
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,16,1,balanced,0.03831552028656006
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,16,1,balanced,0.03052864074707031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,16,1,balanced,0.030941441059112552
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,16,1,balanced,0.0312774395942688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,16,1,balanced,0.031495680809021
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,8,2,power_law_1.2,0.6867788696289063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,16,1,balanced,0.03088831901550293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,16,1,balanced,0.0509555196762085
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,16,1,balanced,0.05424384117126465
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,16,1,balanced,0.06990655899047851
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,16,1,power_law_1.01,0.06230783939361573
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,16,1,power_law_1.01,0.07098559856414795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,16,1,power_law_1.01,0.0752019214630127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,16,1,power_law_1.01,0.08759679794311523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,16,1,power_law_1.01,0.12984895706176758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,16,1,power_law_1.01,0.1298560047149658
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,16,1,power_law_1.01,0.13119935989379883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,16,1,power_law_1.01,0.13182144165039061
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,16,1,power_law_1.01,0.13276288032531738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,16,1,power_law_1.01,0.13541824340820313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,16,1,power_law_1.01,0.14364735603332518
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,16,1,power_law_1.01,0.14753600120544436
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,16,1,balanced,0.08534015655517578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,16,1,power_law_1.01,0.1513145637512207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,16,1,power_law_1.01,0.15605759620666504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,16,1,power_law_1.01,0.15994048118591309
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,16,1,power_law_1.01,0.16958335876464842
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,16,1,power_law_1.01,0.17878591537475585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,16,1,power_law_1.01,0.22445119857788084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,16,1,power_law_1.01,0.2938092803955078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,16,1,power_law_1.01,0.35979328155517576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,16,1,power_law_1.01,0.4428915023803711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,16,1,balanced,0.11339776039123535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,16,1,power_law_1.01,0.6053388977050782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,16,1,power_law_1.01,0.7688934326171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,16,1,balanced,0.14100031852722167
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,16,1,balanced,0.20214719772338868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,16,1,power_law_1.01,0.04374527931213379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,16,1,power_law_1.01,0.050405759811401364
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,16,1,power_law_1.01,0.052871041297912605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,16,1,power_law_1.01,1.1137420654296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,16,1,power_law_1.01,0.07011903762817383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,16,1,power_law_1.01,0.0699609613418579
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,16,1,power_law_1.01,0.07164608001708984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,16,1,power_law_1.01,0.07147327899932862
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,16,1,power_law_1.01,0.07226816177368164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,16,1,balanced,0.2630259132385254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,16,1,power_law_1.01,0.07335807800292968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,16,1,power_law_1.01,0.0751692819595337
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,16,1,power_law_1.01,0.07560512065887451
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,16,1,power_law_1.01,0.07864768028259277
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,16,1,power_law_1.01,0.08040191650390624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,16,1,power_law_1.01,0.08347007751464844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,16,1,power_law_1.01,0.085415678024292
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,16,1,power_law_1.01,0.08913791656494141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,16,1,power_law_1.01,0.0990719985961914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,16,1,power_law_1.01,0.13229311943054198
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,16,1,power_law_1.01,0.15311360359191895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,16,1,power_law_1.01,0.06454847812652588
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,16,1,power_law_1.01,1.4598956298828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,16,1,balanced,0.38757312774658204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,16,1,power_law_1.01,0.18490495681762695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,16,1,balanced,0.5244025421142579
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,16,1,power_law_1.01,0.24923839569091796
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,16,1,power_law_1.01,0.02056512117385864
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,16,1,power_law_1.01,0.024644479751586915
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,16,1,power_law_1.01,0.025002241134643555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,16,1,power_law_1.01,0.0295961594581604
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,16,1,power_law_1.01,0.03080895900726318
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,16,1,power_law_1.01,0.32126655578613283
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,16,1,power_law_1.01,0.031166720390319824
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,16,1,power_law_1.01,0.030903680324554445
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,16,1,power_law_1.01,0.031062400341033934
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,16,1,power_law_1.01,0.03205631971359253
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,16,1,power_law_1.01,0.032828800678253174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,16,1,power_law_1.01,0.4589510345458985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,16,1,power_law_1.01,2.179893798828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,16,1,power_law_1.01,0.597625617980957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,16,1,power_law_1.01,2.8963699340820312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,16,1,power_law_1.01,0.8618208312988281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,16,1,power_law_1.01,1.1303635406494141
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,16,1,power_law_1.01,0.031491200923919674
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,16,1,power_law_1.01,0.03332927942276001
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,16,1,power_law_1.01,0.0351308798789978
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,16,1,power_law_1.01,0.0866483211517334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,16,1,power_law_1.01,0.04091072082519531
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,16,1,power_law_1.01,0.04289087772369385
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,16,1,power_law_1.01,0.04393472194671631
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,16,1,power_law_1.01,0.04854207992553711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,16,1,power_law_1.01,0.056897277832031246
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,16,1,power_law_1.01,0.06883456230163573
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,16,1,power_law_1.2,0.06309055805206298
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,16,1,power_law_1.2,0.07110720157623292
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,16,1,power_law_1.2,0.07529088020324706
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,16,1,power_law_1.01,0.1028435230255127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,16,1,power_law_1.2,0.1278873634338379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,16,1,power_law_1.2,0.08951744079589843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,16,1,power_law_1.2,0.1311359977722168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,16,1,power_law_1.2,0.13278656005859374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,16,1,power_law_1.2,0.13279104232788086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,16,1,power_law_1.2,0.13468095779418945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,16,1,power_law_1.2,0.13537983894348143
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,16,1,power_law_1.2,0.14501248359680177
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,16,1,power_law_1.01,0.1413644790649414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,16,1,power_law_1.2,0.148220796585083
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,16,1,power_law_1.2,0.15154623985290527
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,16,1,power_law_1.2,0.1574963188171387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,16,1,power_law_1.2,0.16096511840820313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,16,1,power_law_1.2,0.17153215408325195
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,16,1,power_law_1.01,0.17904703140258788
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,16,1,power_law_1.2,0.18045055389404296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,16,1,power_law_1.2,0.22430912017822266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,16,1,power_law_1.2,0.29713663101196286
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,16,1,power_law_1.2,0.3618508911132813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,16,1,power_law_1.2,0.44259265899658207
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,16,1,power_law_1.01,0.25344127655029297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,16,1,power_law_1.2,0.6079225540161133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,16,1,power_law_1.2,0.7705913543701172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,16,1,power_law_1.2,0.04158207893371582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,16,1,power_law_1.2,0.04943999767303467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,16,1,power_law_1.2,0.050355200767517094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,16,1,power_law_1.2,0.06296127796173095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,16,1,power_law_1.2,0.06798975944519044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,16,1,power_law_1.2,0.06974976062774658
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,16,1,power_law_1.2,0.0710591983795166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,16,1,power_law_1.2,0.07185088157653809
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,16,1,power_law_1.2,0.07316415786743165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,16,1,power_law_1.2,0.07359936237335205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,16,1,power_law_1.2,0.07475391864776612
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,16,1,power_law_1.2,0.07635327816009521
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,16,1,power_law_1.2,0.0775878381729126
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,16,1,power_law_1.01,0.3317958450317383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,16,1,power_law_1.2,0.08098176002502441
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,16,1,power_law_1.2,0.08236800193786621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,16,1,power_law_1.2,0.08616448402404785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,16,1,power_law_1.2,0.08989888191223144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,16,1,power_law_1.2,0.09931391716003418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,16,1,power_law_1.2,0.13235008239746093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,16,1,power_law_1.2,1.1075360107421877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,16,1,power_law_1.2,0.153056640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,16,1,power_law_1.2,0.18496383666992186
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,16,1,power_law_1.01,0.48684543609619146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,16,1,power_law_1.2,0.2494771194458008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,16,1,power_law_1.2,1.453651885986328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,16,1,power_law_1.2,0.3203513717651367
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,16,1,power_law_1.2,0.02014847993850708
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,16,1,power_law_1.2,0.023812479972839355
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,16,1,power_law_1.2,0.024824960231781008
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,16,1,power_law_1.2,0.029920001029968262
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,16,1,power_law_1.2,0.029831039905548095
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,16,1,power_law_1.2,0.03131711959838867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,16,1,power_law_1.2,0.031303040981292725
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,16,1,power_law_1.2,0.031020801067352295
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,16,1,power_law_1.2,0.031836159229278564
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,16,1,power_law_1.2,0.03252928018569946
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,16,1,power_law_1.2,0.032904319763183594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,16,1,power_law_1.2,0.03401087999343872
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,16,1,power_law_1.2,0.035139200687408445
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,16,1,power_law_1.2,0.04042943954467774
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,16,1,power_law_1.01,0.64080322265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,16,1,power_law_1.2,0.46025150299072265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,16,1,power_law_1.2,2.17495361328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,16,1,power_law_1.2,0.5971750259399414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,16,1,power_law_1.2,0.04435647964477539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,16,1,power_law_1.2,0.859183349609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,16,1,power_law_1.2,0.04592895984649658
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,16,1,power_law_1.2,0.050216960906982425
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,16,1,power_law_1.2,2.8756927490234374
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,16,1,power_law_1.2,0.05747903823852539
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,16,1,power_law_1.2,0.06976319789886475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,16,1,power_law_1.2,1.1310066986083984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,16,1,power_law_1.2,0.08807295799255371
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,16,1,power_law_1.2,0.10597887992858887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,16,balanced,0.05377408027648926
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,16,1,power_law_1.2,0.14566207885742188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,16,balanced,0.052906241416931156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,16,balanced,0.051060481071472166
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,16,1,power_law_1.2,0.18474752426147462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,16,balanced,0.0759552001953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,16,balanced,0.2526643180847168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,16,balanced,0.25833984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,16,balanced,0.24623487472534178
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,16,balanced,0.13445631980895995
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,16,1,power_law_1.2,0.2630240058898926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,16,balanced,0.25107072830200194
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,16,balanced,0.25805952072143556
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,16,balanced,0.25724416732788086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,16,balanced,0.26228992462158207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,16,balanced,0.2667737579345703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,16,balanced,0.2748480033874512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,16,balanced,0.35959678649902344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,16,balanced,0.25462015151977535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,16,balanced,0.26240512847900394
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,16,1,power_law_1.2,0.5016774368286133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,16,balanced,0.46617214202880863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,16,balanced,0.2656972885131836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,16,balanced,0.31710464477539063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,16,balanced,0.34848384857177733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,16,balanced,0.4204812622070312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,16,balanced,0.03911295890808105
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,16,1,power_law_1.2,0.3427481460571289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,16,balanced,0.03911551952362061
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,16,balanced,0.046659841537475585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,16,balanced,0.061644802093505854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,16,balanced,0.7360345458984374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,16,balanced,0.08782464027404785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,16,balanced,0.155918083190918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,16,balanced,0.17998847961425782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,16,balanced,0.16249216079711915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,16,balanced,0.16749696731567382
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,16,balanced,0.1663590431213379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,16,balanced,0.17021440505981444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,16,balanced,0.17281024932861327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,16,balanced,0.17513216018676758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,16,balanced,0.16846208572387694
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,16,balanced,0.17248512268066407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,16,balanced,0.1743462371826172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,16,balanced,0.17913728713989258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,16,balanced,0.18895231246948244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,16,balanced,0.207696647644043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,16,balanced,0.23777151107788086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,16,balanced,0.25610496520996096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,16,balanced,0.30882944107055665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,16,balanced,0.3582783889770508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,16,balanced,0.5119948959350585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,16,balanced,0.6156185531616212
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,16,balanced,0.8672652435302733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,16,balanced,1.116578598022461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,16,balanced,0.593120002746582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,16,balanced,0.6639308929443359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,16,balanced,1.0385113525390626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,16,balanced,1.174856948852539
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,1,16,balanced,0.022749440670013426
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,1,16,balanced,0.02269695997238159
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,1,16,balanced,0.0920576000213623
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,1,16,balanced,0.09245311737060546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,1,16,balanced,0.09182208061218262
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,1,16,balanced,0.09227007865905762
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,1,16,balanced,0.02441472053527832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,1,16,balanced,0.093057279586792
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,1,16,balanced,0.0924287986755371
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,1,16,balanced,0.09321215629577637
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,1,16,balanced,0.09292799949645995
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,1,16,balanced,0.03599488019943237
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,1,16,balanced,0.09266688346862792
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,1,16,balanced,0.09306112289428711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,1,16,balanced,0.056801280975341796
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,1,16,balanced,0.0951859188079834
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,1,16,balanced,0.09512703895568848
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,1,16,balanced,0.09672703742980956
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,1,16,balanced,0.09838591575622559
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,1,16,balanced,0.10359295845031738
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,1,16,balanced,0.10631168365478516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,16,1,power_law_1.2,0.6644812774658203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,16,power_law_1.01,0.13318143844604494
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,1,16,balanced,0.09171199798583984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,1,16,balanced,0.0925055980682373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,16,power_law_1.01,0.2460313606262207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,16,power_law_1.01,0.23024511337280273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,16,power_law_1.01,0.23510528564453126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,16,power_law_1.01,0.21312255859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,16,power_law_1.01,0.24032255172729494
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,16,power_law_1.01,0.21603071212768557
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,16,power_law_1.01,0.22890623092651366
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,16,power_law_1.01,0.23214591979980467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,16,power_law_1.01,0.24703487396240234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,16,power_law_1.01,0.2444851112365723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,16,power_law_1.01,0.3031923294067383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,16,power_law_1.01,0.2445235252380371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,16,power_law_1.01,0.30999935150146485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,16,power_law_1.01,0.3356953430175781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,1,16,balanced,0.14312447547912596
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,16,power_law_1.01,0.24968191146850588
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,1,16,balanced,0.20805120468139648
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,1,16,balanced,0.25929471969604495
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,1,16,balanced,0.15366527557373047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,16,power_law_1.01,0.23777023315429688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,16,power_law_1.01,0.32225536346435546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,16,power_law_1.01,0.431196174621582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,16,power_law_1.01,0.5599001693725586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,16,power_law_1.01,0.08635775566101074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,16,power_law_1.01,0.14703231811523437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,16,power_law_1.01,0.7365401458740235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,16,power_law_1.01,0.14517248153686524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,16,power_law_1.01,0.14872063636779784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,16,power_law_1.01,0.3489344024658203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,16,power_law_1.01,0.15201919555664062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,16,power_law_1.01,0.13438719749450684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,16,power_law_1.01,0.14806528091430665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,16,power_law_1.01,0.1490726375579834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,16,power_law_1.01,0.15288960456848144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,16,power_law_1.01,0.1537600040435791
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,16,power_law_1.01,0.15475199699401856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,16,power_law_1.01,0.15821184158325194
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,16,power_law_1.01,0.16305280685424806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,16,power_law_1.01,0.1564095973968506
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,16,power_law_1.01,0.16516735076904296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,16,power_law_1.01,0.16961919784545898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,16,power_law_1.01,0.17599103927612306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,16,power_law_1.01,0.1957606315612793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,16,power_law_1.01,0.21668224334716796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,16,power_law_1.01,0.27042943954467774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,16,power_law_1.01,0.3031091117858887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,16,power_law_1.01,0.39399166107177735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,16,power_law_1.01,0.509351692199707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,16,power_law_1.01,0.6560332489013672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,16,power_law_1.01,0.9365235137939454
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,1,16,power_law_1.01,0.056363520622253416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,16,power_law_1.01,1.197391357421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,1,16,power_law_1.01,0.09290111541748047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,1,16,power_law_1.01,0.09085184097290039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,1,16,power_law_1.01,0.09172608375549315
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,1,16,power_law_1.01,0.0920691204071045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,16,power_law_1.01,1.771814422607422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,1,16,power_law_1.01,0.0758899211883545
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,1,16,power_law_1.01,0.08187392234802246
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,1,16,power_law_1.01,0.08015104293823241
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,1,16,power_law_1.01,0.08105855941772461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,1,16,power_law_1.01,0.08024191856384277
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,1,16,power_law_1.01,0.08376704216003418
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,1,16,power_law_1.01,0.08454912185668946
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,1,16,power_law_1.01,0.085415678024292
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,1,16,power_law_1.01,0.08609024047851563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,16,power_law_1.01,1.4158323669433595
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,1,16,power_law_1.01,0.08741120338439942
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,16,power_law_1.01,0.7372723388671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,1,16,power_law_1.01,0.09346943855285644
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,1,16,power_law_1.01,0.08416640281677246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,16,power_law_1.01,2.008733367919922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,16,power_law_1.01,1.0323929595947265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,16,power_law_1.01,0.9416934204101561
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,1,16,power_law_1.01,0.09848575592041016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,1,16,power_law_1.01,0.10331520080566406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,1,16,power_law_1.01,0.11856639862060547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,1,16,power_law_1.01,0.1354419231414795
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,1,16,power_law_1.01,0.17462144851684572
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,16,power_law_1.2,0.19306880950927735
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,1,16,power_law_1.01,0.19149696350097656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,16,power_law_1.2,0.13355648040771484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,1,16,power_law_1.01,0.29929088592529296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,16,power_law_1.2,0.24230783462524413
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,1,16,power_law_1.01,0.3161676788330078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,16,power_law_1.2,0.23963903427124023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,16,power_law_1.2,0.2116326332092285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,16,power_law_1.2,0.22469503402709962
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,16,power_law_1.2,0.186060791015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,16,power_law_1.2,0.20357759475708007
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,16,power_law_1.2,0.25900032043457033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,16,power_law_1.2,0.25277439117431644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,16,power_law_1.2,0.3069670486450195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,16,power_law_1.2,0.3219276809692383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,16,power_law_1.2,0.22925184249877928
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,16,power_law_1.2,0.366959342956543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,16,power_law_1.2,0.31147647857666017
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,16,power_law_1.2,0.23720575332641602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,16,power_law_1.2,0.5121599960327148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,16,power_law_1.2,0.24476543426513672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,16,power_law_1.2,0.7699008178710938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,16,power_law_1.2,1.0525004577636718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,16,power_law_1.2,0.2548108863830566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,16,power_law_1.2,1.3141976928710937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,16,power_law_1.2,0.08603263854980468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,16,power_law_1.2,0.14716416358947754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,16,power_law_1.2,0.12584320068359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,16,power_law_1.2,0.14658687591552735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,16,power_law_1.2,1.785479736328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,16,power_law_1.2,0.14740608215332032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,16,power_law_1.2,0.4161369705200196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,16,power_law_1.2,0.14594304084777832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,16,power_law_1.2,0.14891519546508789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,16,power_law_1.2,0.14715264320373536
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,16,power_law_1.2,0.155097599029541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,16,power_law_1.2,0.1499891185760498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,16,power_law_1.2,0.15823488235473632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,16,power_law_1.2,0.33728641510009766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,16,power_law_1.2,0.16540288925170898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,16,power_law_1.2,0.16716672897338866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,16,power_law_1.2,0.16364927291870118
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,16,power_law_1.2,0.1698611259460449
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,16,power_law_1.2,0.1842995262145996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,16,power_law_1.2,0.16435327529907226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,16,power_law_1.2,0.2141926383972168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,16,power_law_1.2,0.260183048248291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,16,power_law_1.2,0.2999308776855469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,16,power_law_1.2,0.3719359970092774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,16,power_law_1.2,0.4386483383178711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,16,power_law_1.2,0.6225331115722657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,16,power_law_1.2,0.7876172637939453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,1,16,power_law_1.01,0.5384230422973633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,1,16,power_law_1.01,0.6845209503173828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,1,16,power_law_1.2,0.05596543788909912
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,16,power_law_1.2,1.0733683013916016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,1,16,power_law_1.2,0.0822771167755127
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,1,16,power_law_1.2,0.0919155216217041
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,1,16,power_law_1.2,0.09116031646728515
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,1,16,power_law_1.2,0.08743807792663574
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,1,16,power_law_1.2,0.08493311882019043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,16,power_law_1.2,1.7162725830078123
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,1,16,power_law_1.2,0.08099072456359864
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,1,16,power_law_1.2,0.07906432151794433
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,1,16,power_law_1.2,0.07754752159118652
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,1,16,power_law_1.2,0.08543871879577637
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,1,16,power_law_1.2,0.08691840171813965
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,1,16,power_law_1.2,0.08478976249694824
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,1,16,power_law_1.2,0.08911999702453613
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,16,power_law_1.2,2.1829927062988284
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,1,16,power_law_1.2,0.09461631774902343
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,1,16,power_law_1.2,0.07313024044036866
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,1,16,power_law_1.2,0.08798208236694335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,16,power_law_1.2,2.3733068847656247
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,16,power_law_1.2,0.6274342346191406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,1,16,power_law_1.2,0.10616319656372071
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,1,16,power_law_1.2,0.11876992225646972
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,1,16,power_law_1.2,0.09636096000671387
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,1,16,power_law_1.2,0.138603515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,1,16,power_law_1.2,0.16708608627319338
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,8,balanced,0.03585920095443725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,8,balanced,0.06307456016540527
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,8,balanced,0.07735680103302002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,8,balanced,0.09849599838256837
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,8,balanced,0.15319552421569824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,8,balanced,0.2992038345336914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,1,16,power_law_1.2,0.20602880477905275
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,8,balanced,0.3734515380859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,8,balanced,0.376577262878418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,8,balanced,0.32268798828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,8,balanced,0.3658380889892578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,8,balanced,0.3054297637939453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,8,balanced,0.35635200500488284
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,8,balanced,0.40153343200683594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,8,balanced,0.40790271759033203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,8,balanced,0.3586265563964844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,8,balanced,0.41177089691162105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,8,balanced,0.3569689559936523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,8,balanced,0.47615486145019525
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,8,balanced,0.43460094451904296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,8,balanced,0.5165631866455078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,8,balanced,0.6056524658203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,8,balanced,0.8207027435302734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,8,balanced,0.9772684478759766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,8,balanced,1.380517120361328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,1,16,power_law_1.2,0.26485376358032225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,8,balanced,0.03718015909194946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,8,balanced,1.766209259033203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,8,balanced,0.03933439970016479
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,8,balanced,0.05412864208221435
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,8,balanced,0.06901504039764404
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,8,balanced,0.10179583549499512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,8,balanced,0.16415359497070312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,8,balanced,0.17257984161376955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,8,balanced,0.1772108840942383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,8,balanced,2.605895690917969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,8,balanced,0.18283264160156248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,8,balanced,0.17215232849121093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,8,balanced,0.176309757232666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,8,balanced,0.17802879333496094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,8,balanced,0.1822643280029297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,8,balanced,0.18544767379760743
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,8,balanced,0.19241600036621093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,8,balanced,0.20905855178833005
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,8,balanced,0.23121023178100586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,8,balanced,0.2517184066772461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,8,balanced,0.2927948760986328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,8,balanced,3.72531982421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,8,balanced,0.3342655944824219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,8,balanced,0.4238079833984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,8,balanced,0.5089958572387696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,8,balanced,0.17887872695922852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,8,balanced,0.7281664276123048
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,8,balanced,0.9001407623291016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,8,balanced,1.3083750915527343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,8,balanced,1.6915621948242188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,1,16,power_law_1.2,0.39059967041015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,1,16,power_law_1.2,0.4591872024536133
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,1,16,power_law_1.2,0.7150502777099609
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,1,16,power_law_1.2,0.9773900604248047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,2,8,balanced,0.02390399932861328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,2,8,balanced,0.09580544471740723
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,2,8,balanced,0.02475519895553589
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,2,8,balanced,0.09660544395446777
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,2,8,balanced,0.09608063697814942
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,2,8,balanced,0.09641728401184083
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,2,8,balanced,0.0962611198425293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,2,8,balanced,0.0273523211479187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,2,8,balanced,0.09724160194396972
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,2,8,balanced,0.0982476806640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,2,8,balanced,0.09819392204284669
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,2,8,balanced,0.0993075180053711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,2,8,balanced,0.10166144371032715
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,2,8,balanced,0.1032870388031006
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,2,8,balanced,0.10590335845947266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,2,8,balanced,0.10907135963439943
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,2,8,balanced,0.11786239624023438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,2,8,balanced,0.12329600334167481
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,2,8,balanced,0.03741055965423584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,2,8,balanced,0.057644801139831545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,8,power_law_1.01,0.09727359771728515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,8,power_law_1.01,0.13946240425109863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,8,power_law_1.01,0.166376953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,8,power_law_1.01,0.18364288330078124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,8,power_law_1.01,0.22399616241455078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,8,power_law_1.01,0.23524095535278322
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,8,power_law_1.01,0.2634956741333008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,8,power_law_1.01,0.25598079681396485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,8,power_law_1.01,0.2813670349121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,8,power_law_1.01,0.27262975692749025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,8,power_law_1.01,0.31351423263549805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,8,power_law_1.01,0.2931033515930176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,8,power_law_1.01,0.27420032501220704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,8,power_law_1.01,0.29129600524902344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,8,power_law_1.01,0.30667135238647464
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,8,power_law_1.01,0.33513473510742187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,8,power_law_1.01,0.3854399871826172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,8,power_law_1.01,0.433436164855957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,8,power_law_1.01,0.47257984161376954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,8,power_law_1.01,0.5476121520996093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,8,power_law_1.01,0.6913830566406249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,8,power_law_1.01,0.8658432006835938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,8,power_law_1.01,1.1443430328369142
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,8,power_law_1.01,1.5384294128417968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,2,8,balanced,0.09564543724060058
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,2,8,balanced,0.09526656150817872
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,2,8,balanced,0.09549695968627929
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,8,power_law_1.01,1.8728370666503906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,8,power_law_1.01,0.061976318359375004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,8,power_law_1.01,0.09076607704162598
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,8,power_law_1.01,0.10889216423034669
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,8,power_law_1.01,0.12338687896728515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,8,power_law_1.01,0.1424563217163086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,8,power_law_1.01,0.13302783966064452
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,8,power_law_1.01,0.15162367820739747
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,8,power_law_1.01,0.15076095581054688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,8,power_law_1.01,0.15600640296936036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,8,power_law_1.01,0.16446720123291017
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,8,power_law_1.01,0.16671871185302733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,8,power_law_1.01,0.1671552085876465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,8,power_law_1.01,0.16405887603759767
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,8,power_law_1.01,0.17022335052490234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,8,power_law_1.01,3.770814819335938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,8,power_law_1.01,0.17256704330444336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,8,power_law_1.01,0.17357311248779297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,8,power_law_1.01,0.18727807998657225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,8,power_law_1.01,0.2098419189453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,8,power_law_1.01,0.2376166343688965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,8,power_law_1.01,0.29532800674438475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,8,power_law_1.01,0.33390464782714846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,8,power_law_1.01,0.4517926406860352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,8,power_law_1.01,0.5180364990234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,8,power_law_1.01,0.7191462707519531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,8,power_law_1.01,0.8875993347167969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,2,8,power_law_1.01,0.039921920299530025
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,2,8,power_law_1.01,0.058501119613647456
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,8,power_law_1.01,1.3308595275878905
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,2,8,power_law_1.01,0.07015552043914794
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,2,8,power_law_1.01,0.07245440006256103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,8,power_law_1.01,3.096192016601562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,2,8,power_law_1.01,0.08165504455566405
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,2,8,power_law_1.01,0.08753536224365235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,8,power_law_1.01,1.716106262207031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,2,8,power_law_1.01,0.08203007698059081
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,2,8,power_law_1.01,0.08863360404968261
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,2,8,power_law_1.01,0.08941951751708985
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,2,8,power_law_1.01,0.08849791526794434
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,2,8,power_law_1.01,0.08454527854919433
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,2,8,power_law_1.01,0.09034879684448242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,2,8,power_law_1.01,0.09235584259033203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,2,8,power_law_1.01,0.09076095581054687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,2,8,power_law_1.01,0.0925875186920166
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,2,8,power_law_1.01,0.09653375625610353
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,2,8,power_law_1.01,0.09788288116455078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,2,8,balanced,0.17016576766967773
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,2,8,balanced,0.26007680892944335
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,2,8,balanced,0.18612096786499024
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,2,8,balanced,0.33215358734130856
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,2,8,power_law_1.01,0.10806655883789062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,2,8,power_law_1.01,0.11722368240356444
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,2,8,power_law_1.01,0.15366016387939455
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,2,8,power_law_1.01,0.1326476764678955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,8,power_law_1.2,0.1409177589416504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,8,power_law_1.2,0.13245439529418945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,8,power_law_1.2,0.10059904098510741
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,8,power_law_1.2,0.17612159729003907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,8,power_law_1.2,0.22570623397827147
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,8,power_law_1.2,0.20244735717773438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,2,8,power_law_1.01,0.19910144805908203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,8,power_law_1.2,0.2548134422302246
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,2,8,power_law_1.01,0.2360051155090332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,8,power_law_1.2,0.2402572822570801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,8,power_law_1.2,0.2718604850769043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,2,8,power_law_1.01,0.33682559967041015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,8,power_law_1.2,0.29017215728759765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,8,power_law_1.2,0.28803583145141604
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,8,power_law_1.2,0.27647743225097654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,8,power_law_1.2,0.29280895233154297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,8,power_law_1.2,0.2947929573059082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,8,power_law_1.2,0.3243123245239258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,8,power_law_1.2,0.3680012893676758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,8,power_law_1.2,0.41079040527343746
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,2,8,power_law_1.01,0.415728645324707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,8,power_law_1.2,0.4292671966552734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,8,power_law_1.2,0.4735871887207031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,8,power_law_1.2,0.6288127899169922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,8,power_law_1.2,0.7083097839355469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,8,power_law_1.2,1.0044544219970704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,8,power_law_1.2,1.113966064453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,8,power_law_1.2,0.06132480144500733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,8,power_law_1.2,0.08829824447631836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,8,power_law_1.2,1.6375360107421877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,8,power_law_1.2,0.08498175621032715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,8,power_law_1.2,0.12100735664367676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,8,power_law_1.2,0.13893631935119627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,8,power_law_1.2,0.12159104347229004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,8,power_law_1.2,2.1795135498046876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,8,power_law_1.2,0.1588582420349121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,8,power_law_1.2,0.15052160263061523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,8,power_law_1.2,0.15502335548400878
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,8,power_law_1.2,0.1627110481262207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,8,power_law_1.2,0.16673408508300783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,8,power_law_1.2,0.167957763671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,8,power_law_1.2,0.16331520080566406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,8,power_law_1.2,0.16978431701660157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,8,power_law_1.2,0.17382144927978516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,8,power_law_1.2,0.17640447616577148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,8,power_law_1.2,0.19095424652099607
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,8,power_law_1.2,0.21512191772460937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,8,power_law_1.2,3.3297601318359376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,8,power_law_1.2,0.24850048065185545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,8,power_law_1.2,0.3144166374206543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,8,power_law_1.2,0.3460211181640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,8,power_law_1.2,0.46392574310302737
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,8,power_law_1.2,0.5209523010253906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,2,8,power_law_1.2,0.040028162002563476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,8,power_law_1.2,0.8379609680175781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,2,8,power_law_1.2,0.05645823955535889
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,8,power_law_1.2,1.0850713348388672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,2,8,power_law_1.2,0.05346047878265381
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,8,power_law_1.2,4.38738037109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,2,8,power_law_1.2,0.07345151901245117
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,2,8,power_law_1.2,0.08344191551208496
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,2,8,power_law_1.2,0.07889535903930664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,2,8,power_law_1.2,0.08093055725097656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,8,power_law_1.2,1.5077708435058592
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,2,8,power_law_1.2,0.08844287872314453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,2,8,power_law_1.2,0.0900774383544922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,2,8,power_law_1.2,0.08840576171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,2,8,power_law_1.2,0.08889344215393066
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,2,8,power_law_1.2,0.08856320381164551
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,2,8,power_law_1.2,0.09591423988342285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,2,8,power_law_1.2,0.09117952346801758
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,2,8,power_law_1.2,0.09518336296081543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,2,8,power_law_1.2,0.10018176078796386
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,8,power_law_1.2,1.960120391845703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,2,8,power_law_1.01,0.6197452926635743
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,2,8,power_law_1.01,0.8058496093749999
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,2,8,power_law_1.2,0.10161919593811035
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,2,8,power_law_1.2,0.11791104316711425
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,2,8,power_law_1.2,0.15270912170410156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,2,8,power_law_1.2,0.13043968200683592
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,2,8,power_law_1.2,0.17919103622436522
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,4,balanced,0.03555711984634399
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,2,8,power_law_1.2,0.2701593589782715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,4,balanced,0.06174464225769043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,4,balanced,0.0756710386276245
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,4,balanced,0.09815296173095703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,4,balanced,0.14523648262023925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,4,balanced,0.30058496475219726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,4,balanced,0.2961459159851074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,4,balanced,0.2932787132263184
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,4,balanced,0.2808550453186035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,4,balanced,0.2844915199279785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,4,balanced,0.35589118957519533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,4,balanced,0.2932159996032715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,4,balanced,0.378922233581543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,4,balanced,0.30656639099121097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,4,balanced,0.3110655975341797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,4,balanced,0.3226073455810547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,4,balanced,0.3385203170776367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,4,balanced,0.3766131210327148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,4,balanced,0.4137561416625976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,4,balanced,0.49455230712890624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,4,balanced,0.5805939102172852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,4,balanced,0.7795788574218749
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,4,balanced,0.9175974273681641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,4,balanced,1.3108326721191408
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,4,balanced,1.6711500549316405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,4,balanced,2.4488627624511716
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,4,balanced,0.03586816072463989
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,4,balanced,0.03914880037307739
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,4,balanced,0.05411456108093262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,4,balanced,0.07945600032806396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,4,balanced,0.1101030445098877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,4,balanced,0.1763737678527832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,4,balanced,0.1733171272277832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,4,balanced,0.17447423934936523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,4,balanced,0.19297920227050783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,4,balanced,0.19457279205322267
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,4,balanced,0.1959859275817871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,4,balanced,0.198787841796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,4,balanced,0.20213375091552735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,4,balanced,3.5098034667968747
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,4,balanced,0.20708095550537112
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,4,balanced,0.20873727798461914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,4,balanced,0.2149196815490723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,4,balanced,0.2250137519836426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,4,balanced,0.24874752044677734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,4,balanced,0.26566656112670894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,4,balanced,0.3051724815368652
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,4,balanced,0.3474124908447266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,4,balanced,0.41555198669433596
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,4,balanced,0.5008025741577148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,4,balanced,0.7393049621582032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,4,balanced,0.8939456176757812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,4,balanced,1.2921279907226562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,4,balanced,1.6673062133789063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,2,8,power_law_1.2,0.3122713661193848
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,2,8,power_law_1.2,0.45158016204833984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,2,8,power_law_1.2,0.5889471817016602
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,2,8,power_law_1.2,0.9009113311767578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,2,8,power_law_1.2,1.1062016296386719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,4,4,balanced,0.02975615978240967
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,4,4,balanced,0.109619197845459
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,4,4,balanced,0.030536959171295164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,4,4,balanced,0.11059455871582032
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,4,4,balanced,0.11051520347595215
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,4,4,balanced,0.11061887741088867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,4,4,balanced,0.1109721565246582
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,4,4,balanced,0.11293696403503417
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,4,4,balanced,0.11292672157287598
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,4,4,balanced,0.11418880462646483
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,4,4,balanced,0.0350272011756897
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,4,4,balanced,0.11488127708435059
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,4,4,balanced,0.11904000282287597
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,4,4,balanced,0.12279040336608886
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,4,4,balanced,0.1283187198638916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,4,4,balanced,0.1329203224182129
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,4,4,balanced,0.14715392112731934
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,4,4,balanced,0.158274564743042
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,4,4,balanced,0.04684160232543945
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,4,4,balanced,0.06859392166137696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,4,power_law_1.01,0.07472767829895019
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,4,power_law_1.01,0.0938815975189209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,4,power_law_1.01,0.10753151893615723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,4,power_law_1.01,0.129367036819458
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,4,power_law_1.01,0.1565503978729248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,4,power_law_1.01,0.17853952407836915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,4,power_law_1.01,0.21956096649169923
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,4,power_law_1.01,0.20942848205566406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,4,power_law_1.01,0.24045568466186523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,4,power_law_1.01,0.2382156753540039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,4,power_law_1.01,0.2500966453552246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,4,power_law_1.01,0.25494144439697264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,4,power_law_1.01,0.26559104919433596
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,4,power_law_1.01,0.280313606262207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,4,power_law_1.01,0.2825331115722656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,4,power_law_1.01,0.2977203178405762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,4,power_law_1.01,0.34814079284667965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,4,power_law_1.01,0.4357555389404297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,4,power_law_1.01,0.4671449661254883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,4,power_law_1.01,0.5516812896728516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,4,power_law_1.01,0.6492889404296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,4,4,balanced,0.10948608398437501
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,4,4,balanced,0.108919038772583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,4,power_law_1.01,0.8437197113037109
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,4,4,balanced,0.10946944236755371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,4,power_law_1.01,1.0740646362304687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,4,power_law_1.01,1.4530854797363282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,4,power_law_1.01,0.049834241867065435
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,4,power_law_1.01,0.06655744075775147
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,4,power_law_1.01,1.813501434326172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,4,power_law_1.01,0.07577216148376464
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,4,power_law_1.01,0.09758720397949219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,4,power_law_1.01,0.11293567657470702
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,4,power_law_1.01,0.12319231986999513
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,4,power_law_1.01,2.710346374511719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,4,power_law_1.01,0.15391231536865235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,4,power_law_1.01,0.1694361686706543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,4,power_law_1.01,0.16997888565063476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,4,power_law_1.01,0.1745919990539551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,4,power_law_1.01,0.17891712188720704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,4,power_law_1.01,0.18446720123291016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,4,power_law_1.01,3.5370904541015626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,4,power_law_1.01,0.18959615707397462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,4,power_law_1.01,0.1523647975921631
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,4,power_law_1.01,0.19527296066284178
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,4,power_law_1.01,0.200948486328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,4,power_law_1.01,0.21051776885986326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,4,power_law_1.01,0.24155263900756835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,4,power_law_1.01,0.2599807929992676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,4,power_law_1.01,0.32314880371093746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,4,power_law_1.01,0.368642578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,4,power_law_1.01,0.46999679565429686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,4,power_law_1.01,0.5698751831054688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,4,power_law_1.01,0.8041165161132813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,4,power_law_1.01,1.0103782653808593
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,4,4,power_law_1.01,0.03504512071609497
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,4,4,power_law_1.01,0.049489917755126955
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,4,4,power_law_1.01,0.06091904163360595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,4,power_law_1.01,1.4481190490722657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,4,4,power_law_1.01,0.0669491195678711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,4,4,power_law_1.01,0.08093440055847168
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,4,4,power_law_1.01,0.07594111919403077
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,4,4,power_law_1.01,0.10098303794860838
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,4,4,power_law_1.01,0.10437503814697266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,4,4,power_law_1.01,0.1055065631866455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,4,power_law_1.01,1.8227328491210937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,4,4,power_law_1.01,0.10318976402282716
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,4,4,power_law_1.01,0.10151040077209472
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,4,4,power_law_1.01,0.10608896255493164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,4,4,power_law_1.01,0.10288640022277833
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,4,4,power_law_1.01,0.10250752449035645
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,4,4,power_law_1.01,0.107007999420166
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,4,4,power_law_1.01,0.11084799766540528
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,4,4,power_law_1.01,0.12006143569946288
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,4,4,balanced,0.2252390480041504
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,4,4,balanced,0.25260799407958984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,4,4,balanced,0.3581158447265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,4,4,balanced,0.4632716751098632
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,4,4,power_law_1.01,0.13412351608276368
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,4,4,power_law_1.01,0.14799360275268553
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,4,4,power_law_1.01,0.21761663436889647
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,4,4,power_law_1.01,0.1806604766845703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,4,power_law_1.2,0.07760767936706543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,4,power_law_1.2,0.09161215782165527
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,4,power_law_1.2,0.09714303970336914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,4,power_law_1.2,0.12359680175781249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,4,power_law_1.2,0.151528959274292
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,4,power_law_1.2,0.1667238426208496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,4,power_law_1.2,0.22813568115234376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,4,power_law_1.2,0.2163711929321289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,4,power_law_1.2,0.2320806312561035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,4,power_law_1.2,0.23956991195678712
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,4,power_law_1.2,0.2490662384033203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,4,power_law_1.2,0.2577267265319824
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,4,4,power_law_1.01,0.2719193649291992
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,4,4,power_law_1.01,0.33789310455322263
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,4,power_law_1.2,0.2783347129821777
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,4,4,power_law_1.01,0.450645751953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,4,power_law_1.2,0.2970508766174317
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,4,power_law_1.2,0.3056563186645508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,4,power_law_1.2,0.39277694702148436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,4,power_law_1.2,0.4344563293457031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,4,power_law_1.2,0.47098239898681643
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,4,power_law_1.2,0.5791270446777343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,4,power_law_1.2,0.705296630859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,4,power_law_1.2,0.8724416351318359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,4,power_law_1.2,1.074645767211914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,4,power_law_1.2,0.2655539131164551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,4,power_law_1.2,1.4492568969726562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,4,power_law_1.2,0.04791039943695068
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,4,power_law_1.2,0.06958975791931152
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,4,power_law_1.2,0.0661900806427002
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,4,4,power_law_1.01,0.6290201568603516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,4,power_law_1.2,1.9624064636230467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,4,power_law_1.2,0.09549695968627929
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,4,power_law_1.2,0.11363840103149414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,4,power_law_1.2,0.11348863601684571
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,4,power_law_1.2,0.15299072265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,4,power_law_1.2,0.16923904418945312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,4,power_law_1.2,0.1738470458984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,4,power_law_1.2,0.1848575973510742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,4,power_law_1.2,0.18220159530639649
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,4,power_law_1.2,0.1916659164428711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,4,power_law_1.2,2.9018072509765624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,4,power_law_1.2,0.1536025619506836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,4,power_law_1.2,0.19763584136962892
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,4,power_law_1.2,0.2036300849914551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,4,power_law_1.2,0.21633663177490234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,4,power_law_1.2,0.24670335769653323
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,4,power_law_1.2,0.2761459159851074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,4,power_law_1.2,0.3379097747802734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,4,power_law_1.2,3.6389144897460937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,4,power_law_1.2,0.3909708786010742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,4,power_law_1.2,0.4882815933227539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,4,power_law_1.2,0.6072972869873047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,4,4,power_law_1.2,0.03410559892654419
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,4,power_law_1.2,0.8335116577148437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,4,power_law_1.2,1.0158745574951173
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,4,4,power_law_1.2,0.04815616130828858
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,4,4,power_law_1.2,0.04960000038146973
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,4,4,power_law_1.2,0.0625548791885376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,4,4,power_law_1.2,0.0752780818939209
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,4,4,power_law_1.2,0.0894553565979004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,4,power_law_1.2,1.5054605102539063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,4,4,power_law_1.2,0.10215423583984376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,4,4,power_law_1.2,0.10070655822753907
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,4,4,power_law_1.2,0.09224703788757324
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,4,4,power_law_1.2,0.10731007575988769
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,4,4,power_law_1.2,0.10622336387634276
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,4,4,power_law_1.2,0.10686079978942871
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,4,4,power_law_1.2,0.10479104042053222
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,4,4,power_law_1.2,0.10603775978088378
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,4,4,power_law_1.2,0.11245823860168458
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,4,4,power_law_1.2,0.11765376091003418
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,4,4,power_law_1.2,0.14775424003601073
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,4,power_law_1.2,1.9347378540039064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,4,power_law_1.2,0.16947328567504882
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,4,4,power_law_1.01,0.9315148925781249
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,4,4,power_law_1.01,1.133623046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,4,4,power_law_1.2,0.12836095809936524
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,4,4,power_law_1.2,0.1801740837097168
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,4,4,power_law_1.2,0.20174463272094725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,8,2,balanced,0.03579391956329346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,8,2,balanced,0.06187903881072998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,8,2,balanced,0.07576064109802247
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,4,4,power_law_1.2,0.2818201637268066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,8,2,balanced,0.09898367881774903
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,8,2,balanced,0.14959103584289551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,8,2,balanced,0.31363712310791014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,8,2,balanced,0.3096038436889649
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,8,2,balanced,0.3120486450195313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,4,4,power_law_1.2,0.34918785095214844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,8,2,balanced,0.3093964767456055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,8,2,balanced,0.3124620819091797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,8,2,balanced,0.3149043273925781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,8,2,balanced,0.31786495208740234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,8,2,balanced,0.31751935958862304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,8,2,balanced,0.32674049377441405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,8,2,balanced,0.3340697479248047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,8,2,balanced,0.34066558837890626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,8,2,balanced,0.3572966384887695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,8,2,balanced,0.3907519912719727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,8,2,balanced,0.41914623260498046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,8,2,balanced,0.5015219116210938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,8,2,balanced,0.5837376022338867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,4,4,power_law_1.2,0.5141120147705078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,8,2,balanced,0.8042291259765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,8,2,balanced,0.9033203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,8,2,balanced,1.2980915832519533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,8,2,balanced,0.03585024118423462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,8,2,balanced,1.6563623046875002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,8,2,balanced,0.040783357620239255
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,8,2,balanced,0.05789055824279785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,8,2,balanced,0.08439167976379394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,8,2,balanced,0.13913087844848632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,8,2,balanced,0.21645952224731446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,8,2,balanced,0.2178175926208496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,8,2,balanced,0.21959423065185546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,8,2,balanced,2.397127685546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,8,2,balanced,0.2211622428894043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,8,2,balanced,0.2228646469116211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,8,2,balanced,0.22590335845947265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,8,2,balanced,0.2560268783569336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,8,2,balanced,0.2582912063598633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,8,2,balanced,0.2632473564147949
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,8,2,balanced,0.2650918388366699
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,8,2,balanced,0.2704447937011719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,8,2,balanced,0.27733631134033204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,8,2,balanced,3.3630297851562503
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,8,2,balanced,0.2930240058898926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,8,2,balanced,0.31138559341430666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,8,2,balanced,0.3474316787719727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,8,2,balanced,0.38377727508544923
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,8,2,balanced,0.45532928466796874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,8,2,balanced,0.5277337646484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,8,2,balanced,0.8052159881591796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,8,2,balanced,0.9472460937499999
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,8,2,balanced,1.3796083068847655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,8,2,balanced,1.7791961669921874
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,4,4,power_law_1.2,0.6092633438110352
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,4,4,power_law_1.2,1.0097689819335938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,4,4,power_law_1.2,1.5160473632812501
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,4,4,power_law_1.2,1.6408460998535155
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,8,2,balanced,0.042018561363220214
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,8,2,balanced,0.04321280002593994
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,8,2,balanced,0.14313983917236328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,8,2,balanced,0.14393471717834472
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,8,2,balanced,0.14444928169250487
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,8,2,balanced,0.050449919700622556
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,8,2,balanced,0.14583552360534668
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,8,2,balanced,0.14600831985473633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,8,2,balanced,0.14763903617858887
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,8,2,balanced,0.1498572826385498
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,8,2,balanced,0.15416576385498046
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,8,2,balanced,0.16635776519775392
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,8,2,balanced,0.17649152755737305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,8,2,balanced,0.18904064178466798
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,8,2,balanced,0.06489344120025635
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,8,2,balanced,0.21442304611206056
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,8,2,balanced,0.2385945510864258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,8,2,balanced,0.0919206428527832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,8,2,balanced,0.1432192039489746
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,8,2,balanced,0.1429811191558838
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,8,2,power_law_1.01,0.06507264137268066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,8,2,power_law_1.01,0.0748582410812378
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,8,2,power_law_1.01,0.08824064254760741
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,8,2,balanced,0.14253952026367187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,8,2,power_law_1.01,0.10743552207946779
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,8,2,power_law_1.01,0.14139264106750488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,8,2,power_law_1.01,0.16958463668823243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,8,2,power_law_1.01,0.2110028839111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,8,2,power_law_1.01,0.21597312927246093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,8,2,power_law_1.01,0.23287424087524414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,8,2,power_law_1.01,0.22802816390991212
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,8,2,power_law_1.01,0.2443929672241211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,8,2,power_law_1.01,0.28496000289916995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,8,2,power_law_1.01,0.2925196838378906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,8,2,power_law_1.01,0.3051340866088867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,8,2,power_law_1.01,0.30883968353271485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,8,2,power_law_1.01,0.3280204772949219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,8,2,power_law_1.01,0.3502425765991211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,8,2,power_law_1.01,0.42843776702880854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,8,2,power_law_1.01,0.4886656188964844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,8,2,power_law_1.01,0.5542963027954102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,8,2,power_law_1.01,0.6606745910644531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,8,2,power_law_1.01,0.8498841857910155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,8,2,power_law_1.01,1.036258544921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,8,2,power_law_1.01,1.3915122985839843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,8,2,power_law_1.01,1.7670080566406248
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,8,2,balanced,0.35916416168212895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,8,2,power_law_1.01,2.5345753479003905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,8,2,power_law_1.01,0.040318717956542974
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,8,2,power_law_1.01,0.05327744007110595
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,8,2,balanced,0.4102873611450195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,8,2,power_law_1.01,3.3069708251953123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,8,2,power_law_1.01,0.12642815589904785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,8,2,power_law_1.01,0.16704511642456055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,8,2,power_law_1.01,0.09195903778076173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,8,2,power_law_1.01,0.06770559787750244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,8,2,power_law_1.01,0.18754816055297852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,8,2,power_law_1.01,0.1955571174621582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,8,2,power_law_1.01,0.19805055618286133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,8,2,power_law_1.01,0.2042265510559082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,8,2,power_law_1.01,0.22984319686889648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,8,2,power_law_1.01,0.24049919128417968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,8,2,power_law_1.01,0.24518911361694337
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,8,2,power_law_1.01,0.23358720779418946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,8,2,power_law_1.01,0.24858495712280276
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,8,2,power_law_1.01,0.26446847915649413
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,8,2,balanced,0.5904000091552735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,8,2,power_law_1.01,0.31528064727783206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,8,2,power_law_1.01,0.28336639404296876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,8,2,power_law_1.01,0.37300479888916016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,8,2,power_law_1.01,0.19232383728027344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,8,2,power_law_1.01,0.4209318542480469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,8,2,power_law_1.01,0.5225433731079102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,8,2,power_law_1.01,0.6388300704956055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,8,2,power_law_1.01,0.8467584228515624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,8,2,power_law_1.01,0.04446080207824707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,8,2,power_law_1.01,1.066915817260742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,8,2,balanced,0.7663910675048828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,8,2,power_law_1.01,1.4936448669433593
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,8,2,power_law_1.01,0.08425215721130372
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,8,2,power_law_1.01,0.061662721633911136
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,8,2,power_law_1.01,0.05118080139160156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,8,2,power_law_1.01,0.07228288173675537
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,8,2,power_law_1.01,0.09454976081848145
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,8,2,power_law_1.01,0.11852543830871583
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,8,2,power_law_1.01,0.12467840194702148
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,8,2,power_law_1.01,0.12807040214538573
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,8,2,power_law_1.01,0.13372672080993653
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,8,2,power_law_1.01,0.14026880264282227
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,8,2,power_law_1.01,0.14233856201171874
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,8,2,power_law_1.01,0.1478540802001953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,8,2,power_law_1.01,0.15402496337890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,8,2,power_law_1.01,0.15545856475830078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,8,2,power_law_1.01,0.16541055679321287
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,8,2,power_law_1.01,1.930292510986328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,8,2,power_law_1.01,0.1307532787322998
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,8,2,power_law_1.01,0.26091520309448246
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,8,2,power_law_1.01,0.2142118453979492
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,8,2,power_law_1.01,0.18958335876464844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,8,2,power_law_1.01,0.3209843063354492
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,8,2,power_law_1.01,0.42704513549804685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,8,2,power_law_1.2,0.060034561157226565
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,8,2,power_law_1.01,0.5264012908935547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,8,2,power_law_1.2,0.0742412805557251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,8,2,power_law_1.2,0.08718463897705078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,8,2,power_law_1.2,0.1026841640472412
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,8,2,power_law_1.2,0.1376255989074707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,8,2,power_law_1.2,0.1598425579071045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,8,2,power_law_1.2,0.20371456146240235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,8,2,power_law_1.2,0.21458303451538088
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,8,2,power_law_1.2,0.224322566986084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,8,2,power_law_1.2,0.22776832580566406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,8,2,power_law_1.2,0.25021055221557614
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,8,2,power_law_1.2,0.2858367919921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,8,2,power_law_1.2,0.2938444709777832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,8,2,power_law_1.2,0.30785408020019533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,8,2,power_law_1.2,0.31811328887939455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,8,2,power_law_1.2,0.33469696044921876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,8,2,power_law_1.2,0.37458816528320316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,8,2,power_law_1.2,0.46434047698974606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,8,2,power_law_1.2,0.492413444519043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,8,2,power_law_1.2,0.5626892852783203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,8,2,power_law_1.2,0.6689036560058594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,8,2,power_law_1.2,0.8565824127197266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,8,2,power_law_1.2,1.0455052947998047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,8,2,power_law_1.2,1.421913604736328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,8,2,power_law_1.2,1.796425018310547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,8,2,power_law_1.2,2.5387532043457033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,8,2,power_law_1.2,0.04115456104278564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,8,2,power_law_1.2,0.05409535884857177
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,8,2,power_law_1.2,0.0653004789352417
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,8,2,power_law_1.2,0.08913023948669434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,8,2,power_law_1.2,0.12143872261047364
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,8,2,power_law_1.2,3.3948159790039063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,8,2,power_law_1.2,0.14607232093811034
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,8,2,power_law_1.2,0.1862246322631836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,8,2,power_law_1.2,0.1917990493774414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,8,2,power_law_1.2,0.19655424118041992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,8,2,power_law_1.2,0.19762432098388671
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,8,2,power_law_1.2,0.2039718437194824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,8,2,power_law_1.2,0.23358335494995117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,8,2,power_law_1.2,0.2348454475402832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,8,2,power_law_1.2,0.24261247634887698
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,8,2,power_law_1.2,0.24724607467651366
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,8,2,power_law_1.2,0.2516979217529297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,8,2,power_law_1.2,0.2730521583557129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,8,2,power_law_1.2,0.29384063720703124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,8,2,power_law_1.2,0.3311449432373047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,8,2,power_law_1.2,0.38423297882080076
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,8,2,power_law_1.2,0.4408051300048828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,8,2,power_law_1.2,0.5470515060424804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,8,2,power_law_1.2,0.6641792297363281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,8,2,power_law_1.2,0.8749555206298828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,8,2,power_law_1.2,1.1156735992431641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,8,2,power_law_1.2,1.5368730163574218
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,8,2,power_law_1.2,0.045646080970764155
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,8,2,power_law_1.2,0.051490559577941894
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,8,2,power_law_1.2,0.05860864162445069
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,8,2,power_law_1.2,1.9988890075683592
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,8,2,power_law_1.2,0.07085440158843995
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,8,2,power_law_1.2,0.07724415779113769
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,8,2,power_law_1.2,0.09571455955505372
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,8,2,power_law_1.2,0.11719807624816894
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,8,2,power_law_1.2,0.1223731231689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,8,2,power_law_1.2,0.12763263702392577
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,8,2,power_law_1.2,0.1295257568359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,8,2,power_law_1.2,0.13648256301879882
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,8,2,power_law_1.2,0.14198143959045412
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,8,2,power_law_1.2,0.14633472442626955
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,8,2,power_law_1.2,0.15149439811706542
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,8,2,power_law_1.2,0.15842432022094727
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,8,2,power_law_1.2,0.1654515266418457
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,8,2,power_law_1.2,0.1860723114013672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,8,2,power_law_1.01,0.8099571228027344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,8,2,power_law_1.01,1.0333837127685546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,8,2,power_law_1.01,1.631846466064453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,8,2,power_law_1.01,1.982488250732422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,8,2,power_law_1.2,0.21220224380493163
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,8,2,power_law_1.2,0.2520051193237305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,8,2,power_law_1.2,0.32156288146972656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,16,1,balanced,0.04474944114685059
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,16,1,balanced,0.07181119918823242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,16,1,balanced,0.08911552429199218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,16,1,balanced,0.11622015953063965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,16,1,balanced,0.17263551712036135
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,16,1,balanced,0.3575302505493164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,16,1,balanced,0.3458163070678711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,8,2,power_law_1.2,0.4181171035766601
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,16,1,balanced,0.3485273742675781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,8,2,power_law_1.2,0.5885158538818359
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,8,2,power_law_1.2,0.7816716766357421
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,16,1,balanced,0.3503129577636719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,16,1,balanced,0.34765953063964844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,16,1,balanced,0.3522675323486328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,16,1,balanced,0.35462207794189454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,16,1,balanced,0.35636737823486325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,16,1,balanced,0.3591052627563477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,16,1,balanced,0.42447551727294924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,16,1,balanced,0.4343564987182617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,16,1,balanced,0.4530035018920898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,16,1,balanced,0.47556095123291015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,16,1,balanced,0.5046105575561524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,16,1,balanced,0.5654528045654297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,16,1,balanced,0.6350886535644531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,16,1,balanced,0.8878189086914062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,16,1,balanced,0.9879180908203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,16,1,balanced,1.3912736511230468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,16,1,balanced,0.046381440162658695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,16,1,balanced,0.05494783878326416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,16,1,balanced,1.7660543823242185
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,16,1,balanced,0.07456704139709472
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,16,1,balanced,0.10913087844848632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,16,1,balanced,0.17662656784057618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,16,1,balanced,0.31806144714355467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,16,1,balanced,0.31880447387695315
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,16,1,balanced,2.5279571533203127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,16,1,balanced,0.3219564819335937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,16,1,balanced,0.32354942321777347
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,16,1,balanced,0.3249203109741211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,16,1,balanced,0.33084415435791015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,16,1,balanced,0.33446975708007814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,16,1,balanced,0.3394393539428711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,16,1,balanced,3.4394439697265624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,16,1,balanced,0.34542911529541015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,16,1,balanced,0.37844799041748045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,16,1,balanced,0.383583984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,16,1,balanced,0.38997249603271483
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,16,1,balanced,0.4079699325561523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,16,1,balanced,0.4225881576538086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,16,1,balanced,0.45479999542236327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,16,1,balanced,0.4874348831176758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,16,1,balanced,0.5593017578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,16,1,balanced,0.6239340972900391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,16,1,balanced,0.9723193359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,16,1,balanced,1.1102508544921874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,16,1,balanced,1.5955250549316407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,16,1,balanced,2.0834501647949217
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,8,2,power_law_1.2,1.3306060791015626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,8,2,power_law_1.2,2.587517395019531
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,8,2,power_law_1.2,3.7483288574218747
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,8,2,power_law_1.2,1.5957708740234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,16,1,balanced,0.07058303833007813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,16,1,balanced,0.06834943771362305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,16,1,balanced,0.21869695663452146
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,16,1,balanced,0.08200127601623534
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,16,1,balanced,0.21963903427124026
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,16,1,balanced,0.09953215599060058
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,16,1,balanced,0.2214451217651367
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,16,1,balanced,0.22262527465820314
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,16,1,balanced,0.224847354888916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,16,1,balanced,0.227455997467041
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,16,1,balanced,0.2301158332824707
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,16,1,balanced,0.23968320846557617
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,16,1,balanced,0.25038719177246094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,16,1,balanced,0.26159103393554684
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,16,1,balanced,0.23377599716186523
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,16,1,balanced,0.2859398460388184
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,16,1,balanced,0.31025728225708005
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,16,1,balanced,0.35997825622558594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,16,1,balanced,0.41112831115722653
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,16,1,balanced,0.13762240409851073
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,16,1,balanced,0.21601728439331053
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,16,1,power_law_1.01,0.06994815826416015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,16,1,power_law_1.01,0.04490816116333008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,16,1,power_law_1.01,0.0894041633605957
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,16,1,balanced,0.21820159912109377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,16,1,power_law_1.01,0.1492518424987793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,16,1,balanced,0.21550464630126953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,16,1,power_law_1.01,0.18840192794799804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,16,1,power_law_1.01,0.2432159996032715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,16,1,power_law_1.01,0.11258048057556151
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,16,1,power_law_1.01,0.25738752365112305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,16,1,power_law_1.01,0.25924224853515626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,16,1,power_law_1.01,0.27450111389160153
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,16,1,power_law_1.01,0.31706880569458007
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,16,1,power_law_1.01,0.32554561614990235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,16,1,power_law_1.01,0.3455379104614258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,16,1,power_law_1.01,0.4038246536254883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,16,1,power_law_1.01,0.4112192153930664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,16,1,power_law_1.01,0.43498241424560546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,16,1,power_law_1.01,0.49038974761962895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,16,1,power_law_1.01,0.5827833557128906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,16,1,power_law_1.01,0.6466700744628906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,16,1,power_law_1.01,0.7320735931396485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,16,1,power_law_1.01,0.24784255981445313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,16,1,power_law_1.01,0.9180825805664062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,16,1,power_law_1.01,1.0887904357910156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,16,1,power_law_1.01,1.4609861755371094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,16,1,power_law_1.01,0.047216639518737794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,16,1,power_law_1.01,1.8371058654785155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,16,1,power_law_1.01,0.05456960201263428
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,16,1,power_law_1.01,0.10888319969177247
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,16,1,power_law_1.01,2.5743865966796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,16,1,power_law_1.01,0.15166720390319824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,16,1,power_law_1.01,0.21949375152587888
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,16,1,power_law_1.01,0.2703609657287598
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,16,1,power_law_1.01,0.27830272674560547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,16,1,power_law_1.01,3.3230612182617185
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,16,1,power_law_1.01,0.07571263790130615
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,16,1,power_law_1.01,0.2827673530578613
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,16,1,power_law_1.01,0.28534912109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,16,1,power_law_1.01,0.2954412841796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,16,1,power_law_1.01,0.3036588859558106
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,16,1,power_law_1.01,0.3085830307006836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,16,1,power_law_1.01,0.31547903060913085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,16,1,power_law_1.01,0.34880001068115235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,16,1,power_law_1.01,0.35507328033447266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,16,1,power_law_1.01,0.36893951416015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,16,1,power_law_1.01,0.3912768173217773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,16,1,power_law_1.01,0.42118209838867193
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,16,1,power_law_1.01,0.47650177001953126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,16,1,power_law_1.01,0.5450009536743164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,16,1,power_law_1.01,0.6539794921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,16,1,power_law_1.01,0.7621011352539062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,16,1,power_law_1.01,1.0008422088623046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,16,1,power_law_1.01,1.2482540893554688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,16,1,power_law_1.01,1.7315930175781251
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,16,1,power_law_1.01,0.06894015789031982
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,16,1,power_law_1.01,0.07217408180236816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,16,1,power_law_1.01,2.2142591857910157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,16,1,power_law_1.01,0.08571071624755859
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,16,1,power_law_1.01,0.09921600341796874
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,16,1,power_law_1.01,0.1215494441986084
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,16,1,balanced,0.637496337890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,16,1,power_law_1.01,0.14082112312316894
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,16,1,balanced,0.7392134094238282
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,16,1,power_law_1.01,0.18262592315673828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,16,1,balanced,1.0691577911376953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,16,1,power_law_1.01,0.1758982467651367
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,16,1,power_law_1.01,0.206625919342041
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,16,1,power_law_1.01,0.18639039993286133
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,16,1,power_law_1.01,0.19811199188232423
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,16,1,power_law_1.01,0.21009792327880858
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,16,1,power_law_1.01,0.21787391662597658
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,16,1,balanced,1.390777587890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,16,1,power_law_1.01,0.23882816314697264
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,16,1,power_law_1.01,0.19007999420166016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,16,1,power_law_1.01,0.23108991622924804
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,16,1,power_law_1.01,0.2586777687072754
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,16,1,power_law_1.01,0.3432825469970703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,16,1,power_law_1.01,0.29806976318359374
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,16,1,power_law_1.01,0.4274393463134766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,16,1,power_law_1.01,0.9253695678710937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,16,1,power_law_1.01,0.7087225341796876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,16,1,power_law_1.2,0.04455296039581299
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,16,1,power_law_1.01,0.5287340927124023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,16,1,power_law_1.2,0.08925632476806641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,16,1,power_law_1.2,0.10957440376281738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,16,1,power_law_1.2,0.13741951942443847
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,16,1,power_law_1.2,0.191975040435791
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,16,1,power_law_1.2,0.2380665588378906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,16,1,power_law_1.01,1.3048664855957033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,16,1,power_law_1.2,0.24970624923706053
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,16,1,power_law_1.2,0.06634880065917968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,16,1,power_law_1.2,0.25313215255737304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,16,1,power_law_1.2,0.26200960159301756
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,16,1,power_law_1.2,0.3153849601745605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,16,1,power_law_1.2,0.2794400024414062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,16,1,power_law_1.01,1.7466400146484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,16,1,power_law_1.2,0.3272889709472656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,16,1,power_law_1.2,0.3560518264770508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,16,1,power_law_1.2,0.40620159149169927
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,16,1,power_law_1.2,0.4201990509033203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,16,1,power_law_1.2,0.44421825408935545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,16,1,power_law_1.2,0.5863308715820312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,16,1,power_law_1.2,0.5312518310546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,16,1,power_law_1.2,0.6543103790283202
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,16,1,power_law_1.2,0.7517510223388671
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,16,1,power_law_1.2,0.9341772460937501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,16,1,power_law_1.2,1.1073843383789064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,16,1,power_law_1.2,0.051674242019653316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,16,1,power_law_1.2,0.054776959419250495
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,16,1,power_law_1.2,0.07756991863250733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,16,1,power_law_1.2,1.4726483154296877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,16,1,power_law_1.2,0.10356608390808106
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,16,1,power_law_1.2,0.14539392471313478
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,16,1,power_law_1.2,1.8445542907714845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,16,1,power_law_1.2,0.20797887802124024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,16,1,power_law_1.2,0.26790143966674806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,16,1,power_law_1.2,0.2760479927062988
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,16,1,power_law_1.2,0.2801055908203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,16,1,power_law_1.2,0.28647680282592775
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,16,1,power_law_1.2,0.29377920150756837
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,16,1,power_law_1.2,2.59376708984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,16,1,power_law_1.2,0.30203775405883787
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,16,1,power_law_1.2,0.3090950393676758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,16,1,power_law_1.2,0.31962432861328127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,16,1,power_law_1.2,0.3525580978393555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,16,1,power_law_1.2,0.3617817687988281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,16,1,power_law_1.2,0.3747078323364258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,16,1,power_law_1.2,0.4061888122558594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,16,1,power_law_1.2,3.347238464355469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,16,1,power_law_1.2,0.43287422180175783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,16,1,power_law_1.2,0.5063084793090821
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,16,1,power_law_1.2,0.5601689529418945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,16,1,power_law_1.2,0.6738912200927734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,16,1,power_law_1.2,0.8050303649902343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,16,1,power_law_1.2,1.0131302642822266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,16,1,power_law_1.2,1.2660269165039062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,16,1,power_law_1.2,0.06818943977355957
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,16,1,power_law_1.2,0.07226943969726562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,16,1,power_law_1.2,1.7490681457519532
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,16,1,power_law_1.2,0.08391551971435547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,16,1,power_law_1.2,0.09671296119689941
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,16,1,power_law_1.2,2.232174072265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,16,1,power_law_1.2,0.11657279968261718
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,16,1,power_law_1.2,0.13627584457397462
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,16,1,power_law_1.2,0.17376447677612306
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,16,1,power_law_1.2,0.18097471237182616
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,16,1,power_law_1.2,0.18592960357666016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,16,1,power_law_1.2,0.1920057678222656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,16,1,power_law_1.2,0.19907455444335936
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,16,1,power_law_1.2,0.2072332763671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,16,1,power_law_1.2,0.22842687606811524
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,16,1,power_law_1.2,0.21470975875854492
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,16,1,power_law_1.2,0.24167999267578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,16,1,power_law_1.2,0.2579609680175781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,16,1,power_law_1.2,0.4019417572021484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,16,1,power_law_1.01,3.4144799804687502
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,16,1,power_law_1.01,2.5746380615234377
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,16,1,power_law_1.2,0.28702783584594727
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,16,1,power_law_1.2,0.34448577880859377
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,16,1,power_law_1.2,0.5474214553833008
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,16,1,power_law_1.2,0.682537612915039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,16,balanced,0.04961535930633545
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,16,1,power_law_1.2,0.9959366607666016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,16,balanced,0.04855936050415039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,16,balanced,0.1184883213043213
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,16,1,power_law_1.2,1.3441433715820312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,16,balanced,0.04443647861480713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,16,balanced,0.21492992401123048
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,16,balanced,0.22459135055541993
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,16,balanced,0.22504192352294922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,16,balanced,0.2248896026611328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,16,balanced,0.23011327743530274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,16,balanced,0.21993215560913085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,16,balanced,0.06993152141571045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,16,balanced,0.23352319717407227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,16,balanced,0.23064319610595704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,16,balanced,0.23376256942749024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,16,balanced,0.24282880783081057
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,16,balanced,0.3142220878601074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,16,balanced,0.3011750411987305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,16,balanced,0.2213030433654785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,16,balanced,0.22777215957641603
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,16,balanced,0.3664025497436524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,16,balanced,0.40824321746826175
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,16,balanced,0.5154444885253906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,16,balanced,0.5774182510375977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,16,balanced,0.9144499206542969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,16,balanced,0.035973119735717776
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,16,balanced,1.0441522979736328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,16,balanced,0.035952639579772946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,16,balanced,0.038396160602569576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,16,balanced,0.05624447822570801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,16,balanced,0.08130047798156739
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,16,balanced,0.13281408309936524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,16,balanced,0.1468172836303711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,16,balanced,0.14321279525756836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,16,balanced,0.1465344047546387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,16,balanced,0.1470207977294922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,16,balanced,0.15050496101379393
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,16,balanced,0.1532710361480713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,16,balanced,0.15648639678955079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,16,balanced,0.14866687774658202
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,16,balanced,0.15238271713256837
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,16,balanced,0.15599488258361816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,16,balanced,0.15981311798095704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,16,balanced,0.16729856491088868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,16,balanced,0.1864729690551758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,16,balanced,0.21252351760864258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,16,balanced,0.23057024002075197
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,16,balanced,0.2782271957397461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,16,balanced,0.32381439208984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,16,balanced,0.27330047607421876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,16,balanced,0.4636249542236328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,16,balanced,0.5569331359863281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,16,balanced,0.780346908569336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,16,balanced,1.0085977935791015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,16,balanced,0.6388019180297851
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,16,1,power_law_1.2,2.064801940917969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,16,1,power_law_1.2,2.7712973022460936
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,1,16,balanced,0.02225152015686035
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,1,16,balanced,0.02290560007095337
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,1,16,balanced,0.02222464084625244
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,1,16,balanced,0.08138751983642578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,1,16,balanced,0.08156031608581543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,1,16,balanced,0.08163071632385253
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,1,16,balanced,0.08112895965576172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,1,16,balanced,0.08144767761230469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,1,16,balanced,0.08180864334106445
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,1,16,balanced,0.08180480003356934
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,1,16,balanced,0.08219648361206054
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,1,16,balanced,0.0820569610595703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,1,16,balanced,0.08225791931152344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,1,16,balanced,0.08253824234008789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,1,16,balanced,0.08245504379272461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,1,16,balanced,0.08450431823730468
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,1,16,balanced,0.08497664451599121
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,1,16,balanced,0.08607872009277344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,1,16,balanced,0.09115008354187012
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,1,16,balanced,0.09463935852050781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,1,16,balanced,0.031596798896789555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,1,16,balanced,0.09940863609313964
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,1,16,balanced,0.06603263854980469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,16,power_law_1.01,0.21507328033447265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,16,power_law_1.01,0.11833600044250488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,16,power_law_1.01,0.19432064056396486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,16,power_law_1.01,0.20906496047973633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,16,power_law_1.01,0.21980031967163086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,16,power_law_1.01,0.1916595268249512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,16,power_law_1.01,0.1957414436340332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,16,power_law_1.01,0.2011609649658203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,16,power_law_1.01,0.19778688430786134
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,16,power_law_1.01,0.2093964767456055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,16,power_law_1.01,0.22029056549072265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,16,power_law_1.01,0.19667583465576172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,16,power_law_1.01,0.225347843170166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,16,power_law_1.01,0.23976959228515624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,16,power_law_1.01,0.2530265617370605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,16,power_law_1.01,0.2685004806518555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,16,power_law_1.01,0.2893440055847168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,16,power_law_1.01,0.2767475128173828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,1,16,balanced,0.1258355236053467
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,1,16,balanced,0.1819955253601074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,16,power_law_1.01,0.5392460632324219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,1,16,balanced,0.1348748779296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,16,power_law_1.01,0.5807385635375977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,16,power_law_1.01,0.6427865600585937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,16,power_law_1.01,0.7643392181396484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,16,power_law_1.01,0.9478643035888672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,16,power_law_1.01,0.07967743873596192
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,16,power_law_1.01,1.297400360107422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,16,power_law_1.01,1.6282815551757814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,16,power_law_1.01,0.1293337631225586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,16,power_law_1.01,0.12935680389404297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,16,power_law_1.01,0.30049280166625975
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,1,16,balanced,0.2271334457397461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,16,power_law_1.01,0.12624640464782716
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,16,power_law_1.01,0.13209343910217286
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,16,power_law_1.01,0.13915136337280273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,16,power_law_1.01,0.125546236038208
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,16,power_law_1.01,0.13354496002197264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,16,power_law_1.01,0.371728630065918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,16,power_law_1.01,0.13230463981628418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,16,power_law_1.01,0.12949376106262206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,16,power_law_1.01,0.1335167980194092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,16,power_law_1.01,0.13739263534545898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,16,power_law_1.01,0.147708158493042
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,16,power_law_1.01,0.1414233589172363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,16,power_law_1.01,0.15026304244995117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,16,power_law_1.01,0.15141504287719726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,16,power_law_1.01,0.19995391845703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,16,power_law_1.01,0.18601215362548829
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,16,power_law_1.01,0.24835071563720704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,16,power_law_1.01,0.2764838409423828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,16,power_law_1.01,0.3652761459350586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,16,power_law_1.01,0.14474623680114745
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,16,power_law_1.01,0.45845504760742184
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,16,power_law_1.01,0.602311668395996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,1,16,power_law_1.01,0.04968575954437256
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,1,16,power_law_1.01,0.0812070369720459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,16,power_law_1.01,0.8811161804199219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,1,16,power_law_1.01,0.08138751983642578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,1,16,power_law_1.01,0.07874048233032227
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,1,16,power_law_1.01,0.08056832313537597
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,1,16,power_law_1.01,0.07115263938903808
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,1,16,power_law_1.01,0.07193600177764893
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,1,16,power_law_1.01,0.0744268798828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,1,16,power_law_1.01,0.07362175941467285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,1,16,power_law_1.01,0.07332352161407471
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,1,16,power_law_1.01,0.07156991958618164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,16,power_law_1.01,1.1804364776611327
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,1,16,power_law_1.01,0.0716864013671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,1,16,power_law_1.01,0.07431680202484131
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,1,16,power_law_1.01,0.07290112018585206
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,1,16,power_law_1.01,0.07847424030303955
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,1,16,power_law_1.01,0.07957248210906982
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,1,16,power_law_1.01,0.08086912155151367
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,1,16,power_law_1.01,0.0824294376373291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,16,power_law_1.01,1.5708633422851563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,1,16,power_law_1.01,0.09004032135009765
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,1,16,power_law_1.01,0.23083648681640626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,1,16,power_law_1.01,0.11135359764099122
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,1,16,power_law_1.01,0.1785036849975586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,1,16,power_law_1.01,0.10370304107666015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,1,16,power_law_1.01,0.14049407958984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,1,16,power_law_1.01,0.30820991516113283
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,16,power_law_1.2,0.18665216445922853
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,16,power_law_1.2,0.20390399932861328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,16,power_law_1.2,0.11867008209228516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,16,power_law_1.2,0.1858483123779297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,16,power_law_1.2,0.18922624588012696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,16,power_law_1.2,0.1975654411315918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,16,power_law_1.2,0.2013055992126465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,16,power_law_1.2,0.21208192825317385
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,16,power_law_1.2,0.21887487411499024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,16,power_law_1.2,0.22222463607788087
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,16,power_law_1.2,0.2517964744567871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,16,power_law_1.2,0.2547007942199707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,16,power_law_1.2,0.2694643211364746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,16,power_law_1.2,0.21303167343139648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,16,power_law_1.2,0.1792844772338867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,16,power_law_1.2,0.25091455459594725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,16,power_law_1.2,0.21518592834472655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,16,power_law_1.2,0.2808025550842285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,16,power_law_1.2,0.4221376037597656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,16,power_law_1.2,0.5084569549560547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,16,power_law_1.2,0.7120563507080078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,16,power_law_1.2,0.9100339508056641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,16,power_law_1.2,1.6058073425292967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,16,power_law_1.2,2.076299591064453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,16,power_law_1.2,0.12702976226806642
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,16,power_law_1.2,0.12548992156982422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,16,power_law_1.2,0.07902463912963867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,16,power_law_1.2,0.13084287643432618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,16,power_law_1.2,0.12103424072265626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,16,power_law_1.2,0.1290649604797363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,16,power_law_1.2,0.12134271621704103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,16,power_law_1.2,0.12314496040344239
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,16,power_law_1.2,0.1302617645263672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,16,power_law_1.2,0.35974273681640623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,16,power_law_1.2,0.13431808471679688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,16,power_law_1.2,0.14682368278503416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,16,power_law_1.2,0.14401023864746093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,16,power_law_1.2,0.1387827205657959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,16,power_law_1.2,0.1479257583618164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,16,power_law_1.2,0.1530303955078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,16,power_law_1.2,0.1578982353210449
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,16,power_law_1.2,0.18800384521484376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,16,power_law_1.2,0.2166707229614258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,16,power_law_1.2,0.2507225608825684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,16,power_law_1.2,0.32086910247802736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,16,power_law_1.2,0.43167488098144535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,16,power_law_1.2,0.5173849487304688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,16,power_law_1.2,0.8031603240966797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,16,power_law_1.2,0.9336589050292968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,1,16,power_law_1.2,0.04958720207214355
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,1,16,power_law_1.2,0.07890431880950928
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,16,power_law_1.2,1.480711669921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,1,16,power_law_1.2,0.07099647998809815
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,1,16,power_law_1.2,0.08171648025512696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,16,power_law_1.2,1.8349554443359373
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,1,16,power_law_1.2,0.07146240234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,1,16,power_law_1.2,0.07056896209716797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,1,16,power_law_1.2,0.0680191993713379
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,1,16,power_law_1.2,0.07024767875671387
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,1,16,power_law_1.2,0.0707699203491211
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,1,16,power_law_1.2,0.07423744201660157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,1,16,power_law_1.2,0.07642623901367188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,1,16,power_law_1.2,0.07768447875976563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,1,16,power_law_1.2,0.07904128074645997
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,1,16,power_law_1.01,0.45388160705566405
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,1,16,power_law_1.2,0.08234880447387696
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,1,16,power_law_1.2,0.074716157913208
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,16,power_law_1.2,0.12325887680053711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,1,16,power_law_1.01,0.5321267318725587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,16,power_law_1.2,0.6173759841918945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,16,power_law_1.2,1.1931929779052735
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,1,16,power_law_1.2,0.08648832321166992
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,1,16,power_law_1.2,0.08744704246520996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,1,16,power_law_1.2,0.07769087791442872
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,1,16,power_law_1.2,0.09953791618347167
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,1,16,power_law_1.2,0.13686911582946776
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,2,8,balanced,0.034213120937347415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,2,8,balanced,0.04944896221160889
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,2,8,balanced,0.06730368137359619
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,2,8,balanced,0.0865817642211914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,2,8,balanced,0.12967424392700194
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,2,8,balanced,0.2660876846313477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,2,8,balanced,0.2951756858825684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,2,8,balanced,0.31869312286376955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,2,8,balanced,0.25942527770996093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,1,16,power_law_1.2,0.16140031814575195
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,1,16,power_law_1.2,0.11449600219726563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,2,8,balanced,0.3287206268310547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,2,8,balanced,0.28766464233398437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,2,8,balanced,0.33696128845214846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,2,8,balanced,0.34428031921386715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,2,8,balanced,0.2661824035644531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,2,8,balanced,0.352305908203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,2,8,balanced,0.3119052886962891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,2,8,balanced,0.3555353546142578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,2,8,balanced,0.41043582916259763
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,2,8,balanced,0.37901824951171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,2,8,balanced,0.453117446899414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,2,8,balanced,0.5304832077026367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,2,8,balanced,0.7230668640136718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,2,8,balanced,0.861465606689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,2,8,balanced,1.2101055908203127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,2,8,balanced,1.5555711364746094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,2,8,balanced,2.286677703857422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,2,8,balanced,0.03406847953796387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,2,8,balanced,0.03583616018295288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,2,8,balanced,0.04468480110168457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,2,8,balanced,0.06170368194580078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,2,8,balanced,0.14615296363830566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,2,8,balanced,0.153821439743042
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,2,8,balanced,0.15837183952331543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,2,8,balanced,3.0308709716796876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,2,8,balanced,0.08923263549804687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,2,8,balanced,0.16099967956542968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,2,8,balanced,0.1630963134765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,2,8,balanced,0.15248512268066405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,2,8,balanced,0.15425151824951172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,1,16,power_law_1.2,0.21490432739257814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,2,8,balanced,0.15664640426635743
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,2,8,balanced,0.16147071838378907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,2,8,balanced,0.16342655181884766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,2,8,balanced,0.16952320098876955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,2,8,balanced,0.1858470344543457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,2,8,balanced,0.20379007339477537
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,2,8,balanced,0.2220800018310547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,2,8,balanced,0.2613337516784668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,2,8,balanced,0.2997670364379883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,2,8,balanced,0.3788787078857422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,2,8,balanced,0.4581375885009765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,2,8,balanced,0.6525030517578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,2,8,balanced,0.8053683471679687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,2,8,balanced,1.1722188568115235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,2,8,balanced,1.5151129150390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,1,16,power_law_1.2,0.30340736389160156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,1,16,power_law_1.2,0.46509441375732424
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,1,16,power_law_1.2,0.6083622360229493
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,1,16,power_law_1.2,0.9906829071044921
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,2,8,balanced,0.023532800674438477
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,2,8,balanced,0.08454015731811523
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,2,8,balanced,0.08493824005126953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,2,8,balanced,0.025629439353942872
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,2,8,balanced,0.08487296104431152
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,2,8,balanced,0.08506112098693848
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,2,8,balanced,0.02358144044876099
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,2,8,balanced,0.0861081600189209
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,2,8,balanced,0.08635775566101074
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,2,8,balanced,0.08643584251403809
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,2,8,balanced,0.08645119667053222
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,2,8,balanced,0.08782336235046387
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,2,8,balanced,0.08865152359008789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,2,8,balanced,0.09159040451049805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,2,8,balanced,0.09405311584472656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,2,8,balanced,0.09592448234558106
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,2,8,balanced,0.033342719078063965
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,2,8,balanced,0.10386560440063478
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,2,8,balanced,0.10785280227661134
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,2,8,balanced,0.05217663764953613
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,2,8,power_law_1.01,0.08287360191345214
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,2,8,power_law_1.01,0.12312576293945313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,2,8,balanced,0.08526080131530762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,2,8,power_law_1.01,0.164769287109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,2,8,balanced,0.08505599975585938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,2,8,power_law_1.01,0.1703398323059082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,2,8,power_law_1.01,0.20775808334350584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,2,8,power_law_1.01,0.24785791397094728
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,2,8,power_law_1.01,0.2660313606262207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,2,8,power_law_1.01,0.25821439743041996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,2,8,balanced,0.08467455863952636
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,2,8,power_law_1.01,0.2628198432922363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,2,8,power_law_1.01,0.27581439971923827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,2,8,power_law_1.01,0.28871936798095704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,2,8,power_law_1.01,0.24762496948242188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,2,8,power_law_1.01,0.28847103118896483
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,2,8,power_law_1.01,0.2707046318054199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,2,8,power_law_1.01,0.28762624740600584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,2,8,power_law_1.01,0.32506881713867186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,2,8,power_law_1.01,0.37158271789550784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,2,8,power_law_1.01,0.4061286544799804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,2,8,power_law_1.01,0.47951488494873046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,2,8,power_law_1.01,0.5665536117553711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,2,8,power_law_1.01,0.7680127716064453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,2,8,power_law_1.01,0.9129574584960938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,2,8,power_law_1.01,1.277692184448242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,2,8,power_law_1.01,1.69056640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,2,8,power_law_1.01,0.058234882354736325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,2,8,power_law_1.01,0.07986559867858886
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,2,8,power_law_1.01,0.10591744422912597
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,2,8,power_law_1.01,0.09506048202514648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,2,8,power_law_1.01,0.12062080383300781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,2,8,power_law_1.01,0.1115993595123291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,2,8,power_law_1.01,2.5269926452636717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,2,8,power_law_1.01,3.103084716796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,2,8,power_law_1.01,0.13765119552612304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,2,8,power_law_1.01,0.15003647804260253
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,2,8,power_law_1.01,0.14605183601379396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,2,8,power_law_1.01,0.144016637802124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,2,8,power_law_1.01,0.15116928100585938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,2,8,power_law_1.01,0.1589311981201172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,2,8,power_law_1.01,0.14952704429626465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,2,8,power_law_1.01,0.14933119773864748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,2,8,power_law_1.01,0.1520857620239258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,2,8,power_law_1.01,0.16218368530273436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,2,8,power_law_1.01,0.18964223861694335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,2,8,power_law_1.01,0.2076505661010742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,2,8,power_law_1.01,0.24837631225585938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,2,8,power_law_1.01,0.15005311965942383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,2,8,power_law_1.01,0.29321088790893557
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,2,8,power_law_1.01,0.21090431213378907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,2,8,power_law_1.01,0.4008806228637695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,2,8,power_law_1.01,0.46214015960693355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,2,8,power_law_1.01,0.6542438507080078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,2,8,power_law_1.01,0.03422976016998291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,2,8,power_law_1.01,0.7831910705566406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,2,8,power_law_1.01,0.0525491189956665
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,2,8,power_law_1.01,0.06406911849975586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,2,8,power_law_1.01,1.0984486389160157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,2,8,power_law_1.01,0.06591743946075439
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,2,8,power_law_1.01,0.07637760162353516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,2,8,power_law_1.01,0.06649343967437744
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,2,8,power_law_1.01,0.07811456203460694
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,2,8,balanced,0.2867647933959961
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,2,8,power_law_1.01,0.07887360095977783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,2,8,power_law_1.01,1.5036505126953126
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,2,8,power_law_1.01,0.0781824016571045
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,2,8,power_law_1.01,0.07990911960601807
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,2,8,power_law_1.01,0.07654528141021728
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,2,8,balanced,0.22572671890258791
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,2,8,power_law_1.01,0.08161151885986329
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,2,8,power_law_1.01,0.07890304088592529
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,2,8,power_law_1.01,0.08795519828796386
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,2,8,balanced,0.16225664138793944
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,2,8,power_law_1.01,0.08939519882202149
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,2,8,power_law_1.01,0.0872435188293457
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,2,8,balanced,0.14894847869873046
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,2,8,power_law_1.01,0.08211584091186523
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,2,8,power_law_1.01,0.13695743560791016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,2,8,power_law_1.01,0.09453311920166016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,2,8,power_law_1.01,0.10099967956542968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,2,8,power_law_1.01,0.17494144439697265
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,2,8,power_law_1.01,0.12210687637329101
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,2,8,power_law_1.2,0.0825113582611084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,2,8,power_law_1.2,0.12338432312011718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,2,8,power_law_1.2,0.1314508819580078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,2,8,power_law_1.2,0.1578265571594238
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,2,8,power_law_1.2,0.21435264587402342
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,2,8,power_law_1.01,0.2992742347717285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,2,8,power_law_1.2,0.23243392944335936
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,2,8,power_law_1.01,0.20352640151977539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,2,8,power_law_1.2,0.24739200592041016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,2,8,power_law_1.2,0.27792896270751954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,2,8,power_law_1.2,0.23689344406127927
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,2,8,power_law_1.2,0.2475596809387207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,2,8,power_law_1.2,0.2900543975830078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,2,8,power_law_1.2,0.2776486396789551
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,2,8,power_law_1.01,0.33567745208740235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,2,8,power_law_1.2,0.26664575576782223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,2,8,power_law_1.2,0.27667583465576173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,2,8,power_law_1.2,0.29120000839233395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,2,8,power_law_1.2,0.30866815567016603
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,2,8,power_law_1.2,0.357336311340332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,2,8,power_law_1.2,0.3828569412231445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,2,8,power_law_1.2,0.4278067016601563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,2,8,power_law_1.2,0.5364518356323242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,2,8,power_law_1.2,0.6495014190673828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,2,8,power_law_1.2,0.8956198120117188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,2,8,power_law_1.2,1.030808334350586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,2,8,power_law_1.2,0.05627520084381103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,2,8,power_law_1.2,0.07770751953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,2,8,power_law_1.2,0.08400768280029297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,2,8,power_law_1.2,1.5765823364257812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,2,8,power_law_1.2,0.09815168380737305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,2,8,power_law_1.2,0.12310784339904786
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,2,8,power_law_1.2,1.921925048828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,2,8,power_law_1.2,0.13126912117004394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,2,8,power_law_1.2,0.13617024421691895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,2,8,power_law_1.2,0.14019200325012207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,2,8,power_law_1.2,0.14526720046997071
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,2,8,power_law_1.2,0.1454259204864502
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,2,8,power_law_1.2,0.14995455741882324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,2,8,power_law_1.2,0.15460479736328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,2,8,power_law_1.2,0.14894847869873046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,2,8,power_law_1.2,0.15284095764160158
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,2,8,power_law_1.2,0.15328767776489258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,2,8,power_law_1.2,0.17120895385742188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,2,8,power_law_1.2,0.19690752029418945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,2,8,power_law_1.2,3.0865869140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,2,8,power_law_1.2,0.22423423767089842
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,2,8,power_law_1.2,0.2647065544128418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,2,8,power_law_1.2,0.3160908889770508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,2,8,power_law_1.2,0.4334988784790039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,2,8,power_law_1.2,0.5005401611328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,2,8,power_law_1.2,0.14974464416503905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,2,8,power_law_1.2,3.7853927612304688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,2,8,power_law_1.2,0.03452415943145752
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,2,8,power_law_1.2,0.05177343845367431
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,2,8,power_law_1.2,0.6660889434814453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,2,8,power_law_1.2,0.054042878150939944
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,2,8,power_law_1.2,0.8698611450195312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,2,8,power_law_1.2,0.06422912120819092
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,2,8,power_law_1.2,0.07210879802703858
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,2,8,power_law_1.2,0.07227519989013673
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,2,8,power_law_1.2,0.07341440200805664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,2,8,power_law_1.2,0.0762879991531372
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,2,8,power_law_1.2,0.07900928020477295
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,2,8,power_law_1.2,0.08075263977050781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,2,8,power_law_1.2,0.08283391952514649
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,2,8,power_law_1.2,1.3665957641601563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,2,8,power_law_1.2,0.08437631607055665
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,2,8,power_law_1.2,0.08395263671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,2,8,power_law_1.2,0.08313599586486817
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,2,8,power_law_1.2,0.08758272171020508
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,2,8,power_law_1.2,0.0838003158569336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,2,8,power_law_1.2,1.603558349609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,2,8,power_law_1.2,0.09365504264831544
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,2,8,power_law_1.2,0.10102656364440918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,2,8,power_law_1.01,0.518732795715332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,2,8,power_law_1.01,0.7260390472412109
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,2,8,power_law_1.2,0.14080256462097168
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,2,8,power_law_1.2,0.1656537628173828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,2,8,power_law_1.2,0.11692543983459473
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,4,4,balanced,0.031692800521850584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,2,8,power_law_1.2,0.21255424499511716
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,4,4,balanced,0.06656000137329102
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,2,8,power_law_1.2,0.2636313629150391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,4,4,balanced,0.12473600387573243
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,2,8,power_law_1.2,0.41757694244384763
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,4,4,balanced,0.2583219146728516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,4,4,balanced,0.25678335189819335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,4,4,balanced,0.26075647354125975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,4,4,balanced,0.24379903793334962
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,4,4,balanced,0.25709440231323244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,4,4,balanced,0.05469823837280273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,4,4,balanced,0.25748863220214846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,4,4,balanced,0.08421631813049316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,4,4,balanced,0.27922815322875977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,4,4,balanced,0.3276019287109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,4,4,balanced,0.26595071792602537
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,4,4,balanced,0.27377407073974613
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,4,4,balanced,0.2797337532043457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,4,4,balanced,0.29459968566894534
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,4,4,balanced,0.32713344573974606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,4,4,balanced,0.36213504791259765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,4,4,balanced,0.4347391891479492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,4,4,balanced,0.5075136184692383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,4,4,balanced,0.6814604949951172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,4,4,balanced,0.8061119842529297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,4,4,balanced,1.139441909790039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,4,4,balanced,0.033089280128479004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,4,4,balanced,0.03671295881271362
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,4,4,balanced,0.047330560684204104
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,4,4,balanced,1.460669403076172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,4,4,balanced,0.07044991970062256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,4,4,balanced,0.09463935852050781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,4,4,balanced,0.15131263732910155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,4,4,balanced,0.15257472038269043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,4,4,balanced,0.15497599601745607
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,4,4,balanced,0.1696294403076172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,4,4,balanced,0.17072383880615233
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,4,4,balanced,0.17194368362426757
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,4,4,balanced,2.134845428466797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,4,4,balanced,0.17538047790527345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,4,4,balanced,0.17741567611694337
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,4,4,balanced,0.18226943969726564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,4,4,balanced,0.1838809585571289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,4,4,balanced,0.18726783752441406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,4,4,balanced,0.1991001510620117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,4,4,balanced,0.21954303741455078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,4,4,balanced,0.23433216094970702
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,4,4,balanced,0.27058303833007813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,4,4,balanced,0.3104729652404785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,4,4,balanced,2.816264953613281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,4,4,balanced,0.37074817657470704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,4,4,balanced,0.4486105728149414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,4,4,balanced,0.6534770965576172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,4,4,balanced,0.7904653167724609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,4,4,balanced,1.1453017425537109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,4,4,balanced,1.4827122497558594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,2,8,power_law_1.2,0.5252646255493164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,2,8,power_law_1.2,0.8414399719238281
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,2,8,power_law_1.2,1.160680923461914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,4,4,balanced,0.028887040615081787
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,4,4,balanced,0.09701120376586914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,4,4,balanced,0.029859840869903564
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,4,4,balanced,0.09731072425842285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,4,4,balanced,0.09784192085266114
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,4,4,balanced,0.09763199806213378
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,4,4,balanced,0.032561919689178466
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,4,4,balanced,0.09928959846496582
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,4,4,balanced,0.1002227210998535
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,4,4,balanced,0.04236671924591064
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,4,4,balanced,0.10051967620849608
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,4,4,balanced,0.06225920200347901
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,4,4,balanced,0.10270848274230956
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,4,4,balanced,0.10528639793395995
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,4,4,balanced,0.1085324764251709
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,4,4,balanced,0.1130073642730713
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,4,4,balanced,0.11762432098388671
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,4,4,balanced,0.1292569637298584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,4,4,balanced,0.1381772804260254
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,4,4,balanced,0.09777664184570313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,4,4,power_law_1.01,0.06575359821319579
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,4,4,power_law_1.01,0.08262656211853028
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,4,4,power_law_1.01,0.10080896377563478
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,4,4,power_law_1.01,0.10934528350830078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,4,4,power_law_1.01,0.1385420799255371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,4,4,power_law_1.01,0.16011903762817384
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,4,4,power_law_1.01,0.20380544662475586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,4,4,power_law_1.01,0.20503679275512696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,4,4,power_law_1.01,0.21705215454101562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,4,4,power_law_1.01,0.2278463935852051
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,4,4,power_law_1.01,0.2248409652709961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,4,4,power_law_1.01,0.22941823959350588
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,4,4,power_law_1.01,0.23493888854980466
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,4,4,power_law_1.01,0.24058752059936522
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,4,4,power_law_1.01,0.25006336212158203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,4,4,power_law_1.01,0.2596441650390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,4,4,power_law_1.01,0.32621696472167966
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,4,4,power_law_1.01,0.37870849609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,4,4,balanced,0.09669247627258301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,4,4,power_law_1.01,0.4049817657470703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,4,4,power_law_1.01,0.46151550292968746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,4,4,power_law_1.01,0.5635865783691406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,4,4,power_law_1.01,0.7376051330566407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,4,4,power_law_1.01,0.882525405883789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,4,4,power_law_1.01,1.247399673461914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,4,4,balanced,0.09739904403686524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,4,4,power_law_1.01,1.631957702636719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,4,4,power_law_1.01,0.039153919219970704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,4,4,power_law_1.01,2.337212219238281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,4,4,power_law_1.01,0.06019199848175048
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,4,4,power_law_1.01,3.033074035644531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,4,4,power_law_1.01,0.07223296165466309
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,4,4,power_law_1.01,0.08547072410583496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,4,4,power_law_1.01,0.10849791526794435
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,4,4,power_law_1.01,0.11358336448669433
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,4,4,power_law_1.01,0.134204158782959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,4,4,power_law_1.01,0.14099328041076659
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,4,4,power_law_1.01,0.15757823944091798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,4,4,power_law_1.01,0.15566592216491698
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,4,4,power_law_1.01,0.15933823585510254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,4,4,power_law_1.01,0.1634739112854004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,4,4,power_law_1.01,0.1659110450744629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,4,4,power_law_1.01,0.16604032516479492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,4,4,power_law_1.01,0.17062143325805662
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,4,4,power_law_1.01,0.17579391479492187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,4,4,power_law_1.01,0.1908172798156738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,4,4,power_law_1.01,0.20988544464111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,4,4,power_law_1.01,0.22919551849365236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,4,4,power_law_1.01,0.2731507110595703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,4,4,power_law_1.01,0.3227084732055664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,4,4,power_law_1.01,0.3984511947631836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,4,4,power_law_1.01,0.5098547363281251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,4,4,power_law_1.01,0.6599616241455079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,4,4,power_law_1.01,0.8509657287597656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,4,4,balanced,0.09699199676513673
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,4,4,power_law_1.01,0.03206912040710449
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,4,4,power_law_1.01,0.044995841979980467
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,4,4,power_law_1.01,0.05278975963592529
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,4,4,power_law_1.01,0.05795328140258789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,4,4,power_law_1.01,1.222302703857422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,4,4,power_law_1.01,0.06558335781097412
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,4,4,power_law_1.01,0.0682431983947754
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,4,4,power_law_1.01,0.08310400009155273
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,4,4,power_law_1.01,0.0835801601409912
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,4,4,power_law_1.01,1.5205657958984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,4,4,power_law_1.01,0.08560511589050293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,4,4,power_law_1.01,0.08792448043823242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,4,4,power_law_1.01,0.09096575736999511
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,4,4,power_law_1.01,0.09272704124450684
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,4,4,power_law_1.01,0.09383296012878418
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,4,4,power_law_1.01,0.09590656280517577
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,4,4,power_law_1.01,0.10144127845764159
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,4,4,power_law_1.01,0.10051967620849608
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,4,4,power_law_1.01,0.10303104400634766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,4,4,balanced,0.19731456756591798
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,4,4,balanced,0.22098304748535158
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,4,4,balanced,0.31325439453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,4,4,balanced,0.40130176544189455
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,4,4,power_law_1.01,0.12751359939575196
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,4,4,power_law_1.01,0.11651071548461914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,4,4,power_law_1.01,0.18518655776977538
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,4,4,power_law_1.01,0.14797311782836914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,4,4,power_law_1.2,0.08226304054260254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,4,4,power_law_1.2,0.0879308795928955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,4,4,power_law_1.2,0.06463359832763672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,4,4,power_law_1.2,0.10963839530944823
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,4,4,power_law_1.2,0.13841535568237304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,4,4,power_law_1.2,0.1582156753540039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,4,4,power_law_1.2,0.21245439529418944
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,4,4,power_law_1.2,0.20919679641723632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,4,4,power_law_1.2,0.21361536026000977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,4,4,power_law_1.2,0.22017663955688477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,4,4,power_law_1.2,0.22526592254638672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,4,4,power_law_1.2,0.22690303802490233
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,4,4,power_law_1.2,0.23457536697387696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,4,4,power_law_1.2,0.24255744934082032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,4,4,power_law_1.2,0.25288576126098633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,4,4,power_law_1.2,0.26506368637084965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,4,4,power_law_1.2,0.3241689682006836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,4,4,power_law_1.2,0.3818892669677735
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,4,4,power_law_1.01,0.2271449661254883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,4,4,power_law_1.2,0.4148953628540039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,4,4,power_law_1.2,0.4847603225708008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,4,4,power_law_1.2,0.6065792083740235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,4,4,power_law_1.2,0.759820785522461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,4,4,power_law_1.01,0.29571712493896485
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,4,4,power_law_1.01,0.4361344146728515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,4,4,power_law_1.2,0.9257292938232421
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,4,4,power_law_1.2,0.03872895956039429
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,4,4,power_law_1.01,0.4958617782592773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,4,4,power_law_1.2,1.3632818603515626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,4,4,power_law_1.2,0.05978623867034912
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,4,4,power_law_1.2,0.06596608161926269
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,4,4,power_law_1.2,0.08477312088012695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,4,4,power_law_1.2,0.10039423942565917
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,4,4,power_law_1.2,1.6471539306640626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,4,4,power_law_1.2,0.11070079803466797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,4,4,power_law_1.2,0.13635199546813964
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,4,4,power_law_1.2,0.15460479736328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,4,4,power_law_1.2,0.14151040077209473
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,4,4,power_law_1.2,0.1555072021484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,4,4,power_law_1.2,0.1571507167816162
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,4,4,power_law_1.2,0.16138111114501955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,4,4,power_law_1.2,0.1640012741088867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,4,4,power_law_1.2,2.414451141357422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,4,4,power_law_1.2,0.168404483795166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,4,4,power_law_1.2,0.1727052879333496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,4,4,power_law_1.2,0.17728384017944337
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,4,4,power_law_1.2,0.19307647705078126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,4,4,power_law_1.2,0.21573375701904296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,4,4,power_law_1.2,0.23453311920166015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,4,4,power_law_1.2,0.28837120056152343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,4,4,power_law_1.2,0.33446784973144533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,4,4,power_law_1.2,0.42817150115966796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,4,4,power_law_1.2,3.3701901245117183
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,4,4,power_law_1.2,0.5122086334228515
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,4,4,power_law_1.2,0.031792640686035156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,4,4,power_law_1.2,0.7028915405273437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,4,4,power_law_1.2,0.04507008075714111
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,4,4,power_law_1.2,0.04705920219421387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,4,4,power_law_1.2,0.8829094696044922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,4,4,power_law_1.2,0.05725696086883545
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,4,4,power_law_1.2,0.06598271846771239
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,4,4,power_law_1.2,0.08028927803039551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,4,4,power_law_1.2,1.290918426513672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,4,4,power_law_1.2,0.08272512435913086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,4,4,power_law_1.2,0.08582655906677246
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,4,4,power_law_1.2,0.08775424003601075
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,4,4,power_law_1.2,0.08837504386901855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,4,4,power_law_1.2,1.7229209899902345
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,4,4,power_law_1.2,0.09387776374816895
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,4,4,power_law_1.2,0.09397120475769043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,4,4,power_law_1.2,0.09664511680603027
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,4,4,power_law_1.2,0.10394880294799805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,4,4,power_law_1.2,0.10507776260375976
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,4,4,power_law_1.2,0.11305215835571289
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,4,4,power_law_1.2,0.14697728157043458
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,4,4,power_law_1.2,0.06854656219482422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,4,4,power_law_1.01,0.8018777465820313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,4,4,power_law_1.01,1.0501798248291014
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,4,4,power_law_1.2,0.12749567985534668
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,4,4,power_law_1.2,0.3109875106811523
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,4,4,power_law_1.2,0.23844863891601564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,8,2,balanced,0.03178240060806274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,8,2,balanced,0.053777918815612794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,8,2,balanced,0.06624639987945556
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,4,4,power_law_1.2,0.5725657653808593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,8,2,balanced,0.08649472236633302
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,8,2,balanced,0.1316108798980713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,8,2,balanced,0.2671295928955078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,8,2,balanced,0.2686169624328613
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,8,2,balanced,0.26952703475952144
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,4,4,power_law_1.2,0.17740543365478517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,8,2,balanced,0.27511680603027344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,8,2,balanced,0.2715519905090332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,8,2,balanced,0.27321088790893555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,8,2,balanced,0.2741708755493164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,8,2,balanced,0.2778879928588867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,8,2,balanced,0.28404735565185546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,8,2,balanced,0.28749696731567387
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,4,4,power_law_1.2,0.41159294128417967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,8,2,balanced,0.292043514251709
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,8,2,balanced,0.31328384399414066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,8,2,balanced,0.33707393646240236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,8,2,balanced,0.3597721481323242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,8,2,balanced,0.4386931228637695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,8,2,balanced,0.5067200088500977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,8,2,balanced,0.6865395355224609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,8,2,balanced,0.8009369659423828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,8,2,balanced,0.033704960346221925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,8,2,balanced,1.1275904083251953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,8,2,balanced,0.03790080070495606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,8,2,balanced,0.05051519870758057
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,8,2,balanced,0.07596288204193116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,8,2,balanced,1.4346649169921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,8,2,balanced,0.12136704444885253
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,8,2,balanced,0.19400960922241212
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,8,2,balanced,0.1922470474243164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,8,2,balanced,0.19241344451904296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,8,2,balanced,0.19364095687866212
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,8,2,balanced,0.19533952713012695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,8,2,balanced,0.19750400543212893
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,8,2,balanced,2.0738726806640626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,8,2,balanced,0.22510847091674804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,8,2,balanced,0.22622079849243165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,8,2,balanced,0.2307468795776367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,8,2,balanced,0.23188608169555663
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,8,2,balanced,0.23432064056396484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,8,2,balanced,0.24304639816284182
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,8,2,balanced,0.25729791641235356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,8,2,balanced,0.27446144104003906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,8,2,balanced,0.3048575973510742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,8,2,balanced,2.732025451660156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,8,2,balanced,0.34000511169433595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,8,2,balanced,0.40346111297607423
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,8,2,balanced,0.4668377685546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,8,2,balanced,0.7074752044677735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,8,2,balanced,0.8342476654052735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,8,2,balanced,1.2113868713378906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,8,2,balanced,1.5665728759765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,4,4,power_law_1.2,0.7498700714111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,4,4,power_law_1.2,1.2749005126953126
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,4,4,power_law_1.2,1.5904832458496094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,8,2,balanced,0.04159615993499756
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,8,2,balanced,0.04759039878845215
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,8,2,balanced,0.042383360862731936
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,8,2,balanced,0.1280345630645752
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,8,2,balanced,0.1283827209472656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,8,2,balanced,0.12927488327026365
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,8,2,balanced,0.12985728263854981
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,8,2,balanced,0.12926719665527345
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,8,2,balanced,0.1314035224914551
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,8,2,balanced,0.13297151565551757
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,8,2,balanced,0.05959167957305909
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,8,2,balanced,0.1375551986694336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,8,2,balanced,0.08393471717834473
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,8,2,balanced,0.14215295791625976
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,8,2,balanced,0.14728320121765137
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,8,2,balanced,0.15694208145141603
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,8,2,balanced,0.16636159896850586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,8,2,balanced,0.1875584030151367
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,8,2,balanced,0.2091929626464844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,8,2,balanced,0.12790016174316404
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,8,2,power_law_1.01,0.03465728044509887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,8,2,power_law_1.01,0.06705023765563964
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,8,2,power_law_1.01,0.08093824386596679
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,8,2,power_law_1.01,0.09558783531188965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,8,2,power_law_1.01,0.1244966411590576
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,8,2,balanced,0.12764415740966797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,8,2,power_law_1.01,0.14775679588317872
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,8,2,power_law_1.01,0.18669567108154297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,8,2,power_law_1.01,0.18852479934692384
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,8,2,power_law_1.01,0.18913280487060546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,8,2,power_law_1.01,0.2024051284790039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,8,2,power_law_1.01,0.21153791427612303
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,8,2,power_law_1.01,0.25144832611083984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,8,2,power_law_1.01,0.2565171241760254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,8,2,power_law_1.01,0.2685132789611816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,8,2,power_law_1.01,0.26829824447631834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,8,2,power_law_1.01,0.2802252769470215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,8,2,power_law_1.01,0.3041779136657715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,8,2,power_law_1.01,0.3739315032958984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,8,2,power_law_1.01,0.42370689392089844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,8,2,power_law_1.01,0.48149761199951174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,8,2,power_law_1.01,0.5553817749023438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,8,2,power_law_1.01,0.7214617919921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,8,2,power_law_1.01,0.8785932922363282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,8,2,power_law_1.01,1.2241702270507813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,8,2,power_law_1.01,1.530022430419922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,8,2,power_law_1.01,2.208248291015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,8,2,power_law_1.01,0.03703167915344238
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,8,2,power_law_1.01,0.04312448024749756
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,8,2,balanced,0.1276595211029053
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,8,2,power_law_1.01,0.0640064001083374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,8,2,power_law_1.01,2.8835736083984376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,8,2,power_law_1.01,0.08342399597167968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,8,2,power_law_1.01,0.11753087997436525
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,8,2,power_law_1.01,0.13107199668884278
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,8,2,power_law_1.01,0.16063743591308594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,8,2,power_law_1.01,0.16855552673339844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,8,2,power_law_1.01,0.17127040863037107
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,8,2,power_law_1.01,0.17152896881103516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,8,2,power_law_1.01,0.17796096801757813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,8,2,power_law_1.01,0.20011775970458984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,8,2,power_law_1.01,0.20502656936645508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,8,2,power_law_1.01,0.21417728424072266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,8,2,power_law_1.01,0.21238655090332031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,8,2,power_law_1.01,0.21844863891601562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,8,2,power_law_1.01,0.23104639053344728
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,8,2,power_law_1.01,0.24434560775756836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,8,2,power_law_1.01,0.27011455535888673
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,8,2,power_law_1.01,0.31756927490234377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,8,2,power_law_1.01,0.36849536895751955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,8,2,power_law_1.01,0.4532953643798828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,8,2,power_law_1.01,0.5652019119262695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,8,2,power_law_1.01,0.7369100952148437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,8,2,power_law_1.01,0.9220620727539062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,8,2,power_law_1.01,0.04289023876190186
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,8,2,power_law_1.01,0.047386879920959475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,8,2,power_law_1.01,1.289564208984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,8,2,power_law_1.01,0.057043199539184575
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,8,2,power_law_1.01,0.06509312152862548
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,8,2,power_law_1.01,0.07606143951416015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,8,2,power_law_1.01,0.08747520446777343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,8,2,power_law_1.01,1.6666905212402345
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,8,2,power_law_1.01,0.11059583663940428
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,8,2,power_law_1.01,0.11131775856018067
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,8,2,power_law_1.01,0.11292287826538086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,8,2,power_law_1.01,0.11651328086853027
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,8,2,power_law_1.01,0.12076671600341797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,8,2,power_law_1.01,0.12301440238952636
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,8,2,power_law_1.01,0.131146240234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,8,2,power_law_1.01,0.13661824226379396
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,8,2,power_law_1.01,0.1435929584503174
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,8,2,power_law_1.01,0.10951295852661133
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,8,2,power_law_1.01,0.1496025562286377
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,8,2,balanced,0.3577036666870117
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,8,2,balanced,0.31302528381347655
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,8,2,balanced,0.515365104675293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,8,2,balanced,0.6612992095947265
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,8,2,power_law_1.01,0.17512832641601564
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,8,2,power_law_1.01,0.20107904434204102
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,8,2,power_law_1.01,0.23535871505737305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,8,2,power_law_1.01,0.5271270370483399
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,8,2,power_law_1.01,0.28419328689575196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,8,2,power_law_1.2,0.033835520744323735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,8,2,power_law_1.2,0.06507904052734374
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,8,2,power_law_1.01,0.38650878906249997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,8,2,power_law_1.2,0.07497600078582764
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,8,2,power_law_1.2,0.09019136428833008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,8,2,power_law_1.2,0.11854592323303223
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,8,2,power_law_1.01,0.7493670654296876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,8,2,power_law_1.2,0.1287795162200928
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,8,2,power_law_1.2,0.17178752899169922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,8,2,power_law_1.2,0.18128511428833008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,8,2,power_law_1.2,0.18868223190307618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,8,2,power_law_1.2,0.1996019172668457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,8,2,power_law_1.2,0.21640447616577146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,8,2,power_law_1.2,0.2485055923461914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,8,2,power_law_1.2,0.25613824844360356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,8,2,power_law_1.2,0.2643878364562988
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,8,2,power_law_1.2,0.2708774375915527
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,8,2,power_law_1.2,0.2777164840698242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,8,2,power_law_1.2,0.32739585876464844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,8,2,power_law_1.2,0.38220672607421874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,8,2,power_law_1.2,0.42663040161132815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,8,2,power_law_1.2,0.4894003295898438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,8,2,power_law_1.2,0.5668659210205078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,8,2,power_law_1.2,0.730439682006836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,8,2,power_law_1.2,0.8864409637451172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,8,2,power_law_1.2,1.2567628479003905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,8,2,power_law_1.2,0.0366758394241333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,8,2,power_law_1.2,1.5646873474121095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,8,2,power_law_1.2,0.04283135890960693
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,8,2,power_law_1.2,0.05924223899841309
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,8,2,power_law_1.2,0.07918335914611815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,8,2,power_law_1.2,0.11179264068603516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,8,2,power_law_1.2,2.258059539794922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,8,2,power_law_1.2,0.12274815559387206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,8,2,power_law_1.2,0.1618739128112793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,8,2,power_law_1.2,0.16763008117675782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,8,2,power_law_1.2,0.16779775619506837
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,8,2,power_law_1.2,0.1727142333984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,8,2,power_law_1.2,0.17511423110961916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,8,2,power_law_1.2,0.1982271957397461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,8,2,power_law_1.2,3.0109298706054686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,8,2,power_law_1.2,0.20155263900756837
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,8,2,power_law_1.2,0.20854143142700193
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,8,2,power_law_1.2,0.212487678527832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,8,2,power_law_1.2,0.2161868858337402
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,8,2,power_law_1.2,0.23851135253906247
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,8,2,power_law_1.2,0.2524620819091797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,8,2,power_law_1.2,0.2810982322692871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,8,2,power_law_1.2,0.3452492904663086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,8,2,power_law_1.2,0.3862374496459961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,8,2,power_law_1.2,0.4669811248779297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,8,2,power_law_1.2,0.5751948928833007
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,8,2,power_law_1.2,0.7631769561767578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,8,2,power_law_1.2,0.9445836639404297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,8,2,power_law_1.2,1.3141439819335938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,8,2,power_law_1.2,0.04275712013244629
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,8,2,power_law_1.2,0.047884798049926756
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,8,2,power_law_1.2,1.6982949829101563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,8,2,power_law_1.2,0.054275841712951654
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,8,2,power_law_1.2,0.06257791996002197
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,8,2,power_law_1.2,0.07323520183563233
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,8,2,power_law_1.2,0.08454655647277833
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,8,2,power_law_1.2,0.10966783523559571
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,8,2,power_law_1.2,0.11264512062072754
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,8,2,power_law_1.2,0.11275903701782226
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,8,2,power_law_1.2,0.11770879745483398
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,8,2,power_law_1.2,0.12262144088745117
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,8,2,power_law_1.2,0.1258291244506836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,8,2,power_law_1.2,0.1366988754272461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,8,2,power_law_1.2,0.14290176391601564
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,8,2,power_law_1.2,0.14969344139099122
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,8,2,power_law_1.2,0.1634444808959961
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,8,2,power_law_1.2,0.19707775115966797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,8,2,power_law_1.2,0.23724031448364258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,8,2,power_law_1.2,0.10775424003601075
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,8,2,power_law_1.01,1.0104064178466796
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,8,2,power_law_1.01,1.2852838134765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,8,2,power_law_1.01,1.833790740966797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,8,2,power_law_1.2,0.7430182647705078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,8,2,power_law_1.2,0.2902553558349609
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,8,2,power_law_1.2,0.3668288040161133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,16,1,balanced,0.04114431858062744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,16,1,balanced,0.06531199932098389
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,8,2,power_law_1.2,0.5356556701660156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,16,1,balanced,0.07826432228088379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,16,1,balanced,0.09984383583068848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,16,1,balanced,0.1459776020050049
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,16,1,balanced,0.30174144744873044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,16,1,balanced,0.3008460807800293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,16,1,balanced,0.29835136413574215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,16,1,balanced,0.30433471679687496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,16,1,balanced,0.31138816833496097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,16,1,balanced,0.3071878433227539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,16,1,balanced,0.3096268844604492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,16,1,balanced,0.3166124725341797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,16,1,balanced,0.31907072067260744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,16,1,balanced,0.36828094482421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,16,1,balanced,0.375997428894043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,16,1,balanced,0.38532032012939454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,16,1,balanced,0.4104102325439453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,16,1,balanced,0.43714046478271484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,16,1,balanced,0.4894591903686523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,16,1,balanced,0.5520364761352539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,16,1,balanced,0.7653215789794923
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,16,1,balanced,0.8646892547607422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,16,1,balanced,1.2068716430664064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,16,1,balanced,1.5296377563476562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,16,1,balanced,0.04278463840484619
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,16,1,balanced,0.04925248146057129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,16,1,balanced,0.06553855895996094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,16,1,balanced,2.1774598693847658
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,16,1,balanced,0.09749119758605958
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,16,1,balanced,0.15616640090942382
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,16,1,balanced,0.27724031448364256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,16,1,balanced,0.2797311973571778
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,8,2,power_law_1.2,1.0387213134765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,16,1,balanced,0.280565128326416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,16,1,balanced,0.28285247802734376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,16,1,balanced,0.2855411148071289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,16,1,balanced,0.28872959136962895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,16,1,balanced,0.2930886459350586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,16,1,balanced,2.82594482421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,16,1,balanced,0.2969280052185058
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,16,1,balanced,0.30147199630737304
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,8,2,power_law_1.2,1.447097625732422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,16,1,balanced,0.331091194152832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,16,1,balanced,0.3350777435302734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,16,1,balanced,0.34223041534423826
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,16,1,balanced,0.3557651138305664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,16,1,balanced,0.3699967956542969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,16,1,balanced,0.3998220825195312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,16,1,balanced,0.4278086471557617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,16,1,balanced,0.48866943359375004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,16,1,balanced,0.548017921447754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,16,1,balanced,0.8523878479003907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,16,1,balanced,0.970348129272461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,16,1,balanced,1.3964505004882812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,16,1,balanced,1.812742462158203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,8,2,power_law_1.2,2.619928283691406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,16,1,balanced,0.06981503963470459
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,16,1,balanced,0.06688511848449707
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,16,1,balanced,0.07785920143127442
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,16,1,balanced,0.09272128105163574
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,16,1,balanced,0.12656384468078613
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,16,1,balanced,0.19569856643676758
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,16,1,balanced,0.19773183822631837
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,16,1,balanced,0.19640512466430665
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,16,1,balanced,0.19368383407592774
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,16,1,balanced,0.19945728302001953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,16,1,balanced,0.20100671768188477
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,16,1,balanced,0.19564607620239258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,16,1,balanced,0.20687744140625003
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,16,1,balanced,0.21102464675903318
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,16,1,balanced,0.2041721534729004
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,16,1,balanced,0.231026554107666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,16,1,balanced,0.22130111694335936
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,16,1,balanced,0.27138368606567387
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,16,1,balanced,0.31497152328491207
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,16,1,balanced,0.25119039535522464
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,16,1,balanced,0.35770816802978517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,16,1,power_law_1.01,0.04303167819976807
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,16,1,power_law_1.01,0.05147776126861572
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,16,1,power_law_1.01,0.07852352142333985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,16,1,power_law_1.01,0.09988800048828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,16,1,balanced,0.19132415771484373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,16,1,power_law_1.01,0.12786687850952147
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,16,1,power_law_1.01,0.21019519805908202
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,16,1,power_law_1.01,0.2172480010986328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,16,1,power_law_1.01,0.22074752807617187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,16,1,power_law_1.01,0.22312768936157226
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,16,1,balanced,0.19350784301757812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,16,1,power_law_1.01,0.2366169548034668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,16,1,power_law_1.01,0.2736972808837891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,16,1,power_law_1.01,0.28153856277465816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,16,1,power_law_1.01,0.2995948791503906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,16,1,power_law_1.01,0.35041984558105466
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,16,1,power_law_1.01,0.3579878234863281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,16,1,power_law_1.01,0.3795315170288086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,16,1,power_law_1.01,0.4275795364379883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,16,1,power_law_1.01,0.5039596939086914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,16,1,power_law_1.01,0.5630214309692383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,16,1,power_law_1.01,0.6310688018798828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,16,1,power_law_1.01,0.7964179229736328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,16,1,power_law_1.01,0.1637228775024414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,16,1,power_law_1.01,0.9459929656982421
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,16,1,power_law_1.01,1.2683545684814452
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,16,1,power_law_1.01,1.5834988403320314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,16,1,power_law_1.01,2.236890869140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,16,1,power_law_1.01,0.045148801803588864
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,16,1,power_law_1.01,0.05001152038574219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,16,1,power_law_1.01,0.09723711967468261
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,16,1,power_law_1.01,0.13507776260375975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,16,1,power_law_1.01,0.06930816173553467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,16,1,power_law_1.01,2.8814245605468747
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,16,1,power_law_1.01,0.1885625648498535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,16,1,power_law_1.01,0.23687999725341796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,16,1,power_law_1.01,0.2446348762512207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,16,1,power_law_1.01,0.2475961685180664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,16,1,power_law_1.01,0.2504332733154297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,16,1,power_law_1.01,0.2595232009887695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,16,1,power_law_1.01,0.2658841514587402
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,16,1,power_law_1.01,0.27119167327880855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,16,1,power_law_1.01,0.2774176025390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,16,1,power_law_1.01,0.30618623733520506
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,16,1,power_law_1.01,0.3087750434875488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,16,1,power_law_1.01,0.3232640075683594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,16,1,power_law_1.01,0.34279422760009765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,16,1,power_law_1.01,0.3677299118041992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,16,1,power_law_1.01,0.416734733581543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,16,1,power_law_1.01,0.4755148696899414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,16,1,power_law_1.01,0.5719795227050781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,16,1,power_law_1.01,0.8754265594482422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,16,1,power_law_1.01,0.6667142486572265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,16,1,power_law_1.01,1.0917375946044923
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,16,1,balanced,0.5514995193481445
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,16,1,balanced,0.6430489349365234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,16,1,power_law_1.01,1.5139768981933595
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,16,1,balanced,0.9285491180419921
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,16,1,power_law_1.01,0.06975488185882568
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,16,1,power_law_1.01,0.06672255992889405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,16,1,power_law_1.01,1.9322149658203123
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,16,1,power_law_1.01,0.0819705581665039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,16,1,power_law_1.01,0.09329471588134766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,16,1,balanced,1.2037926483154298
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,16,1,power_law_1.01,0.11241663932800292
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,16,1,power_law_1.01,0.12809856414794923
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,16,1,power_law_1.01,0.15739520072937013
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,16,1,power_law_1.01,0.16768640518188477
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,16,1,power_law_1.01,0.16415359497070312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,16,1,power_law_1.01,0.17040000915527345
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,16,1,power_law_1.01,0.17761920928955077
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,16,1,power_law_1.01,0.1877702331542969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,16,1,power_law_1.01,0.18390464782714844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,16,1,power_law_1.01,0.1941823959350586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,16,1,power_law_1.01,0.20482496261596678
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,16,1,power_law_1.01,0.21233728408813474
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,16,1,power_law_1.01,0.22827520370483398
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,16,1,power_law_1.01,0.3028857612609863
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,16,1,power_law_1.01,0.2642144012451172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,16,1,power_law_1.01,0.45951488494873044
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,16,1,power_law_1.01,0.3801375961303711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,16,1,power_law_1.01,0.6122598266601562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,16,1,power_law_1.2,0.040663042068481446
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,16,1,power_law_1.01,0.7809331512451172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,16,1,power_law_1.2,0.05407104015350341
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,16,1,power_law_1.2,0.07884607791900636
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,16,1,power_law_1.2,0.0981446361541748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,16,1,power_law_1.2,0.1216703987121582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,16,1,power_law_1.2,0.15870911598205567
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,16,1,power_law_1.2,0.20780288696289065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,16,1,power_law_1.2,0.21271423339843748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,16,1,power_law_1.2,0.2184543991088867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,16,1,power_law_1.2,0.22466047286987303
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,16,1,power_law_1.01,1.1686573028564453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,16,1,power_law_1.2,0.24094463348388673
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,16,1,power_law_1.2,0.27276800155639647
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,16,1,power_law_1.2,0.2841190338134766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,16,1,power_law_1.2,0.30909952163696286
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,16,1,power_law_1.01,1.4966744995117187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,16,1,power_law_1.2,0.35297790527343753
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,16,1,power_law_1.2,0.36528255462646486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,16,1,power_law_1.2,0.38400833129882816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,16,1,power_law_1.2,0.45273086547851565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,16,1,power_law_1.2,0.5673356628417968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,16,1,power_law_1.2,0.5064915084838868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,16,1,power_law_1.2,0.6510508728027344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,16,1,power_law_1.2,0.8232307434082031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,16,1,power_law_1.2,0.960271987915039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,16,1,power_law_1.2,0.04450047969818115
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,16,1,power_law_1.2,1.2809298706054686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,16,1,power_law_1.2,0.05072959899902344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,16,1,power_law_1.2,1.5988613891601564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,16,1,power_law_1.2,0.06955904006958008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,16,1,power_law_1.2,0.09425151824951172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,16,1,power_law_1.2,0.1301356792449951
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,16,1,power_law_1.2,0.1855251121520996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,16,1,power_law_1.2,0.23480255126953126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,16,1,power_law_1.2,0.24184383392333983
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,16,1,power_law_1.2,0.24600383758544925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,16,1,power_law_1.2,0.2520710372924805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,16,1,power_law_1.2,0.25807296752929687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,16,1,power_law_1.2,0.2643398475646973
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,16,1,power_law_1.2,2.890666198730469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,16,1,power_law_1.2,0.2709049606323242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,16,1,power_law_1.2,2.2504454040527344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,16,1,power_law_1.2,0.2808870315551758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,16,1,power_law_1.2,0.30880256652832033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,16,1,power_law_1.2,0.31478784561157225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,16,1,power_law_1.2,0.3277695846557617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,16,1,power_law_1.2,0.35486080169677736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,16,1,power_law_1.2,0.37790721893310547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,16,1,power_law_1.2,0.4406623840332031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,16,1,power_law_1.2,0.4885305786132813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,16,1,power_law_1.2,0.5892959976196289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,16,1,power_law_1.2,0.7052460479736328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,16,1,power_law_1.2,0.8863788604736328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,16,1,power_law_1.2,1.1068972778320312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,16,1,power_law_1.2,0.06659200191497802
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,16,1,power_law_1.2,1.525992889404297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,16,1,power_law_1.2,0.0693452787399292
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,16,1,power_law_1.2,1.9513063049316408
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,16,1,power_law_1.2,0.08058431625366211
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,16,1,power_law_1.2,0.10798975944519043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,16,1,power_law_1.2,0.09120575904846191
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,16,1,power_law_1.2,0.1245574378967285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,16,1,power_law_1.2,0.15560640335083006
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,16,1,power_law_1.2,0.16649408340454103
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,16,1,power_law_1.2,0.17180479049682618
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,16,1,power_law_1.2,0.17782079696655273
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,16,1,power_law_1.2,0.18500160217285155
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,16,1,power_law_1.2,0.19053632736206055
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,16,1,power_law_1.2,0.21650367736816406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,16,1,power_law_1.2,0.22767616271972657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,16,1,power_law_1.2,0.2518111991882324
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,16,1,power_law_1.2,0.20450304031372069
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,16,1,power_law_1.01,2.2265248107910156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,16,1,power_law_1.01,2.9462030029296877
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,16,1,power_law_1.2,0.30704511642456056
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,16,1,power_law_1.2,0.3614451217651367
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,16,1,power_law_1.2,0.47388160705566407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,16,1,power_law_1.2,0.6061139297485352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,16,balanced,0.020188159942626953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,16,balanced,0.020200960636138916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,16,1,power_law_1.2,0.16157888412475585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,16,balanced,0.02029695987701416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,16,balanced,0.02821631908416748
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,16,1,power_law_1.2,0.8641600036621094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,16,balanced,0.03678976058959961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,16,balanced,0.04417535781860352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,16,balanced,0.04403456211090088
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,16,balanced,0.04539519786834717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,16,balanced,0.049443840980529785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,16,balanced,0.05048448085784912
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,16,balanced,0.054740481376647956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,16,balanced,0.05395071983337403
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,16,balanced,0.05079808235168457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,16,balanced,0.05558144092559815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,16,balanced,0.05607808113098145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,16,balanced,0.06091135978698731
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,16,balanced,0.0727513599395752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,16,balanced,0.07971199989318847
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,16,balanced,0.09737471580505372
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,16,balanced,0.11024640083312989
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,16,balanced,0.143056640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,16,balanced,0.17722240447998047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,16,balanced,0.25844608306884764
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,16,balanced,0.328271369934082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,16,balanced,0.46966400146484377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,16,balanced,0.023957760334014894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,16,balanced,0.6153587341308594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,16,balanced,0.024005119800567624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,16,balanced,0.024691200256347655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,16,balanced,0.025296640396118165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,16,balanced,0.026580479145050052
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,16,balanced,0.03134207963943482
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,16,balanced,0.03287935972213745
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,16,balanced,0.03254015922546387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,16,balanced,0.03290240049362182
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,16,balanced,0.03347071886062622
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,16,balanced,0.034206719398498536
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,16,balanced,0.03587584018707275
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,16,balanced,0.03673856019973755
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,16,balanced,0.032545280456542966
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,16,balanced,0.03474303960800171
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,16,balanced,0.036481280326843266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,16,balanced,0.03805567979812622
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,16,balanced,0.04153600215911865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,16,balanced,0.04904191970825195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,16,balanced,0.05738368034362793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,16,balanced,0.06790143966674804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,16,balanced,0.09389183998107911
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,16,balanced,0.11151488304138184
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,16,balanced,0.15933695793151856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,16,balanced,0.20723712921142576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,16,balanced,0.2921868705749512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,16,balanced,0.3872857666015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,16,1,power_law_1.2,1.1560198211669923
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,16,1,power_law_1.2,1.7609829711914062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,16,balanced,0.021434879302978514
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,1,16,balanced,0.016191999912261962
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,1,16,balanced,0.01608191967010498
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,1,16,balanced,0.01698815941810608
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,1,16,balanced,0.0170905601978302
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,1,16,balanced,0.01696768045425415
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,1,16,balanced,0.017300479412078858
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,1,16,balanced,0.016979199647903443
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,1,16,balanced,0.01737344026565552
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,1,16,balanced,0.017359360456466674
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,1,16,balanced,0.0174233603477478
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,1,16,balanced,0.017351679801940918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,1,16,balanced,0.01728767991065979
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,1,16,balanced,0.017388800382614138
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,1,16,balanced,0.018237440586090087
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,1,16,balanced,0.017422080039978027
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,1,16,balanced,0.018145279884338377
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,1,16,balanced,0.01871232032775879
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,1,16,balanced,0.019041279554367067
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,16,1,power_law_1.2,2.4350994873046874
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,1,16,balanced,0.01620736002922058
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,1,16,balanced,0.01660159945487976
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,1,16,balanced,0.016197119951248166
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,1,16,balanced,0.02519295930862427
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,1,16,balanced,0.025576961040496827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,16,power_law_1.01,0.02671231985092163
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,16,power_law_1.01,0.02805248022079468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,16,power_law_1.01,0.027969279289245606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,16,power_law_1.01,0.029196801185607912
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,16,power_law_1.01,0.04100863933563233
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,16,power_law_1.01,0.04739071846008301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,16,power_law_1.01,0.051891198158264165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,16,power_law_1.01,0.050220799446105954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,16,power_law_1.01,0.053013758659362795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,16,power_law_1.01,0.053695998191833495
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,16,power_law_1.01,0.05661568164825439
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,16,power_law_1.01,0.05861504077911377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,16,power_law_1.01,0.058525438308715816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,16,power_law_1.01,0.061951999664306645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,16,power_law_1.01,0.06336639881134033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,16,power_law_1.01,0.0669478416442871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,16,power_law_1.01,0.07521152019500732
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,16,power_law_1.01,0.08664575576782227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,16,power_law_1.01,0.10032896041870117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,16,power_law_1.01,0.1284774398803711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,16,power_law_1.01,0.1601625633239746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,16,power_law_1.01,0.22961408615112305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,16,power_law_1.01,0.2782643127441406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,16,power_law_1.01,0.4149068832397461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,1,16,balanced,0.031343359947204594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,16,power_law_1.01,0.5652326583862305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,16,power_law_1.01,0.025212159156799318
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,16,power_law_1.01,0.8456396484375001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,16,power_law_1.01,0.025687038898468018
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,16,power_law_1.01,0.025599999427795412
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,16,power_law_1.01,0.026073598861694337
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,16,power_law_1.01,0.030855679512023927
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,16,power_law_1.01,0.03134464025497437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,1,16,balanced,0.03838592052459717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,16,power_law_1.01,1.1911808013916017
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,16,power_law_1.01,0.0327948808670044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,16,power_law_1.01,0.0349452805519104
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,16,power_law_1.01,0.03591808080673218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,16,power_law_1.01,0.03305088043212891
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,1,16,balanced,0.05013120174407959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,16,power_law_1.01,0.03589504003524781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,16,power_law_1.01,0.034999039173126224
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,16,power_law_1.01,0.03655424118041992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,16,power_law_1.01,0.04001023769378662
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,16,power_law_1.01,0.04367616176605225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,16,power_law_1.01,0.051774721145629886
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,16,power_law_1.01,0.06196095943450928
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,16,power_law_1.01,0.08187264442443848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,16,power_law_1.01,0.10355711936950684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,16,power_law_1.01,0.037447679042816165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,16,power_law_1.01,0.12868479728698728
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,16,power_law_1.01,0.03294719934463501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,16,power_law_1.01,0.1723404884338379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,16,power_law_1.01,0.2355392074584961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,16,power_law_1.01,0.2861734390258789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,1,16,power_law_1.01,0.01729024052619934
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,1,16,power_law_1.01,0.017008639574050903
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,1,16,power_law_1.01,0.017181440591812133
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,1,16,power_law_1.01,0.016974079608917236
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,1,16,power_law_1.01,0.01661695957183838
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,16,power_law_1.01,0.5035084915161133
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,1,16,power_law_1.01,0.017326079607009888
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,16,power_law_1.01,0.656126708984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,1,16,power_law_1.01,0.01730944037437439
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,1,16,power_law_1.01,0.017132799625396728
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,1,16,power_law_1.01,0.01732223987579346
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,1,16,power_law_1.01,0.017411839962005616
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,1,16,power_law_1.01,0.01740928053855896
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,1,16,power_law_1.01,0.01747712016105652
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,1,16,power_law_1.01,0.017757439613342287
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,1,16,power_law_1.01,0.017831679582595825
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,1,16,power_law_1.01,0.01951616048812866
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,1,16,balanced,0.060172801017761235
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,1,16,power_law_1.01,0.0344320011138916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,1,16,power_law_1.01,0.020441598892211914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,1,16,power_law_1.01,0.017420799732208253
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,1,16,power_law_1.01,0.022744319438934325
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,1,16,power_law_1.01,0.02518656015396118
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,1,16,power_law_1.01,0.04286335945129395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,1,16,power_law_1.01,0.028536319732666016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,16,power_law_1.2,0.026429440975189213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,16,power_law_1.2,0.026004478931427
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,16,power_law_1.2,0.026819839477539065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,16,power_law_1.2,0.029024000167846682
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,16,power_law_1.2,0.03843967914581299
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,16,power_law_1.2,0.043683838844299314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,16,power_law_1.2,0.04829696178436279
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,16,power_law_1.2,0.05101823806762695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,16,power_law_1.2,0.05124095916748047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,16,power_law_1.2,0.05429376125335693
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,16,power_law_1.2,0.057140479087829585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,16,power_law_1.2,0.05738368034362793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,16,power_law_1.2,0.059924478530883785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,16,power_law_1.2,0.06139008045196533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,16,power_law_1.2,0.06578815937042236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,16,power_law_1.2,0.06959104061126709
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,16,power_law_1.2,0.07594367980957031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,16,power_law_1.2,0.09542016029357911
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,16,power_law_1.2,0.10527104377746581
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,16,power_law_1.2,0.13905664443969729
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,16,power_law_1.2,0.20109567642211915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,16,power_law_1.2,0.26082176208496094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,16,power_law_1.2,0.3540300750732422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,16,power_law_1.2,0.5000806427001953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,16,power_law_1.2,0.6739263916015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,1,16,power_law_1.01,0.05057151794433594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,1,16,power_law_1.01,0.07006336212158203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,1,16,power_law_1.01,0.07926911830902099
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,1,16,power_law_1.01,0.10173952102661132
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,16,power_law_1.2,0.025969920158386228
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,16,power_law_1.2,0.026163198947906495
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,1,16,power_law_1.01,0.1342732810974121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,16,power_law_1.2,0.025585920810699464
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,16,power_law_1.2,0.03128576040267944
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,16,power_law_1.2,0.032951040267944334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,16,power_law_1.2,0.03369856119155884
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,16,power_law_1.2,0.03425535917282104
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,16,power_law_1.2,0.03702912092208862
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,16,power_law_1.2,0.03255424022674561
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,16,power_law_1.2,0.03699199914932251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,16,power_law_1.2,0.031750400066375736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,16,power_law_1.2,0.03395456075668335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,16,power_law_1.2,0.036231679916381834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,16,power_law_1.2,0.038348801136016845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,16,power_law_1.2,0.040783357620239255
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,16,power_law_1.2,0.04554751873016357
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,16,power_law_1.2,0.05713151931762696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,16,power_law_1.2,0.06733183860778809
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,16,power_law_1.2,0.08453760147094727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,16,power_law_1.2,0.10959360122680664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,16,power_law_1.2,0.1653785514831543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,16,power_law_1.2,1.3215437316894532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,16,power_law_1.2,0.21518335342407227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,16,power_law_1.2,0.9876927947998047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,1,16,power_law_1.2,0.01710976004600525
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,16,power_law_1.2,0.024788479804992675
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,1,16,power_law_1.2,0.016970239877700806
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,1,16,power_law_1.2,0.016976640224456788
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,1,16,power_law_1.2,0.016944639682769776
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,1,16,power_law_1.2,0.016564480066299438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,1,16,power_law_1.2,0.01733888030052185
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,1,16,power_law_1.2,0.01738368034362793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,1,16,power_law_1.2,0.017007360458374025
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,1,16,power_law_1.2,0.017036800384521482
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,1,16,power_law_1.2,0.01740288019180298
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,1,16,power_law_1.2,0.017460479736328124
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,1,16,power_law_1.2,0.017496320009231566
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,1,16,power_law_1.2,0.017456640005111695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,1,16,power_law_1.2,0.018214399814605715
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,1,16,power_law_1.2,0.01993600010871887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,16,power_law_1.2,0.4024550247192383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,16,power_law_1.2,0.5714636611938476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,16,power_law_1.2,0.30029184341430665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,16,power_law_1.2,0.7245081329345704
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,1,16,power_law_1.2,0.019925760030746458
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,1,16,power_law_1.2,0.021283841133117674
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,1,16,power_law_1.2,0.023965439796447753
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,1,16,power_law_1.2,0.026659839153289795
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,1,16,power_law_1.2,0.03319295883178711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,8,balanced,0.018627840280532836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,8,balanced,0.019842560291290286
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,8,balanced,0.02066431999206543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,8,balanced,0.022781438827514648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,8,balanced,0.02846719980239868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,8,balanced,0.04286464214324951
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,8,balanced,0.04459648132324219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,8,balanced,0.04661119937896728
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,8,balanced,0.047791361808776855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,8,balanced,0.0474726390838623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,8,balanced,0.05022719860076904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,8,balanced,0.049603838920593266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,8,balanced,0.055901441574096676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,8,balanced,0.05803008079528808
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,8,balanced,0.06206592082977295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,8,balanced,0.06472064018249511
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,8,balanced,0.07006847858428955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,8,balanced,0.08566783905029297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,1,16,power_law_1.2,0.03988735914230347
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,8,balanced,0.09708288192749023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,8,balanced,0.12643199920654297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,8,balanced,0.15105024337768555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,8,balanced,0.20366592407226564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,8,balanced,0.25746175765991214
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,8,balanced,0.37757823944091795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,8,balanced,0.4881548690795898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,8,balanced,0.7093875122070312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,8,balanced,0.02310143947601318
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,8,balanced,0.02389888048171997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,8,balanced,0.9329881286621093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,8,balanced,0.02461951971054077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,8,balanced,0.024842240810394288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,8,balanced,0.02807552099227905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,8,balanced,0.03273216009140015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,8,balanced,0.03441920042037964
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,8,balanced,0.03608704090118408
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,8,balanced,0.03704960107803344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,8,balanced,0.03836416006088257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,8,balanced,0.03374847888946533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,8,balanced,0.03581311941146851
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,8,balanced,0.03747584104537964
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,8,balanced,0.0389516806602478
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,8,balanced,0.039509758949279786
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,8,balanced,0.0403276777267456
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,8,balanced,0.04358016014099121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,8,balanced,0.05545728206634522
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,8,balanced,0.06380928039550782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,8,balanced,0.08436736106872558
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,8,balanced,0.0991756820678711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,8,balanced,0.13951359748840333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,8,balanced,0.17434112548828123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,8,balanced,0.24875263214111326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,8,balanced,0.33263870239257814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,8,balanced,0.47767295837402346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,8,balanced,0.6236633682250977
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,1,16,power_law_1.2,0.052334079742431636
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,1,16,power_law_1.2,0.06342656135559081
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,1,16,power_law_1.2,0.10709376335144043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,1,16,power_law_1.2,0.151528959274292
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,1,16,power_law_1.2,0.09561087608337401
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,2,8,balanced,0.017386239767074586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,2,8,balanced,0.020994560718536375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,2,8,balanced,0.021091840267181396
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,2,8,balanced,0.021103360652923585
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,2,8,balanced,0.017822719812393188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,2,8,balanced,0.02145535945892334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,2,8,balanced,0.02108288049697876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,2,8,balanced,0.0212608003616333
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,2,8,balanced,0.021442561149597167
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,2,8,balanced,0.02148224115371704
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,2,8,balanced,0.021320960521697997
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,2,8,balanced,0.02147327899932861
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,2,8,balanced,0.02147327899932861
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,2,8,balanced,0.021914880275726318
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,2,8,balanced,0.02313472032546997
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,2,8,balanced,0.01821568012237549
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,2,8,balanced,0.023572480678558348
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,2,8,balanced,0.019429119825363158
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,2,8,balanced,0.021112320423126222
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,2,8,balanced,0.0210752010345459
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,1,16,power_law_1.2,0.22977151870727539
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,2,8,balanced,0.020207359790802
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,8,power_law_1.01,0.021125121116638185
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,8,power_law_1.01,0.024597759246826174
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,2,8,balanced,0.0336627197265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,2,8,balanced,0.032257280349731444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,8,power_law_1.01,0.029331200122833252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,8,power_law_1.01,0.0313369607925415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,8,power_law_1.01,0.04549248218536377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,8,power_law_1.01,0.047362561225891116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,8,power_law_1.01,0.04963456153869629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,8,power_law_1.01,0.04918399810791015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,8,power_law_1.01,0.05841536045074462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,8,power_law_1.01,0.06297215938568115
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,8,power_law_1.01,0.05561855792999267
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,8,power_law_1.01,0.05729407787322997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,8,power_law_1.01,0.06073855876922607
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,8,power_law_1.01,0.06383999824523925
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,2,8,balanced,0.04356863975524902
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,8,power_law_1.01,0.08387328147888183
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,8,power_law_1.01,0.09785856246948242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,8,power_law_1.01,0.12445952415466308
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,8,power_law_1.01,0.15369215965270996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,2,8,balanced,0.05268095970153809
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,8,power_law_1.01,0.22263040542602539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,8,power_law_1.01,0.05565567970275879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,8,power_law_1.01,0.2699839973449707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,8,power_law_1.01,0.0703001594543457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,8,power_law_1.01,0.3848025512695313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,8,power_law_1.01,0.48294528961181643
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,8,power_law_1.01,0.02608383893966675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,8,power_law_1.01,0.024467198848724364
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,2,8,balanced,0.06842368125915528
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,8,power_law_1.01,0.02492288112640381
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,8,power_law_1.01,0.7608576202392578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,8,power_law_1.01,0.02683903932571411
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,8,power_law_1.01,0.028122880458831788
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,8,power_law_1.01,0.032700159549713136
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,8,power_law_1.01,0.9689011383056642
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,8,power_law_1.01,0.03572479963302612
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,8,power_law_1.01,0.03464960098266602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,8,power_law_1.01,0.03662591934204101
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,8,power_law_1.01,0.03714816093444824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,8,power_law_1.01,0.03837055921554565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,8,power_law_1.01,0.0378278398513794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,8,power_law_1.01,0.03950592041015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,8,power_law_1.01,0.03715327978134155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,8,power_law_1.01,0.04076672077178955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,8,power_law_1.01,0.04357120037078858
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,8,power_law_1.01,0.056423678398132324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,8,power_law_1.01,0.06177919864654541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,8,power_law_1.01,0.09048959732055664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,8,power_law_1.01,0.07875072002410889
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,8,power_law_1.01,0.16777599334716797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,8,power_law_1.01,0.13738240242004393
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,8,power_law_1.01,0.245794563293457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,8,power_law_1.01,0.2970342445373535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,8,power_law_1.01,0.0319270396232605
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,2,8,power_law_1.01,0.018983680009841918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,2,8,power_law_1.01,0.019827200174331667
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,8,power_law_1.01,0.4673715209960937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,2,8,power_law_1.01,0.020677120685577394
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,2,8,power_law_1.01,0.020579841136932373
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,2,8,power_law_1.01,0.020638720989227293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,2,8,power_law_1.01,0.020471038818359374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,8,power_law_1.01,0.6050995254516602
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,2,8,power_law_1.01,0.021038079261779787
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,2,8,power_law_1.01,0.020665600299835205
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,2,8,power_law_1.01,0.020665600299835205
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,2,8,power_law_1.01,0.020701439380645753
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,2,8,power_law_1.01,0.021012480258941653
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,2,8,power_law_1.01,0.020660479068756104
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,2,8,power_law_1.01,0.021118719577789307
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,2,8,power_law_1.01,0.021451520919799804
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,2,8,power_law_1.01,0.02381824016571045
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,2,8,balanced,0.08415616035461426
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,2,8,power_law_1.01,0.03917056083679199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,8,power_law_1.01,0.025520639419555662
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,2,8,power_law_1.01,0.022731521129608155
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,2,8,power_law_1.01,0.02082175970077515
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,2,8,power_law_1.01,0.02848639965057373
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,2,8,power_law_1.01,0.03508991956710815
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,2,8,power_law_1.01,0.058853120803833
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,8,power_law_1.2,0.021040639877319335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,8,power_law_1.2,0.023319039344787598
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,2,8,power_law_1.01,0.08295552253723144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,8,power_law_1.2,0.024542720317840577
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,8,power_law_1.2,0.028454399108886717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,8,power_law_1.2,0.044431362152099606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,8,power_law_1.2,0.045503997802734376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,8,power_law_1.2,0.0541593599319458
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,2,8,power_law_1.01,0.025323519706726073
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,2,8,power_law_1.01,0.051454720497131345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,8,power_law_1.2,0.05661056041717529
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,8,power_law_1.2,0.05669375896453858
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,8,power_law_1.2,0.06222335815429687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,8,power_law_1.2,0.059147520065307616
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,8,power_law_1.2,0.059084801673889166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,8,power_law_1.2,0.06179711818695068
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,8,power_law_1.2,0.06393343925476073
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,8,power_law_1.2,0.07383808135986328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,8,power_law_1.2,0.08797823905944824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,8,power_law_1.2,0.10120960235595704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,8,power_law_1.2,0.049708800315856935
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,8,power_law_1.2,0.12805120468139647
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,8,power_law_1.2,0.16391807556152344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,8,power_law_1.2,0.22702592849731446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,8,power_law_1.2,0.4092083358764649
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,8,power_law_1.2,0.27940479278564456
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,8,power_law_1.2,0.024369919300079347
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,8,power_law_1.2,0.025148160457611084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,8,power_law_1.2,0.028857600688934327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,8,power_law_1.2,0.026380798816680907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,8,power_law_1.2,0.0263590407371521
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,8,power_law_1.2,0.5596620941162109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,8,power_law_1.2,0.02788991928100586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,8,power_law_1.2,0.03296895980834961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,8,power_law_1.2,0.03393151998519898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,8,power_law_1.2,0.03206912040710449
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,8,power_law_1.2,0.03500927925109863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,8,power_law_1.2,0.03623552083969116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,8,power_law_1.2,0.0392793607711792
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,8,power_law_1.2,0.036863999366760256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,8,power_law_1.2,0.03578111886978149
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,8,power_law_1.2,0.038689279556274415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,8,power_law_1.2,0.03996927976608276
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,8,power_law_1.2,0.04486527919769287
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,8,power_law_1.2,0.05769216060638428
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,8,power_law_1.2,0.06499968051910401
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,8,power_law_1.2,0.08454527854919433
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,8,power_law_1.2,0.10359295845031738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,8,power_law_1.2,0.14957183837890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,8,power_law_1.2,0.04108672142028809
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,8,power_law_1.2,0.18494976043701172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,8,power_law_1.2,0.8308415985107421
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,8,power_law_1.2,0.3449612808227539
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,2,8,power_law_1.2,0.01903488039970398
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,2,8,power_law_1.2,0.019028480052948
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,2,8,power_law_1.2,0.020391678810119628
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,2,8,power_law_1.2,0.02067199945449829
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,2,8,power_law_1.2,0.02062079906463623
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,2,8,power_law_1.2,0.02061311960220337
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,2,8,power_law_1.2,0.020675840377807616
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,2,8,power_law_1.2,0.020651519298553467
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,2,8,power_law_1.2,0.02098304033279419
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,2,8,power_law_1.2,0.020663039684295652
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,2,8,power_law_1.2,0.021116158962249755
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,2,8,power_law_1.01,0.09963264465332031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,2,8,power_law_1.2,0.021072640419006347
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,2,8,power_law_1.2,0.02110208034515381
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,2,8,power_law_1.2,0.021893119812011717
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,2,8,power_law_1.2,0.02271104097366333
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,2,8,power_law_1.2,0.02311295986175537
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,2,8,power_law_1.01,0.21672447204589843
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,2,8,power_law_1.01,0.15261440277099608
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,8,power_law_1.2,0.2778137588500976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,8,power_law_1.2,0.5199244689941407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,2,8,power_law_1.2,0.024343039989471436
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,2,8,power_law_1.2,0.027307519912719725
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,2,8,power_law_1.2,0.03323519945144653
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,2,8,power_law_1.2,0.04093440055847168
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,2,8,power_law_1.2,0.048419837951660154
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,2,8,power_law_1.2,0.06502912044525147
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,2,8,power_law_1.2,0.07560448169708252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,4,4,balanced,0.01985919952392578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,4,balanced,0.021488640308380127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,4,balanced,0.024330239295959472
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,4,balanced,0.03260416030883789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,4,balanced,0.032152318954467775
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,4,balanced,0.05395455837249756
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,4,balanced,0.05327871799468994
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,4,balanced,0.05572864055633545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,4,balanced,0.0573145580291748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,4,balanced,0.0535756778717041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,4,balanced,0.054091520309448235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,4,balanced,0.05673984050750732
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,4,balanced,0.05787136077880859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,4,4,balanced,0.021038079261779787
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,4,balanced,0.06047743797302246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,4,balanced,0.06357120037078857
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,4,balanced,0.06898431777954102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,4,balanced,0.08132351875305176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,4,balanced,0.09288319587707519
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,4,balanced,0.11928319931030273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,4,balanced,0.14140288352966307
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,4,balanced,0.18925439834594726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,4,balanced,0.23751039505004884
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,4,balanced,0.3458406448364258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,2,8,power_law_1.2,0.11185791969299316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,4,balanced,0.45047679901123044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,4,balanced,0.024357120990753173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,4,balanced,0.024792320728302
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,4,balanced,0.6643315124511718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,4,balanced,0.0255295991897583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,4,balanced,0.02840320110321045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,4,balanced,0.8656742095947265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,4,balanced,0.03258624076843262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,4,balanced,0.03419519901275635
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,4,balanced,0.039134719371795655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,4,balanced,0.040293121337890626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,4,balanced,0.042398719787597655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,4,balanced,0.04364031791687012
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,4,balanced,0.039110400676727296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,4,balanced,0.0403110408782959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,4,balanced,0.040816640853881835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,4,balanced,0.042339839935302735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,4,balanced,0.042699518203735354
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,4,balanced,0.04485504150390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,4,balanced,0.04850560188293457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,4,balanced,0.05633024215698242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,4,balanced,0.0654963207244873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,4,balanced,0.08448896408081055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,4,balanced,0.09672191619873047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,4,balanced,0.13882495880126952
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,2,8,power_law_1.2,0.14585856437683103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,4,balanced,0.1695142364501953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,4,balanced,0.2515519905090332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,4,balanced,0.3190975952148437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,8,power_law_1.2,1.0414733123779296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,4,balanced,0.4710553741455078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,4,balanced,0.616174087524414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,8,power_law_1.2,0.6541632080078126
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,2,8,power_law_1.2,0.21389184951782228
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,2,8,power_law_1.2,0.2983654403686523
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,4,4,balanced,0.020595200061798095
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,4,4,balanced,0.0251584005355835
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,4,4,balanced,0.02520960092544556
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,4,4,balanced,0.02519808053970337
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,4,4,balanced,0.023047680854797366
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,4,4,balanced,0.02228224039077759
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,4,4,balanced,0.02558336019515991
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,4,4,balanced,0.02099071979522705
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,4,4,balanced,0.025544960498809815
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,4,4,balanced,0.025564160346984867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,4,4,balanced,0.02551039934158325
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,4,4,balanced,0.025744640827178956
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,4,4,balanced,0.024936959743499756
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,4,4,balanced,0.025806078910827635
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,4,4,balanced,0.027627520561218262
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,4,4,balanced,0.027622399330139158
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,4,4,balanced,0.029594879150390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,4,4,balanced,0.03167232036590576
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,4,4,balanced,0.025185279846191407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,4,4,balanced,0.025223679542541504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,4,4,power_law_1.01,0.021456639766693115
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,4,4,power_law_1.01,0.022391040325164795
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,4,4,balanced,0.025166079998016355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,4,power_law_1.01,0.023162879943847657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,4,power_law_1.01,0.028691198825836185
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,4,power_law_1.01,0.031535360813140866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,4,power_law_1.01,0.03337599992752075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,4,power_law_1.01,0.04959616184234619
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,4,power_law_1.01,0.049363198280334475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,4,power_law_1.01,0.05333375930786133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,4,power_law_1.01,0.0554201602935791
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,4,power_law_1.01,0.060652799606323235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,4,power_law_1.01,0.05402880191802979
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,4,power_law_1.01,0.05638656139373779
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,4,power_law_1.01,0.06007423877716065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,4,power_law_1.01,0.06226687908172608
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,4,power_law_1.01,0.06457215785980225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,4,power_law_1.01,0.07177343845367432
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,4,power_law_1.01,0.08200063705444335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,4,power_law_1.01,0.0943769645690918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,4,4,balanced,0.06403840065002442
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,4,power_law_1.01,0.12
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,4,4,balanced,0.04607999801635742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,4,power_law_1.01,0.14485887527465818
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,4,4,balanced,0.04936575889587402
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,4,power_law_1.01,0.1968921661376953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,4,4,balanced,0.09943679809570313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,4,power_law_1.01,0.25910144805908203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,4,4,balanced,0.07595647811889648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,4,power_law_1.01,0.36866432189941406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,4,power_law_1.01,0.026168320178985592
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,4,power_law_1.01,0.48881408691406253
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,4,power_law_1.01,0.02676352024078369
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,4,power_law_1.01,0.029926400184631347
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,4,power_law_1.01,0.6956416320800781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,4,power_law_1.01,0.03047039985656738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,4,power_law_1.01,0.03863552093505859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,4,power_law_1.01,0.8916915130615234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,4,power_law_1.01,0.04400512218475342
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,4,power_law_1.01,0.043308801651000976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,4,power_law_1.01,0.0405350399017334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,4,power_law_1.01,0.0431001615524292
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,4,power_law_1.01,0.04566271781921387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,4,power_law_1.01,0.04953343868255615
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,4,power_law_1.01,0.05876095771789551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,4,power_law_1.01,0.06850048065185546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,4,power_law_1.01,0.0871884822845459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,4,power_law_1.01,0.1047001552581787
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,4,power_law_1.01,0.15187711715698243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,4,power_law_1.01,0.18616704940795897
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,4,power_law_1.01,0.2609868812561035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,4,power_law_1.01,0.03355007886886597
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,4,power_law_1.01,0.34477825164794923
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,4,power_law_1.01,0.49960063934326177
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,4,4,power_law_1.01,0.0207807993888855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,4,power_law_1.01,0.6501811218261719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,4,power_law_1.01,0.04106624126434326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,4,power_law_1.01,0.04032127857208252
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,4,4,power_law_1.01,0.022274560928344726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,4,power_law_1.01,0.0402188777923584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,4,4,power_law_1.01,0.022666239738464357
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,4,4,power_law_1.01,0.023503360748291017
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,4,4,power_law_1.01,0.023778560161590575
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,4,4,power_law_1.01,0.024426240921020508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,4,power_law_1.01,0.042017278671264646
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,4,4,power_law_1.01,0.025012478828430173
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,4,4,power_law_1.01,0.025157120227813724
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,4,4,power_law_1.01,0.02482304096221924
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,4,4,power_law_1.01,0.025163519382476806
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,4,4,power_law_1.01,0.025437440872192386
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,4,4,power_law_1.01,0.02556544065475464
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,4,4,power_law_1.01,0.025927679538726805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,4,4,power_law_1.01,0.02636672019958496
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,4,4,power_law_1.01,0.04046592235565186
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,4,4,power_law_1.01,0.025356800556182862
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,4,4,balanced,0.12673279762268067
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,4,power_law_1.01,0.025624320507049558
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,4,4,power_law_1.01,0.03053567886352539
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,4,4,power_law_1.01,0.028866560459136964
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,4,4,power_law_1.01,0.048313598632812496
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,4,4,power_law_1.01,0.03738111972808838
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,4,4,power_law_1.01,0.056720638275146486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,4,4,power_law_1.2,0.020610558986663818
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,4,4,power_law_1.2,0.021006081104278564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,4,power_law_1.2,0.023161599636077883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,4,power_law_1.2,0.02934528112411499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,4,power_law_1.2,0.03105792045593262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,4,power_law_1.2,0.03641472101211548
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,4,4,power_law_1.01,0.07276927947998046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,4,power_law_1.2,0.04630784034729004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,4,power_law_1.2,0.053614082336425783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,4,power_law_1.2,0.049617919921874996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,4,power_law_1.2,0.05836544036865234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,4,power_law_1.2,0.055082240104675294
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,4,power_law_1.2,0.05240064144134522
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,4,power_law_1.2,0.057310719490051266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,4,power_law_1.2,0.05949567794799805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,4,power_law_1.2,0.06280064105987548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,4,power_law_1.2,0.07324031829833985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,4,power_law_1.2,0.0662886381149292
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,4,4,power_law_1.01,0.12854144096374512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,4,power_law_1.2,0.08484352111816407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,4,power_law_1.2,0.09677951812744141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,4,power_law_1.2,0.1220479965209961
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,4,4,power_law_1.01,0.09156224250793457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,4,power_law_1.2,0.14748543739318848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,4,power_law_1.2,0.1984422492980957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,4,power_law_1.2,0.36648830413818356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,4,power_law_1.2,0.2570047950744629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,4,power_law_1.2,0.025935359001159668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,4,power_law_1.2,0.02683135986328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,4,power_law_1.2,0.029707520008087157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,4,power_law_1.2,0.030830080509185794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,4,power_law_1.2,0.024825599193572998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,4,power_law_1.2,0.03917056083679199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,4,power_law_1.2,0.03995392084121704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,4,power_law_1.2,0.041150717735290526
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,4,power_law_1.2,0.04274303913116455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,4,power_law_1.2,0.0431987190246582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,4,power_law_1.2,0.03994879961013794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,4,power_law_1.2,0.04087552070617676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,4,power_law_1.2,0.04231296062469482
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,4,power_law_1.2,0.0435916805267334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,4,power_law_1.2,0.046094079017639164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,4,power_law_1.2,0.7208767700195312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,4,power_law_1.2,0.03380991935729981
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,4,power_law_1.2,0.04995584011077881
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,4,power_law_1.2,0.05960959911346435
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,4,power_law_1.2,0.06875008106231689
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,4,power_law_1.2,0.08907135963439941
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,4,power_law_1.2,0.10734848022460937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,4,power_law_1.2,0.1533785629272461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,4,power_law_1.2,0.1892736053466797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,4,power_law_1.2,0.4795366287231445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,4,power_law_1.2,0.3535103988647461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,4,4,power_law_1.2,0.02082047939300537
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,4,4,power_law_1.2,0.021488640308380127
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,4,4,power_law_1.01,0.17315967559814455
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,4,4,power_law_1.01,0.284956169128418
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,4,4,power_law_1.2,0.022406399250030518
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,4,4,power_law_1.2,0.022739200592041015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,4,4,power_law_1.2,0.023359999656677247
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,4,4,power_law_1.2,0.02487936019897461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,4,4,power_law_1.2,0.025080320835113523
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,4,4,power_law_1.2,0.025204479694366455
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,4,4,power_law_1.2,0.024796159267425538
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,4,4,power_law_1.2,0.0247705602645874
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,4,4,power_law_1.2,0.02513535976409912
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,4,4,power_law_1.2,0.02553855895996094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,4,4,power_law_1.2,0.025582079887390134
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,4,4,power_law_1.2,0.02606591939926147
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,4,4,power_law_1.2,0.030186240673065183
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,4,4,power_law_1.01,0.24083200454711914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,4,power_law_1.2,0.27538944244384767
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,4,power_law_1.2,0.5137625503540039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,4,4,power_law_1.2,0.03383039951324463
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,4,4,power_law_1.2,0.04199935913085938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,4,4,power_law_1.2,0.029475839138031007
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,4,4,power_law_1.2,0.043957757949829104
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,4,4,power_law_1.2,0.05851647853851318
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,4,4,power_law_1.2,0.06815743923187256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,4,power_law_1.2,0.9359833526611329
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,16,balanced,0.028476159572601318
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,16,balanced,0.02813184022903442
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,16,balanced,0.030975999832153323
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,16,balanced,0.05853824138641357
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,16,balanced,0.07348351955413819
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,16,balanced,0.1096883201599121
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,4,4,power_law_1.2,0.09658495903015137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,16,balanced,0.11236351966857909
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,16,balanced,0.11432959556579589
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,4,4,power_law_1.2,0.12567296028137206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,16,balanced,0.11794560432434083
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,16,balanced,0.12158847808837892
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,16,balanced,0.1256588840484619
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,16,balanced,0.1306060791015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,16,balanced,0.12103679656982422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,16,balanced,0.12722047805786132
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,16,balanced,0.12958463668823242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,16,balanced,0.13814911842346192
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,16,balanced,0.15112704277038574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,16,balanced,0.11621376037597657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,16,balanced,0.16188159942626953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,16,balanced,0.21947263717651366
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,16,balanced,0.23249664306640624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,16,balanced,0.33417217254638676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,16,balanced,0.40395648956298824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,16,balanced,0.5645888137817383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,16,balanced,0.7202239990234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,16,balanced,1.0492774200439452
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,16,balanced,0.030150399208068845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,16,balanced,0.030995199680328368
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,16,balanced,1.3750732421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,16,balanced,0.031709439754486084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,16,balanced,0.04996736049652099
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,16,balanced,0.05694975852966309
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,16,balanced,0.059031038284301764
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,16,balanced,0.06059264183044434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,16,balanced,0.06108416080474853
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,16,balanced,0.03196799993515015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,16,balanced,0.054849281311035156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,16,balanced,0.0633996820449829
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,4,4,power_law_1.2,0.17495935440063476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,16,balanced,0.06679423809051513
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,16,balanced,0.07136640071868897
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,16,balanced,0.062471680641174324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,16,balanced,0.06525184154510498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,16,balanced,0.0671129608154297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,16,balanced,0.07428351879119874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,16,balanced,0.08001279830932617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,16,balanced,0.08743295669555665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,16,balanced,0.10569472312927246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,16,balanced,0.12303232192993163
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,16,balanced,0.1766771125793457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,16,balanced,0.2089151954650879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,16,balanced,0.2868633651733398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,16,balanced,0.37166080474853513
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,16,balanced,0.526635513305664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,16,balanced,0.6924044799804687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,4,4,power_law_1.2,0.22347007751464845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,4,power_law_1.2,0.6696781158447266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,4,4,power_law_1.2,0.30387456893920894
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,1,16,balanced,0.019031039476394652
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,1,16,balanced,0.018615039587020873
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,1,16,balanced,0.033606400489807134
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,1,16,balanced,0.03158015966415405
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,1,16,balanced,0.033342719078063965
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,1,16,balanced,0.031978240013122564
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,1,16,balanced,0.03418495893478394
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,1,16,balanced,0.020701439380645753
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,1,16,balanced,0.03246975898742675
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,1,16,balanced,0.034146559238433835
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,1,16,balanced,0.034028799533843995
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,1,16,balanced,0.03398400068283081
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,1,16,balanced,0.03459455966949463
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,1,16,balanced,0.034835200309753414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,1,16,balanced,0.035732479095458985
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,1,16,balanced,0.03873791933059693
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,1,16,balanced,0.03818880081176758
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,1,16,balanced,0.019166719913482667
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,1,16,balanced,0.032231678962707525
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,1,16,balanced,0.032616961002349856
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,1,16,balanced,0.033011200428009035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,16,power_law_1.01,0.07366399765014649
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,16,power_law_1.01,0.06993279933929444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,16,power_law_1.01,0.07214719772338866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,16,power_law_1.01,0.07351808071136474
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,16,power_law_1.01,0.10143103599548339
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,16,power_law_1.01,0.10847488403320313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,16,power_law_1.01,0.11166336059570312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,16,power_law_1.01,0.11562111854553223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,16,power_law_1.01,0.1179097557067871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,16,power_law_1.01,0.1189081573486328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,4,4,power_law_1.2,0.4177395248413086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,16,power_law_1.01,0.12767871856689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,16,power_law_1.01,0.127642879486084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,16,power_law_1.01,0.13018112182617186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,16,power_law_1.01,0.12834943771362303
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,16,power_law_1.01,0.13290111541748045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,16,power_law_1.01,0.1395353603363037
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,16,power_law_1.01,0.15322367668151854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,16,power_law_1.01,0.18883840560913084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,16,power_law_1.01,0.2355200004577637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,16,power_law_1.01,0.3086771202087403
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,16,power_law_1.01,0.3753395080566406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,16,power_law_1.01,0.5266393661499024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,16,power_law_1.01,0.6912255859374999
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,16,power_law_1.01,1.0321510314941407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,16,power_law_1.01,1.2672537231445313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,16,power_law_1.01,2.057936706542969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,16,power_law_1.01,0.04903168201446533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,16,power_law_1.01,0.049237761497497555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,16,power_law_1.01,0.04991744041442871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,16,power_law_1.01,2.6586663818359377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,16,power_law_1.01,0.0477184009552002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,16,power_law_1.01,0.05081215858459472
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,16,power_law_1.01,0.055979518890380865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,16,power_law_1.01,0.05909120082855225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,16,power_law_1.01,0.06141823768615723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,16,power_law_1.01,0.06245376110076904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,16,power_law_1.01,0.06621312141418458
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,16,power_law_1.01,0.0735807991027832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,1,16,balanced,0.06506495952606202
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,16,power_law_1.01,0.06761343955993653
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,1,16,balanced,0.048853759765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,1,16,balanced,0.05154047966003418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,16,power_law_1.01,0.06744832038879395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,16,power_law_1.01,0.07245567798614502
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,16,power_law_1.01,0.075731201171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,16,power_law_1.01,0.09329664230346679
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,16,power_law_1.01,0.08237695693969727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,16,power_law_1.01,0.12076031684875488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,16,power_law_1.01,0.15813376426696776
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,16,power_law_1.01,0.180947208404541
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,1,16,balanced,0.10364416122436523
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,1,16,balanced,0.07786367893218994
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,16,power_law_1.01,0.3485043334960937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,16,power_law_1.01,0.05731584072113037
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,16,power_law_1.01,0.4633152008056641
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,1,16,power_law_1.01,0.03173120021820068
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,1,16,power_law_1.01,0.03123456001281738
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,1,16,balanced,0.13041279792785646
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,1,16,power_law_1.01,0.02839423894882202
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,1,16,power_law_1.01,0.024732160568237304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,16,power_law_1.01,0.9436595153808593
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,1,16,power_law_1.01,0.030054399967193602
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,1,16,power_law_1.01,0.031045119762420654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,16,power_law_1.01,0.25472768783569333
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,1,16,power_law_1.01,0.03059200048446655
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,1,16,power_law_1.01,0.03216639995574951
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,1,16,power_law_1.01,0.031552000045776366
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,1,16,power_law_1.01,0.03405951976776123
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,1,16,power_law_1.01,0.03390464067459106
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,16,power_law_1.01,1.2077145385742187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,1,16,power_law_1.01,0.03379456043243408
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,1,16,power_law_1.01,0.0361843204498291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,16,power_law_1.01,0.6153087997436524
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,1,16,power_law_1.01,0.03222143888473511
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,1,16,power_law_1.01,0.0324070405960083
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,1,16,power_law_1.01,0.03529599905014038
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,1,16,power_law_1.01,0.04118271827697754
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,1,16,power_law_1.01,0.03752703905105591
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,1,16,power_law_1.01,0.04923903942108154
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,1,16,power_law_1.01,0.05718656063079834
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,1,16,power_law_1.01,0.0629478406906128
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,16,power_law_1.2,0.07074560165405273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,16,power_law_1.2,0.07285759925842286
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,16,power_law_1.2,0.07234816074371338
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,16,power_law_1.2,0.07012224197387695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,16,power_law_1.2,0.10926336288452147
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,16,power_law_1.2,0.11222911834716798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,16,power_law_1.2,0.11527423858642578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,16,power_law_1.2,0.11908224105834961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,16,power_law_1.2,0.09880448341369628
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,16,power_law_1.2,0.12161919593811035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,16,power_law_1.2,0.12729215621948242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,1,16,power_law_1.01,0.08023296356201172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,16,power_law_1.2,0.1309875202178955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,16,power_law_1.2,0.12811264038085937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,16,power_law_1.2,0.13123583793640137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,16,power_law_1.2,0.13558527946472168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,16,power_law_1.2,0.14201727867126465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,16,power_law_1.2,0.15417856216430664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,16,power_law_1.2,0.22820480346679686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,16,power_law_1.2,0.26761728286743164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,16,power_law_1.2,0.3553472137451172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,16,power_law_1.2,0.45833473205566405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,16,power_law_1.2,0.6314419174194337
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,16,power_law_1.2,0.8137753295898438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,16,power_law_1.2,1.2396428680419922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,16,power_law_1.2,1.6304524230957032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,16,power_law_1.2,0.04848127841949463
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,16,power_law_1.2,0.04632192134857178
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,16,power_law_1.2,0.049052162170410155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,16,power_law_1.2,0.04479360103607178
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,16,power_law_1.2,0.046132478713989254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,16,power_law_1.2,0.05433087825775147
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,16,power_law_1.2,0.05774208068847656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,16,power_law_1.2,0.05958144187927246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,16,power_law_1.2,0.06105984210968017
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,16,power_law_1.2,0.06484352111816406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,16,power_law_1.2,0.06884096145629884
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,16,power_law_1.2,0.07097216129302979
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,16,power_law_1.2,0.06484608173370361
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,16,power_law_1.2,0.06955520153045655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,16,power_law_1.2,0.07658239841461181
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,16,power_law_1.2,0.0789132785797119
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,16,power_law_1.2,0.08172672271728516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,16,power_law_1.2,0.109617919921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,16,power_law_1.2,0.13060352325439453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,16,power_law_1.2,0.1762944030761719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,16,power_law_1.2,0.22819456100463867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,16,power_law_1.2,0.290567684173584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,16,power_law_1.2,0.35974399566650395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,1,16,power_law_1.01,0.10360704421997072
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,1,16,power_law_1.01,0.1518950366973877
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,1,16,power_law_1.01,0.2796684837341309
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,1,16,power_law_1.01,0.1773324775695801
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,1,16,power_law_1.2,0.031065599918365477
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,1,16,power_law_1.01,0.412993278503418
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,1,16,power_law_1.2,0.026298880577087402
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,1,16,power_law_1.2,0.03113343954086304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,16,power_law_1.2,0.557151985168457
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,1,16,power_law_1.2,0.028378880023956297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,1,16,power_law_1.2,0.024477438926696776
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,1,16,power_law_1.2,0.02980736017227173
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,1,16,power_law_1.2,0.030988800525665283
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,1,16,power_law_1.2,0.031591680049896244
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,1,16,power_law_1.2,0.03091455936431885
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,1,16,power_law_1.2,0.03222912073135376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,1,16,power_law_1.2,0.03089024066925049
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,1,16,power_law_1.2,0.03162879943847656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,1,16,power_law_1.2,0.03445888042449951
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,1,16,power_law_1.2,0.03500416040420532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,16,power_law_1.2,0.7493901062011719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,16,power_law_1.2,2.3227751159667966
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,16,power_law_1.2,1.106688003540039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,16,power_law_1.2,1.53939453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,1,16,power_law_1.2,0.0369868803024292
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,1,16,power_law_1.2,0.03756671905517578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,1,16,power_law_1.2,0.0384934401512146
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,1,16,power_law_1.2,0.045667839050292966
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,1,16,power_law_1.2,0.05312384128570556
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,16,power_law_1.2,3.319820861816406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,1,16,power_law_1.2,0.06172927856445313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,8,balanced,0.026371839046478274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,8,balanced,0.027180800437927245
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,8,balanced,0.04321663856506348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,8,balanced,0.055484161376953126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,8,balanced,0.07713280200958253
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,8,balanced,0.10950655937194824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,8,balanced,0.11347840309143067
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,8,balanced,0.11560704231262206
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,1,16,power_law_1.2,0.08018943786621094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,8,balanced,0.11926527976989747
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,8,balanced,0.12652799606323242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,8,balanced,0.11570048332214355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,8,balanced,0.12103936195373535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,8,balanced,0.12261376380920411
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,8,balanced,0.1290329647064209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,8,balanced,0.13320192337036132
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,8,balanced,0.13903871536254883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,8,balanced,0.14434816360473632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,8,balanced,0.17437055587768555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,8,balanced,0.19236352920532226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,8,balanced,0.24699392318725585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,8,balanced,0.29120256423950197
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,8,balanced,0.4269132614135742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,8,balanced,0.528611831665039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,8,balanced,0.7576051330566406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,8,balanced,0.9763251495361327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,8,balanced,1.4379327392578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,8,balanced,0.028400640487670898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,8,balanced,0.029648640155792237
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,8,balanced,0.03128448009490967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,8,balanced,0.03512063980102539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,8,balanced,1.8966604614257812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,8,balanced,0.05433472156524658
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,8,balanced,0.06374527931213378
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,8,balanced,0.06632319927215577
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,8,balanced,0.06937983989715577
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,8,balanced,0.07005951881408692
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,8,balanced,0.06318336009979249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,8,balanced,0.06755712032318115
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,8,balanced,0.06893311977386475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,8,balanced,0.06170495986938477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,8,balanced,0.07234687805175781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,8,balanced,0.07353343963623046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,8,balanced,0.07507840156555176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,8,balanced,0.08510592460632324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,8,balanced,0.09831423759460449
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,1,16,power_law_1.2,0.11552384376525879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,8,balanced,0.11298175811767579
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,8,balanced,0.14384896278381348
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,1,16,power_law_1.2,0.14560511589050293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,8,balanced,0.17023231506347655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,8,balanced,0.251856632232666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,8,balanced,0.30801408767700195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,8,balanced,0.42724864959716796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,8,balanced,0.5562611389160156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,8,balanced,0.8062105560302735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,8,balanced,1.0618547058105468
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,1,16,power_law_1.2,0.21670015335083007
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,1,16,power_law_1.2,0.25015680313110356
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,1,16,power_law_1.2,0.4043328094482422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,2,8,balanced,0.01911679983139038
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,2,8,balanced,0.03584511995315552
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,2,8,balanced,0.01982208013534546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,2,8,balanced,0.03614336013793945
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,2,8,balanced,0.020742399692535402
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,2,8,balanced,0.03699199914932251
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,2,8,balanced,0.036392960548400875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,2,8,balanced,0.036652801036834715
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,2,8,balanced,0.036221439838409426
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,2,8,balanced,0.0367142391204834
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,2,8,balanced,0.03695103883743286
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,2,8,balanced,0.038042879104614256
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,2,8,balanced,0.039426560401916506
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,2,8,balanced,0.02309760093688965
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,2,8,balanced,0.0403110408782959
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,2,8,balanced,0.03662208080291748
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,2,8,balanced,0.041994237899780275
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,2,8,balanced,0.044199681282043456
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,2,8,balanced,0.036373760700225834
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,2,8,balanced,0.03682687997817993
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,2,8,balanced,0.03695487976074219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,8,power_law_1.01,0.054746880531311035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,8,power_law_1.01,0.05844480037689209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,8,power_law_1.01,0.058940157890319825
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,8,power_law_1.01,0.06751359939575195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,8,power_law_1.01,0.06835968017578124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,8,power_law_1.01,0.10548224449157714
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,8,power_law_1.01,0.11003520011901854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,8,power_law_1.01,0.11108991622924805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,8,power_law_1.01,0.11471360206604005
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,8,power_law_1.01,0.12007167816162109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,8,power_law_1.01,0.12286335945129394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,8,power_law_1.01,0.11769344329833983
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,8,power_law_1.01,0.11586175918579103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,8,power_law_1.01,0.12375935554504394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,8,power_law_1.01,0.1289791965484619
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,8,power_law_1.01,0.13523584365844726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,8,power_law_1.01,0.1447539234161377
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,2,8,balanced,0.09471872329711914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,2,8,balanced,0.06068863868713379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,8,power_law_1.01,0.1787276840209961
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,2,8,balanced,0.07796224117279052
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,8,power_law_1.01,0.20006015777587888
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,8,power_law_1.01,0.2615014457702637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,8,power_law_1.01,0.33867008209228516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,2,8,balanced,0.1290329647064209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,8,power_law_1.01,0.47464958190917966
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,8,power_law_1.01,0.6227225494384766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,2,8,balanced,0.05776639938354492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,8,power_law_1.01,0.8579763031005859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,8,power_law_1.01,0.03179136037826538
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,8,power_law_1.01,0.04078464031219482
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,8,power_law_1.01,1.1111078643798828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,8,power_law_1.01,0.04664576053619385
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,8,power_law_1.01,0.04457727909088135
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,8,power_law_1.01,0.06081664085388183
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,8,power_law_1.01,0.05063807964324951
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,8,power_law_1.01,0.061145601272583006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,8,power_law_1.01,0.0640614414215088
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,8,power_law_1.01,0.06587007999420166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,8,power_law_1.01,0.0694105577468872
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,8,power_law_1.01,1.6328166198730467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,8,power_law_1.01,0.07302527904510499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,8,power_law_1.01,0.07000319957733155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,8,power_law_1.01,0.06555903911590576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,8,power_law_1.01,0.07006847858428955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,8,power_law_1.01,0.07250815868377686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,8,power_law_1.01,0.0751526403427124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,8,power_law_1.01,0.08217472076416016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,8,power_law_1.01,0.09829631805419922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,8,power_law_1.01,0.10969728469848632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,8,power_law_1.01,2.204302062988281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,8,power_law_1.01,0.1568575954437256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,8,power_law_1.01,0.18231552124023437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,8,power_law_1.01,0.2536038398742676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,8,power_law_1.01,0.3368652725219726
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,2,8,power_law_1.01,0.022389760017395018
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,8,power_law_1.01,0.45139713287353517
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,2,8,power_law_1.01,0.027292160987854003
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,8,power_law_1.01,0.6040089416503907
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,2,8,power_law_1.01,0.027562239170074464
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,2,8,power_law_1.01,0.02722048044204712
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,2,8,power_law_1.01,0.034533119201660155
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,2,8,power_law_1.01,0.03276160001754761
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,8,power_law_1.01,0.8265894317626954
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,2,8,power_law_1.01,0.03338752031326294
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,2,8,power_law_1.01,0.03374592065811157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,2,8,power_law_1.01,0.03450367927551269
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,2,8,power_law_1.01,0.0354803204536438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,2,8,power_law_1.01,0.03540735960006714
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,8,power_law_1.01,1.0471923065185547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,2,8,power_law_1.01,0.03666431903839111
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,2,8,power_law_1.01,0.03802752017974854
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,2,8,power_law_1.01,0.041948161125183105
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,2,8,power_law_1.01,0.0426367998123169
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,2,8,power_law_1.01,0.028838400840759275
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,2,8,balanced,0.16660991668701172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,1,16,power_law_1.2,0.5166259384155273
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,2,8,power_law_1.01,0.039944961071014404
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,2,8,power_law_1.01,0.05489664077758789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,2,8,power_law_1.01,0.06383999824523925
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,2,8,power_law_1.01,0.07308032035827636
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,2,8,power_law_1.01,0.10005120277404785
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,2,8,power_law_1.01,0.04828671932220459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,8,power_law_1.2,0.0545740795135498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,8,power_law_1.2,0.05530752182006836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,8,power_law_1.2,0.06412032127380371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,8,power_law_1.2,0.06782976150512696
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,2,8,power_law_1.01,0.11994367599487306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,8,power_law_1.2,0.10828927993774413
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,8,power_law_1.2,0.06958847999572754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,8,power_law_1.2,0.11426048278808594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,8,power_law_1.2,0.11572992324829101
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,8,power_law_1.2,0.11583616256713866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,8,power_law_1.2,0.12222463607788085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,8,power_law_1.2,0.11951231956481934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,8,power_law_1.2,0.11943679809570314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,8,power_law_1.2,0.1261401557922363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,8,power_law_1.2,0.13569279670715334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,8,power_law_1.2,0.1503001594543457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,8,power_law_1.2,0.1849318313598633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,8,power_law_1.2,0.10935423851013185
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,8,power_law_1.2,0.22162944793701173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,8,power_law_1.2,0.30359935760498047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,8,power_law_1.2,0.37472766876220703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,8,power_law_1.2,0.1320742416381836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,8,power_law_1.2,0.507512321472168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,8,power_law_1.2,0.6391257476806641
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,2,8,power_law_1.01,0.193121280670166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,8,power_law_1.2,0.03179392099380493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,8,power_law_1.2,0.03698303937911988
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,8,power_law_1.2,0.044501757621765135
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,8,power_law_1.2,0.04851712226867676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,8,power_law_1.2,0.04259840011596679
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,8,power_law_1.2,0.05947264194488525
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,8,power_law_1.2,0.0610368013381958
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,8,power_law_1.2,0.06547584056854247
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,8,power_law_1.2,0.06645376205444335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,8,power_law_1.2,0.06782847881317139
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,8,power_law_1.2,0.07668992042541503
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,8,power_law_1.2,0.9065382385253906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,8,power_law_1.2,0.06582143783569336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,8,power_law_1.2,0.06667776107788086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,8,power_law_1.2,0.0729203176498413
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,8,power_law_1.2,0.07384575843811035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,8,power_law_1.2,0.0767257595062256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,8,power_law_1.2,0.08204159736633301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,8,power_law_1.2,0.10214528083801269
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,8,power_law_1.2,0.11811455726623535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,8,power_law_1.2,0.1554700756072998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,8,power_law_1.2,1.2456371307373046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,8,power_law_1.2,0.18697343826293947
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,8,power_law_1.2,0.2671564865112305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,8,power_law_1.2,0.46932735443115237
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,8,power_law_1.2,0.32110721588134766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,8,power_law_1.2,0.6022963333129883
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,2,8,power_law_1.2,0.02199552059173584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,2,8,power_law_1.2,0.024426240921020508
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,2,8,power_law_1.2,0.0275161600112915
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,2,8,power_law_1.2,0.02993664026260376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,2,8,power_law_1.2,0.027201280593872067
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,2,8,power_law_1.2,0.0328601598739624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,2,8,power_law_1.2,0.033841919898986814
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,2,8,power_law_1.2,0.03451904058456421
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,2,8,power_law_1.2,0.03515775918960572
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,2,8,power_law_1.2,0.035340800285339355
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,2,8,power_law_1.2,0.03784192085266113
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,2,8,power_law_1.2,0.03745791912078857
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,2,8,power_law_1.2,0.03871104001998901
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,2,8,power_law_1.2,0.039563519954681395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,8,power_law_1.2,1.7798310852050783
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,2,8,power_law_1.01,0.18767360687255857
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,2,8,power_law_1.01,0.3464076614379883
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,2,8,power_law_1.01,0.4359795379638672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,8,power_law_1.2,0.8828224182128906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,2,8,power_law_1.2,0.07391615867614745
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,8,power_law_1.2,1.2613158416748047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,2,8,power_law_1.2,0.04234496116638183
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,2,8,power_law_1.2,0.04239615917205811
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,2,8,power_law_1.2,0.044399361610412594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,2,8,power_law_1.2,0.05591807842254639
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,2,8,power_law_1.2,0.06316927909851075
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,2,8,power_law_1.2,0.09095040321350098
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,4,balanced,0.026800639629364013
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,4,balanced,0.02725503921508789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,4,balanced,0.0423795223236084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,4,balanced,0.05699584007263183
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,4,balanced,0.078023681640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,4,balanced,0.08070272445678711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,4,balanced,0.11010560035705566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,4,balanced,0.11348352432250977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,4,balanced,0.11513343811035157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,4,balanced,0.12040191650390626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,4,balanced,0.11105792045593262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,4,balanced,0.11400959968566896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,4,balanced,0.11731072425842286
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,4,balanced,0.12031104087829589
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,4,balanced,0.12359552383422852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,4,balanced,0.1283724784851074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,4,balanced,0.1359488010406494
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,4,balanced,0.15658495903015138
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,4,balanced,0.17824512481689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,4,balanced,0.2227840042114258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,4,balanced,0.2654579162597656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,8,power_law_1.2,2.358423004150391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,4,balanced,0.3798937606811523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,4,balanced,0.4874099349975586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,4,balanced,0.701717758178711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,4,balanced,0.9018342590332031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,4,balanced,1.3320895385742186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,4,balanced,0.02844928026199341
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,4,balanced,0.029765119552612306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,4,balanced,0.034252800941467286
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,4,balanced,0.04183040142059326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,4,balanced,1.750689239501953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,4,balanced,0.06186880111694336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,4,balanced,0.0645747184753418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,4,balanced,0.07296512126922608
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,4,balanced,0.07624320030212403
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,4,balanced,0.0767897605895996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,4,balanced,0.08070912361145019
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,4,balanced,0.07086976051330567
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,4,balanced,0.07309440135955811
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,4,balanced,0.07514111995697022
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,4,balanced,0.07759488105773926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,4,balanced,0.07912960052490234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,4,balanced,0.0822054386138916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,4,balanced,0.08655488014221191
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,4,balanced,0.10203519821166993
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,4,balanced,0.11276927947998047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,4,balanced,0.1414732837677002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,4,balanced,0.1649817657470703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,4,balanced,0.24506496429443358
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,4,balanced,0.29359487533569334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,4,balanced,0.4094963073730469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,4,balanced,0.5361088180541993
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,4,balanced,0.7739955139160156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,2,8,power_law_1.2,0.11914752006530763
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,4,balanced,1.0306803131103517
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,2,8,power_law_1.2,0.1506713581085205
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,2,8,power_law_1.2,0.22696575164794922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,2,8,power_law_1.2,0.3120780754089355
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,4,4,balanced,0.02231168031692505
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,4,4,balanced,0.022717440128326417
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,4,4,balanced,0.04353536128997802
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,4,4,balanced,0.04417920112609863
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,4,4,balanced,0.0440831995010376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,4,4,balanced,0.04425343990325928
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,4,4,balanced,0.044303359985351565
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,4,4,balanced,0.02430720090866089
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,4,4,balanced,0.044592638015747074
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,4,4,balanced,0.04478464126586914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,4,4,balanced,0.0448473596572876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,4,4,balanced,0.04624767780303955
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,4,4,balanced,0.046496000289916996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,4,4,balanced,0.04952832221984863
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,4,4,balanced,0.05188735961914063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,4,4,balanced,0.055042557716369636
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,4,4,balanced,0.058583040237426755
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,4,4,balanced,0.04777088165283203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,2,8,power_law_1.2,0.6734809875488281
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,2,8,power_law_1.2,0.46727809906005857
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,4,4,balanced,0.027560958862304686
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,4,4,balanced,0.0435097599029541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,4,power_law_1.01,0.03703808069229126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,4,power_law_1.01,0.04852608203887939
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,4,power_law_1.01,0.05418496131896973
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,4,power_law_1.01,0.05999743938446045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,4,power_law_1.01,0.06738944053649902
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,4,power_law_1.01,0.08097408294677735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,4,power_law_1.01,0.10967679977416993
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,4,power_law_1.01,0.11172736167907715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,4,power_law_1.01,0.11524352073669433
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,4,power_law_1.01,0.11872127532958984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,4,power_law_1.01,0.11704319953918456
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,4,power_law_1.01,0.11204992294311525
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,4,power_law_1.01,0.11433471679687499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,4,power_law_1.01,0.1216038417816162
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,4,power_law_1.01,0.12400511741638183
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,4,power_law_1.01,0.13010432243347167
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,4,power_law_1.01,0.14119039535522462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,4,power_law_1.01,0.165532169342041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,4,power_law_1.01,0.18739456176757813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,4,power_law_1.01,0.2534592056274414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,4,power_law_1.01,0.3213670349121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,4,4,balanced,0.043884801864624026
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,4,power_law_1.01,0.4168896102905274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,4,power_law_1.01,0.5408499145507812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,4,power_law_1.01,0.7519347381591797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,4,power_law_1.01,0.9930137634277344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,4,power_law_1.01,1.4515493774414063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,4,power_law_1.01,0.0306713604927063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,4,power_law_1.01,0.03769471883773804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,4,power_law_1.01,0.041251840591430666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,4,power_law_1.01,1.9362509155273435
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,4,power_law_1.01,0.04550015926361084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,4,power_law_1.01,0.052263679504394533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,4,power_law_1.01,0.06151936054229736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,4,power_law_1.01,0.06959360122680665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,4,power_law_1.01,0.0720255994796753
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,4,power_law_1.01,0.0767091178894043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,4,power_law_1.01,0.08050559997558594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,4,power_law_1.01,0.07713664054870606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,4,power_law_1.01,0.07220736026763916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,4,power_law_1.01,0.07358975887298584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,4,power_law_1.01,0.0777894401550293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,4,power_law_1.01,0.07769855976104736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,4,power_law_1.01,0.08135295867919921
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,4,power_law_1.01,0.08805376052856445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,4,power_law_1.01,0.10469504356384278
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,4,power_law_1.01,0.1192319965362549
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,4,power_law_1.01,0.16082687377929689
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,4,power_law_1.01,0.189304313659668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,4,power_law_1.01,0.24982784271240233
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,4,power_law_1.01,0.3251699066162109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,4,power_law_1.01,0.4548185729980469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,4,power_law_1.01,0.5641561508178711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,4,power_law_1.01,0.8105728149414062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,4,4,power_law_1.01,0.023559679985046385
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,4,4,power_law_1.01,0.025582079887390134
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,4,power_law_1.01,1.097754898071289
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,4,4,power_law_1.01,0.027799038887023925
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,4,4,power_law_1.01,0.029662721157073975
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,4,4,power_law_1.01,0.030975999832153323
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,4,4,power_law_1.01,0.04021376132965088
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,4,4,power_law_1.01,0.04197760105133057
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,4,4,balanced,0.07933440208435058
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,4,4,power_law_1.01,0.04228991985321045
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,4,4,power_law_1.01,0.04345600128173828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,4,4,power_law_1.01,0.043692798614501954
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,4,4,power_law_1.01,0.042325758934021
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,4,4,power_law_1.01,0.04564991950988769
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,4,4,balanced,0.08747136116027832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,4,4,power_law_1.01,0.04415103912353516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,4,4,power_law_1.01,0.04627967834472656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,4,4,power_law_1.01,0.051799039840698245
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,4,4,power_law_1.01,0.054978561401367185
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,4,4,balanced,0.11592320442199706
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,4,4,balanced,0.14923775672912598
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,4,4,balanced,0.20830848693847653
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,4,4,balanced,0.27182592391967775
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,4,4,power_law_1.01,0.0822873592376709
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,4,4,power_law_1.01,0.04944511890411377
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,4,4,power_law_1.01,0.0674393606185913
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,4,4,power_law_1.01,0.10715776443481446
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,4,4,power_law_1.01,0.13458687782287598
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,4,power_law_1.2,0.030188798904418945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,4,power_law_1.2,0.04533760070800781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,4,4,power_law_1.01,0.18235008239746092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,4,power_law_1.2,0.05531648159027099
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,4,power_law_1.2,0.05623040199279785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,4,power_law_1.2,0.06449024200439453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,4,4,power_law_1.01,0.061538558006286624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,4,power_law_1.2,0.07639039993286133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,4,power_law_1.2,0.10762240409851073
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,4,power_law_1.2,0.11033856391906738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,4,power_law_1.2,0.11573760032653808
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,4,power_law_1.2,0.11786751747131348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,4,power_law_1.2,0.11758336067199707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,4,power_law_1.2,0.11252223968505859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,4,power_law_1.2,0.11557888031005858
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,4,power_law_1.2,0.12005120277404786
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,4,power_law_1.2,0.12592000007629395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,4,power_law_1.2,0.13163392066955565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,4,power_law_1.2,0.14192255973815918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,4,power_law_1.2,0.16695552825927734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,4,power_law_1.2,0.18477439880371094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,4,power_law_1.2,0.25956607818603517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,4,power_law_1.2,0.32916351318359377
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,4,4,power_law_1.01,0.2364863967895508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,4,power_law_1.2,0.4424307250976563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,4,power_law_1.2,0.5412620925903321
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,4,power_law_1.2,0.03014784097671509
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,4,power_law_1.2,0.03454976081848145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,4,power_law_1.2,0.04078080177307129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,4,power_law_1.2,0.04412288188934326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,4,power_law_1.2,0.047168002128601075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,4,power_law_1.2,0.8346841430664063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,4,power_law_1.2,0.07139584064483642
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,4,power_law_1.2,0.07391744136810303
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,4,power_law_1.2,0.07555200099945068
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,4,power_law_1.2,0.07889791965484619
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,4,power_law_1.2,0.07918848037719727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,4,power_law_1.2,0.05844607830047608
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,4,power_law_1.2,0.07217919826507568
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,4,power_law_1.2,0.07441792011260986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,4,power_law_1.2,0.07851136207580567
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,4,power_law_1.2,0.08041855812072754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,4,power_law_1.2,0.08298368453979492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,4,power_law_1.2,0.08914431571960449
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,4,power_law_1.2,0.10612095832824706
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,4,power_law_1.2,0.1223423957824707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,4,power_law_1.2,0.16566528320312499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,4,power_law_1.2,0.19274368286132812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,4,power_law_1.2,0.3323942565917969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,4,power_law_1.2,0.2527513694763184
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,4,power_law_1.2,1.0555110168457031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,4,power_law_1.2,0.46696319580078127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,4,power_law_1.2,0.5671295928955078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,4,power_law_1.2,1.5544844055175782
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,4,4,power_law_1.2,0.026398720741271975
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,4,4,power_law_1.2,0.0276147198677063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,4,4,power_law_1.2,0.029249279499053954
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,4,4,power_law_1.2,0.031203839778900143
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,4,4,power_law_1.2,0.03933439970016479
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,4,4,power_law_1.2,0.04195199966430664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,4,4,power_law_1.2,0.04260863780975342
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,4,4,power_law_1.2,0.04205952167510986
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,4,4,power_law_1.2,0.042873601913452145
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,4,4,power_law_1.2,0.04375040054321289
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,4,4,power_law_1.2,0.046292481422424314
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,4,4,power_law_1.2,0.047425279617309565
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,4,4,power_law_1.01,0.3406438446044922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,4,4,power_law_1.2,0.05346303939819336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,4,4,power_law_1.2,0.05490943908691406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,4,4,power_law_1.01,0.4517004776000976
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,4,4,power_law_1.01,0.6305523300170899
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,4,power_law_1.2,0.8746841430664063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,4,4,power_law_1.2,0.023959040641784668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,4,power_law_1.2,1.1357504272460939
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,4,4,power_law_1.2,0.0502348804473877
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,4,4,power_law_1.2,0.05942783832550049
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,4,4,power_law_1.2,0.07538943767547608
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,4,4,power_law_1.2,0.07425280094146729
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,4,4,power_law_1.2,0.10746111869812011
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,4,power_law_1.2,1.9587930297851563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,8,2,balanced,0.025347840785980225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,8,2,balanced,0.027549440860748294
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,8,2,balanced,0.044313597679138186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,8,2,balanced,0.057354240417480475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,8,2,balanced,0.07858304023742677
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,8,2,balanced,0.08004863739013672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,8,2,balanced,0.07977344036102295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,8,2,balanced,0.08117119789123535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,8,2,balanced,0.11574399948120118
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,8,2,balanced,0.1165171241760254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,8,2,balanced,0.11739904403686523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,8,2,balanced,0.12180224418640137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,8,2,balanced,0.12301183700561524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,8,2,balanced,0.12697088241577148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,8,2,balanced,0.13003007888793944
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,8,2,balanced,0.13414784431457522
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,8,2,balanced,0.1431552028656006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,8,2,balanced,0.16307327270507813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,8,2,balanced,0.18116479873657226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,8,2,balanced,0.22543104171752928
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,8,2,balanced,0.2629350471496582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,8,2,balanced,0.3960166549682617
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,4,4,power_law_1.2,0.12895872116088866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,8,2,balanced,0.4784832000732422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,8,2,balanced,0.6839257812499999
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,8,2,balanced,0.8782335662841797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,4,4,power_law_1.2,0.1785113525390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,8,2,balanced,1.2893696594238282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,8,2,balanced,0.02849663972854614
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,8,2,balanced,0.030602240562438966
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,8,2,balanced,0.04989312171936035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,8,2,balanced,0.07392767906188966
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,8,2,balanced,1.7012313842773437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,8,2,balanced,0.07433216094970703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,8,2,balanced,0.07516160011291503
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,8,2,balanced,0.07597439765930177
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,8,2,balanced,0.08721920013427734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,8,2,balanced,0.08805631637573243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,8,2,balanced,0.08917247772216796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,8,2,balanced,0.0920576000213623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,8,2,balanced,0.09347071647644042
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,8,2,balanced,0.09517824172973632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,8,2,balanced,0.09565312385559081
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,8,2,balanced,0.09928704261779785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,8,2,balanced,0.10460288047790527
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,8,2,balanced,0.11500160217285156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,4,4,power_law_1.2,0.24554624557495117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,8,2,balanced,0.12649855613708497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,8,2,balanced,0.14942848205566406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,8,2,balanced,0.17513216018676758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,8,2,balanced,0.26327423095703123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,8,2,balanced,0.3094156837463379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,8,2,balanced,0.44522880554199223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,8,2,balanced,0.5805350494384766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,8,2,balanced,0.8473088073730469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,8,2,balanced,0.035427839756011964
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,8,2,balanced,1.1073011016845702
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,4,4,power_law_1.2,0.36420478820800783
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,4,4,power_law_1.2,0.666899185180664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,4,4,power_law_1.2,0.5127308654785157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,8,2,balanced,0.02970880031585693
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,8,2,balanced,0.05873023986816406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,8,2,balanced,0.030524160861968998
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,8,2,balanced,0.05912447929382324
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,8,2,balanced,0.05911295890808106
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,8,2,balanced,0.05917568206787109
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,8,2,balanced,0.05923327922821044
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,8,2,balanced,0.03332223892211914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,8,2,balanced,0.06079487800598145
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,8,2,balanced,0.03974015951156616
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,8,2,balanced,0.0613593578338623
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,8,2,balanced,0.06203392028808594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,8,2,balanced,0.0635750389099121
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,8,2,balanced,0.06482687950134278
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,8,2,balanced,0.06826367855072021
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,8,2,balanced,0.07132415771484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,8,2,balanced,0.0780787181854248
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,8,2,balanced,0.08368895530700685
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,8,2,balanced,0.05800191879272461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,8,2,balanced,0.05885439872741699
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,8,2,balanced,0.059180798530578616
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,8,2,power_law_1.01,0.026346240043640133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,8,2,power_law_1.01,0.04311808109283447
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,8,2,power_law_1.01,0.0482803201675415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,8,2,power_law_1.01,0.05483136177062988
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,8,2,power_law_1.01,0.061708798408508306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,8,2,power_law_1.01,0.07625855922698974
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,8,2,power_law_1.01,0.07597055912017822
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,8,2,power_law_1.01,0.07783167839050294
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,8,2,power_law_1.01,0.11194879531860351
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,8,2,power_law_1.01,0.11421824455261231
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,8,2,power_law_1.01,0.11624320030212403
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,8,2,power_law_1.01,0.11965696334838867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,8,2,power_law_1.01,0.12153087615966798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,8,2,power_law_1.01,0.12788607597351073
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,8,2,power_law_1.01,0.13124608039855956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,8,2,power_law_1.01,0.1365452766418457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,8,2,power_law_1.01,0.14635135650634765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,8,2,power_law_1.01,0.16661632537841797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,8,2,balanced,0.1349350357055664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,8,2,balanced,0.11926527976989747
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,8,2,power_law_1.01,0.1872371292114258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,8,2,balanced,0.2354086494445801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,8,2,power_law_1.01,0.24024831771850588
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,8,2,power_law_1.01,0.30777343749999997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,8,2,power_law_1.01,0.4049356842041016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,8,2,power_law_1.01,0.5129459381103516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,8,2,power_law_1.01,0.7283519744873047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,8,2,balanced,0.1827827262878418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,8,2,power_law_1.01,0.02882688045501709
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,8,2,power_law_1.01,0.9208255767822265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,8,2,power_law_1.01,0.03368319988250733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,8,2,power_law_1.01,0.04199552059173584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,8,2,power_law_1.01,0.05092991828918457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,8,2,power_law_1.01,0.05544447898864746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,8,2,power_law_1.01,0.06789120197296142
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,8,2,power_law_1.01,0.0716326379776001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,8,2,power_law_1.01,1.376794891357422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,8,2,balanced,0.3329663848876953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,8,2,power_law_1.01,0.08171775817871094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,8,2,power_law_1.01,0.08446335792541504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,8,2,power_law_1.01,0.08660351753234863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,8,2,power_law_1.01,0.08844672203063965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,8,2,power_law_1.01,0.09140607833862305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,8,2,power_law_1.01,1.7805325317382814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,8,2,power_law_1.01,0.09634431838989257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,8,2,power_law_1.01,0.09875840187072754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,8,2,power_law_1.01,0.10605183601379395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,8,2,power_law_1.01,0.12742015838623047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,8,2,power_law_1.01,0.14952704429626465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,8,2,power_law_1.01,0.18182655334472658
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,8,2,power_law_1.01,0.21300224304199217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,8,2,power_law_1.01,0.27864448547363285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,8,2,power_law_1.01,0.3517171096801758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,8,2,power_law_1.01,0.48913406372070317
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,8,2,power_law_1.01,0.6269567871093751
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,8,2,power_law_1.01,0.030883839130401613
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,8,2,power_law_1.01,0.03248512029647827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,8,2,power_law_1.01,0.9115468597412109
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,8,2,power_law_1.01,0.03631103992462158
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,8,2,power_law_1.01,0.0389683198928833
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,8,2,power_law_1.01,0.04276735782623291
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,8,2,power_law_1.01,0.05491072177886963
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,8,2,power_law_1.01,0.05646207809448243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,8,2,power_law_1.01,1.175582733154297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,8,2,power_law_1.01,0.05638527870178223
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,8,2,power_law_1.01,0.057217278480529786
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,8,2,power_law_1.01,0.058277120590209966
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,8,2,power_law_1.01,0.05849215984344482
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,8,2,power_law_1.01,0.06137343883514405
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,8,2,power_law_1.01,0.06266496181488038
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,8,2,power_law_1.01,0.06442751884460449
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,8,2,power_law_1.01,0.0674675178527832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,8,2,power_law_1.01,0.09645952224731445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,8,2,power_law_1.01,0.07260799884796143
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,8,2,balanced,0.4352486419677734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,4,4,power_law_1.2,0.9782489776611328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,8,2,power_law_1.01,0.07324672222137452
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,8,2,power_law_1.01,0.09161600112915039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,8,2,power_law_1.01,0.10286720275878906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,8,2,power_law_1.01,0.07903999805450439
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,8,2,power_law_1.01,0.13364607810974122
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,8,2,power_law_1.01,0.15719807624816895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,8,2,power_law_1.2,0.029433600902557373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,8,2,power_law_1.2,0.04236927986145019
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,8,2,power_law_1.2,0.04769152164459228
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,8,2,power_law_1.2,0.052439041137695305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,8,2,power_law_1.2,0.059020800590515135
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,8,2,power_law_1.2,0.07422207832336426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,8,2,power_law_1.2,0.07582079887390136
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,8,2,power_law_1.2,0.07960063934326171
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,8,2,power_law_1.2,0.11393792152404784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,8,2,power_law_1.2,0.11374848365783692
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,8,2,power_law_1.2,0.11591423988342284
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,8,2,power_law_1.2,0.11895551681518554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,8,2,power_law_1.2,0.12256640434265136
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,8,2,power_law_1.2,0.1284671974182129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,8,2,power_law_1.2,0.13265664100646973
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,8,2,power_law_1.2,0.13721088409423826
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,8,2,power_law_1.2,0.14665216445922852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,8,2,power_law_1.2,0.1683647918701172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,8,2,power_law_1.2,0.1880691146850586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,8,2,power_law_1.2,0.23950336456298826
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,8,2,power_law_1.2,0.3130073547363281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,8,2,power_law_1.2,0.41078910827636717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,8,2,power_law_1.2,0.5134284973144532
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,8,2,power_law_1.01,0.22753664016723635
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,8,2,power_law_1.2,0.7571609497070313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,8,2,power_law_1.2,0.9319987487792968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,8,2,power_law_1.01,0.30411520004272463
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,8,2,power_law_1.2,0.03507200002670288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,8,2,power_law_1.2,0.04144256114959717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,8,2,power_law_1.2,0.050067200660705566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,8,2,power_law_1.2,0.053831682205200196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,8,2,power_law_1.2,0.06815999984741211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,8,2,power_law_1.2,0.07071360111236571
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,8,2,power_law_1.2,0.07300352096557618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,8,2,power_law_1.2,0.0835814380645752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,8,2,power_law_1.2,0.08718079566955567
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,8,2,power_law_1.2,0.08871552467346192
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,8,2,power_law_1.2,0.03133440017700195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,8,2,power_law_1.2,0.09174400329589844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,8,2,power_law_1.2,0.09337344169616699
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,8,2,power_law_1.2,0.09639552116394043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,8,2,power_law_1.2,0.09788031578063965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,8,2,power_law_1.2,0.10139264106750487
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,8,2,power_law_1.2,0.10790143966674806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,8,2,power_law_1.2,0.13362688064575196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,8,2,power_law_1.2,0.15027071952819823
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,8,2,power_law_1.2,0.18278911590576172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,8,2,power_law_1.2,0.21603328704833985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,8,2,power_law_1.2,0.3537638473510742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,8,2,power_law_1.01,0.396416015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,8,2,power_law_1.2,0.2827878379821777
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,8,2,power_law_1.01,0.6178406524658203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,8,2,power_law_1.01,0.8305190277099609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,8,2,power_law_1.2,0.6475737762451172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,8,2,power_law_1.2,0.49168769836425785
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,8,2,power_law_1.2,0.032960000038146975
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,8,2,power_law_1.2,0.03589632034301758
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,8,2,power_law_1.2,0.03778304100036621
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,8,2,power_law_1.2,0.04169727802276611
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,8,2,power_law_1.01,1.1402803039550782
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,8,2,power_law_1.2,0.05439104080200195
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,8,2,power_law_1.2,0.054931201934814454
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,8,2,power_law_1.2,0.058822398185729985
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,8,2,power_law_1.2,0.030906879901885988
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,8,2,power_law_1.2,0.059018239974975586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,8,2,power_law_1.2,0.06057472229003906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,8,2,power_law_1.2,0.062357759475708006
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,8,2,power_law_1.2,0.06382847785949707
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,8,2,power_law_1.2,0.06974207878112793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,8,2,power_law_1.2,0.057681918144226074
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,8,2,power_law_1.2,0.07765247821807861
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,8,2,power_law_1.2,1.3900006103515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,8,2,power_law_1.2,0.9331801605224609
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,8,2,power_law_1.2,0.07413375854492188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,8,2,power_law_1.2,0.08575872421264649
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,8,2,power_law_1.2,0.10413696289062499
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,8,2,power_law_1.2,0.11759615898132325
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,8,2,power_law_1.2,0.1676684761047363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,8,2,power_law_1.2,1.7969215393066407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,8,2,power_law_1.2,1.1829055786132812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,16,1,balanced,0.0311955189704895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,16,1,balanced,0.035750401020050046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,16,1,balanced,0.05494463920593262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,16,1,balanced,0.07409984111785889
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,16,1,balanced,0.10229567527770995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,16,1,balanced,0.10385536193847655
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,8,2,power_law_1.2,0.26658304214477535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,16,1,balanced,0.10433792114257812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,8,2,power_law_1.2,0.1968307113647461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,16,1,balanced,0.10682368278503418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,16,1,balanced,0.1104588794708252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,16,1,balanced,0.15483712196350097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,16,1,balanced,0.15734272003173827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,16,1,balanced,0.16104448318481446
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,8,2,power_law_1.2,0.40292736053466793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,16,1,balanced,0.16391359329223634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,16,1,balanced,0.16772031784057617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,16,1,balanced,0.17528127670288085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,16,1,balanced,0.19396543502807617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,16,1,balanced,0.10643327713012696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,16,1,balanced,0.10835968017578126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,16,1,balanced,0.2109996795654297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,16,1,balanced,0.2570144081115723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,16,1,balanced,0.29164480209350585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,16,1,balanced,0.4042937469482422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,16,1,balanced,0.5043155288696288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,16,1,balanced,0.7039667510986328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,8,2,power_law_1.2,0.5337996673583985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,16,1,balanced,0.035399680137634275
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,16,1,balanced,0.03983680009841919
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,16,1,balanced,0.8966745758056641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,16,1,balanced,0.0484665584564209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,16,1,balanced,0.06718719959259033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,16,1,balanced,0.1096275234222412
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,16,1,balanced,0.10917823791503907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,16,1,balanced,0.10893376350402832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,16,1,balanced,0.11103808403015136
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,16,1,balanced,1.303768310546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,16,1,balanced,0.1110649585723877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,16,1,balanced,0.11293120384216308
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,16,1,balanced,0.11477375984191895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,16,1,balanced,0.12906047821044922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,16,1,balanced,0.13078783988952636
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,16,1,balanced,0.13364864349365235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,16,1,balanced,0.1341766357421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,16,1,balanced,0.1350451183319092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,16,1,balanced,0.13996288299560547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,16,1,balanced,0.15036224365234374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,16,1,balanced,1.6988192749023436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,16,1,balanced,0.16097087860107423
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,16,1,balanced,0.18201471328735352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,16,1,balanced,0.20416704177856446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,16,1,balanced,0.3101548767089844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,16,1,balanced,0.35062911987304685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,16,1,balanced,0.4974790573120117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,16,1,balanced,0.6502413177490235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,16,1,balanced,0.9395807647705079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,16,1,balanced,1.230108184814453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,8,2,power_law_1.2,0.8522662353515624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,16,1,balanced,0.04391808032989502
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,16,1,balanced,0.04575488090515137
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,16,1,balanced,0.09170944213867187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,16,1,balanced,0.08866175651550293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,16,1,balanced,0.05046144008636475
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,16,1,balanced,0.06455167770385742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,16,1,balanced,0.09196160316467286
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,16,1,balanced,0.09081727981567382
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,16,1,balanced,0.09442496299743652
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,16,1,balanced,0.08833344459533692
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,8,2,power_law_1.2,1.1990016174316405
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,16,1,balanced,0.09423295974731445
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,16,1,balanced,0.09566528320312499
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,16,1,balanced,0.09541760444641113
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,16,1,balanced,0.09995967864990234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,16,1,balanced,0.09840831756591797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,16,1,balanced,0.11085184097290039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,16,1,balanced,0.11778240203857422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,16,1,balanced,0.12813376426696776
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,16,1,balanced,0.1459481620788574
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,16,1,balanced,0.08926912307739257
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,16,1,balanced,0.10283519744873046
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,16,1,balanced,0.22083967208862304
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,16,1,balanced,0.24904064178466795
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,16,1,balanced,0.3535539245605469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,16,1,power_law_1.01,0.06642111778259277
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,16,1,balanced,0.45104831695556646
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,16,1,power_law_1.01,0.07750592231750489
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,16,1,power_law_1.01,0.09299072265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,16,1,power_law_1.01,0.033276159763336186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,16,1,power_law_1.01,0.09714816093444825
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,16,1,power_law_1.01,0.10003711700439452
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,16,1,power_law_1.01,0.10380031585693358
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,16,1,power_law_1.01,0.10714816093444823
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,16,1,power_law_1.01,0.11401599884033202
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,16,1,power_law_1.01,0.1533478355407715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,16,1,power_law_1.01,0.04850751876831055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,16,1,power_law_1.01,0.15590271949768067
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,16,1,power_law_1.01,0.0577184009552002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,16,1,power_law_1.01,0.16186943054199218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,16,1,power_law_1.01,0.1653670310974121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,16,1,power_law_1.01,0.16993600845336915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,16,1,power_law_1.01,0.1796313667297363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,16,1,power_law_1.01,0.1996531105041504
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,16,1,balanced,0.653022689819336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,16,1,power_law_1.01,0.21851264953613284
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,16,1,power_law_1.01,0.26019519805908203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,16,1,power_law_1.01,0.3207187271118164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,16,1,power_law_1.01,0.430519027709961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,16,1,power_law_1.01,0.5207321548461914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,16,1,power_law_1.01,0.03569855928421021
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,16,1,power_law_1.01,0.04020415782928467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,16,1,power_law_1.01,0.7279122924804687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,16,1,power_law_1.01,0.05413119792938233
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,16,1,power_law_1.01,0.9228659057617188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,16,1,power_law_1.01,0.06401984214782715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,16,1,power_law_1.01,0.08187711715698243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,16,1,power_law_1.01,0.10415040016174318
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,16,1,power_law_1.01,0.09907903671264648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,16,1,power_law_1.01,0.10393471717834472
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,16,1,power_law_1.01,0.10807744026184082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,16,1,power_law_1.01,1.3233164978027343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,16,1,power_law_1.01,0.10939647674560546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,16,1,power_law_1.01,0.11190208435058593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,16,1,power_law_1.01,0.12783488273620605
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,16,1,balanced,0.8597555541992188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,16,1,power_law_1.01,0.1300825595855713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,16,1,power_law_1.01,0.1335654354095459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,16,1,power_law_1.01,0.13559679985046386
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,16,1,power_law_1.01,0.13804032325744628
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,16,1,power_law_1.01,1.7245843505859377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,16,1,power_law_1.01,0.147576961517334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,16,1,power_law_1.01,0.1614726448059082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,16,1,power_law_1.01,0.19032831192016603
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,16,1,power_law_1.01,0.22135808944702146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,16,1,power_law_1.01,0.25818304061889646
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,16,1,power_law_1.01,0.32504638671874997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,16,1,power_law_1.01,0.404310417175293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,8,2,power_law_1.2,1.668706512451172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,16,1,power_law_1.01,0.5524224090576172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,16,1,power_law_1.01,0.6973139190673828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,16,1,power_law_1.01,0.044858241081237794
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,16,1,power_law_1.01,0.04723392009735107
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,16,1,power_law_1.01,0.05250048160552978
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,16,1,power_law_1.01,0.981335678100586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,16,1,power_law_1.01,0.06533376216888429
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,16,1,power_law_1.01,0.08174079895019531
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,16,1,power_law_1.01,0.08381312370300294
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,16,1,power_law_1.01,1.2764083099365233
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,16,1,power_law_1.01,0.0880947208404541
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,16,1,power_law_1.01,0.08905407905578613
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,16,1,power_law_1.01,0.09185855865478516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,16,1,power_law_1.01,0.09533760070800781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,16,1,power_law_1.01,0.09672191619873047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,16,1,power_law_1.01,0.10719807624816895
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,16,1,power_law_1.01,0.12371264457702638
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,16,1,power_law_1.01,0.09921855926513672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,16,1,power_law_1.01,0.05836607933044433
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,16,1,power_law_1.01,0.11754303932189943
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,16,1,power_law_1.01,0.13529727935791017
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,16,1,power_law_1.01,0.16160383224487304
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,16,1,power_law_1.01,0.2518662452697754
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,16,1,power_law_1.01,0.19224384307861328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,16,1,power_law_1.01,0.31629695892333987
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,16,1,power_law_1.2,0.048704638481140136
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,16,1,power_law_1.2,0.057118721008300774
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,16,1,power_law_1.01,0.4456607818603516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,16,1,power_law_1.2,0.06413311958312988
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,16,1,power_law_1.2,0.09327103614807128
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,16,1,power_law_1.2,0.07635200023651123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,16,1,power_law_1.2,0.09733695983886718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,16,1,power_law_1.2,0.10164799690246581
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,16,1,power_law_1.2,0.03205440044403076
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,16,1,power_law_1.2,0.10493887901306151
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,16,1,power_law_1.2,0.10868543624877929
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,16,1,power_law_1.2,0.1540998363494873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,16,1,power_law_1.2,0.15666751861572265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,16,1,power_law_1.2,0.16297792434692382
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,16,1,power_law_1.2,0.16607488632202147
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,16,1,power_law_1.2,0.17051776885986328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,16,1,power_law_1.2,0.17991615295410157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,16,1,power_law_1.2,0.11480319976806641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,16,1,power_law_1.2,0.2001241683959961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,16,1,power_law_1.2,0.2209644889831543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,16,1,power_law_1.2,0.26080320358276365
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,16,1,power_law_1.2,0.3209222412109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,16,1,power_law_1.2,0.4301472091674805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,16,1,power_law_1.2,0.5236217498779296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,16,1,power_law_1.2,0.7294137573242188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,16,1,power_law_1.01,0.5773267364501953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,16,1,power_law_1.2,0.036120960712432856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,16,1,power_law_1.2,0.04182911872863769
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,16,1,power_law_1.2,0.05180607795715332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,16,1,power_law_1.01,0.849345932006836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,16,1,power_law_1.2,0.06310336112976074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,16,1,power_law_1.2,0.9245458984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,16,1,power_law_1.2,0.10185088157653807
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,16,1,power_law_1.2,0.10446975708007813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,16,1,power_law_1.2,0.07876736164093018
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,16,1,power_law_1.2,0.106693115234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,16,1,power_law_1.01,1.1202828979492188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,16,1,power_law_1.2,0.10912960052490235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,16,1,power_law_1.2,0.1102246379852295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,16,1,power_law_1.2,0.11199935913085937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,16,1,power_law_1.2,0.1303667163848877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,16,1,power_law_1.2,0.12928192138671873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,16,1,power_law_1.2,0.1369337558746338
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,16,1,power_law_1.2,0.14138367652893066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,16,1,power_law_1.2,0.14963135719299317
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,16,1,power_law_1.2,0.17255104064941407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,16,1,power_law_1.2,0.19261119842529298
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,16,1,power_law_1.2,0.2279916763305664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,16,1,power_law_1.2,0.13574208259582518
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,16,1,power_law_1.2,0.26159488677978515
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,16,1,power_law_1.01,1.685063018798828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,16,1,power_law_1.2,0.3258860778808594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,16,1,power_law_1.01,2.2753971862792968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,16,1,power_law_1.2,0.044803199768066404
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,16,1,power_law_1.2,0.40923583984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,16,1,power_law_1.2,0.04758143901824951
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,16,1,power_law_1.2,0.05215231895446777
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,16,1,power_law_1.2,0.05703680038452148
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,16,1,power_law_1.2,0.06271488189697265
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,16,1,power_law_1.2,0.08684224128723145
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,16,1,power_law_1.2,0.08325632095336914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,16,1,power_law_1.2,0.08898431777954101
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,16,1,power_law_1.2,0.09323519706726074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,16,1,power_law_1.2,0.5550233459472655
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,16,1,power_law_1.2,0.09332672119140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,16,1,power_law_1.2,0.09773695945739747
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,16,1,power_law_1.2,0.10049407958984376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,16,1,power_law_1.2,0.10385984420776366
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,16,1,power_law_1.2,0.12298496246337891
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,16,1,power_law_1.2,0.13188672065734863
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,16,1,power_law_1.2,0.1484166431427002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,16,1,power_law_1.2,0.7024320220947265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,16,1,power_law_1.2,1.326003875732422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,16,1,power_law_1.2,0.11559552192687987
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,16,1,power_law_1.2,0.18777984619140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,16,1,power_law_1.2,0.23056127548217775
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,16,1,power_law_1.2,0.9811654663085937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,16,1,power_law_1.2,0.3158752059936524
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,16,1,power_law_1.2,0.40649345397949216
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,16,1,power_law_1.2,1.729875793457031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,1,16,balanced,0.06670464038848876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,1,16,balanced,0.06660352230072022
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,1,16,balanced,0.07792895793914795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,1,16,balanced,0.09477375984191895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,1,16,balanced,0.16349184036254882
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,16,1,power_law_1.2,0.5905376052856445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,1,16,balanced,0.23170816421508791
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,1,16,balanced,0.2405913543701172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,1,16,balanced,0.2339878463745117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,1,16,balanced,0.23889280319213868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,1,16,balanced,0.2362726402282715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,1,16,balanced,0.23294464111328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,1,16,balanced,0.25084672927856444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,1,16,balanced,0.27520639419555665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,1,16,balanced,0.25649280548095704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,1,16,balanced,0.2654745674133301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,1,16,balanced,0.25664255142211917
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,1,16,balanced,0.2802432060241699
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,1,16,balanced,0.3235686492919922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,16,1,power_law_1.2,0.7770899200439454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,1,16,balanced,0.36884479522705077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,1,16,balanced,0.461187858581543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,1,16,balanced,0.5016281509399414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,1,16,balanced,0.6794662475585938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,1,16,balanced,0.8563391876220703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,1,16,balanced,1.1003852844238282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,1,16,balanced,1.4343948364257812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,1,16,balanced,0.03745791912078857
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,1,16,balanced,0.037480959892272944
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,1,16,balanced,0.04536448001861572
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,1,16,balanced,0.060244479179382325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,1,16,balanced,2.1533541870117188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,1,16,balanced,0.10388992309570313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,1,16,balanced,0.12220928192138672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,1,16,balanced,0.12596863746643067
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,1,16,balanced,0.12353023529052734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,1,16,balanced,0.1264844799041748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,1,16,balanced,0.12897151947021485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,1,16,balanced,0.1315392017364502
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,1,16,balanced,3.0764749145507815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,1,16,balanced,0.13408512115478516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,1,16,balanced,0.1406003189086914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,1,16,balanced,0.13154560089111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,16,1,power_law_1.2,1.1587782287597657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,1,16,balanced,0.13433088302612303
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,1,16,balanced,0.13771519660949708
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,1,16,balanced,0.1420697593688965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,1,16,balanced,0.1524454402923584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,1,16,balanced,0.17408767700195313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,1,16,balanced,0.1989504051208496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,1,16,balanced,0.21847040176391602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,1,16,balanced,0.3066355133056641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,1,16,balanced,0.3555865478515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,1,16,balanced,0.499214096069336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,1,16,balanced,0.6356224060058594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,1,16,balanced,0.854101791381836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,1,16,balanced,1.1395520019531251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,16,1,power_law_1.2,1.2745184326171874
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,16,1,power_law_1.2,1.5703353881835938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,1,16,balanced,0.022314240932464598
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,1,16,balanced,0.06581632137298585
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,1,16,balanced,0.022329599857330323
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,1,16,balanced,0.03984639883041382
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,1,16,balanced,0.02410239934921265
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,1,16,balanced,0.06585216045379638
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,1,16,balanced,0.06611328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,1,16,balanced,0.06636288166046142
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,1,16,balanced,0.06644864082336426
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,1,16,balanced,0.06614016056060791
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,1,16,balanced,0.06699776172637939
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,1,16,balanced,0.0669478416442871
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,1,16,balanced,0.06737664222717285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,16,1,power_law_1.2,2.4065644836425784
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,1,16,balanced,0.06775296211242676
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,1,16,balanced,0.06866047859191896
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,1,16,balanced,0.06924032211303711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,1,16,balanced,0.058583040237426755
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,1,16,balanced,0.0737779188156128
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,1,16,balanced,0.0659878396987915
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,1,16,balanced,0.06562687873840331
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,1,16,balanced,0.07127808094024658
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,1,16,balanced,0.10407296180725098
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,1,16,power_law_1.01,0.14219648361206055
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,1,16,balanced,0.13400959968566895
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,1,16,balanced,0.09804287910461426
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,1,16,balanced,0.20974847793579104
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,1,16,power_law_1.01,0.18388736724853516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,1,16,power_law_1.01,0.1592089557647705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,1,16,power_law_1.01,0.1894732856750488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,1,16,power_law_1.01,0.2082905578613281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,1,16,power_law_1.01,0.22483200073242188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,1,16,power_law_1.01,0.23592191696166992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,1,16,power_law_1.01,0.23038848876953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,1,16,power_law_1.01,0.2284351921081543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,1,16,power_law_1.01,0.23787008285522462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,1,16,power_law_1.01,0.25375871658325194
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,1,16,power_law_1.01,0.25691776275634765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,1,16,power_law_1.01,0.26386688232421873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,1,16,power_law_1.01,0.30131456375122073
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,1,16,power_law_1.01,0.31538431167602543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,1,16,power_law_1.01,0.3599884796142578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,1,16,power_law_1.01,0.38409343719482425
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,1,16,power_law_1.01,0.4704102325439453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,1,16,power_law_1.01,0.5805926513671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,1,16,power_law_1.01,0.6894566345214843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,1,16,power_law_1.01,1.0178816223144531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,1,16,power_law_1.01,0.2033625602722168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,1,16,power_law_1.01,1.2331878662109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,1,16,balanced,0.17061119079589843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,1,16,power_law_1.01,0.10202239990234374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,1,16,power_law_1.01,0.11056768417358398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,1,16,power_law_1.01,0.09973631858825684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,1,16,power_law_1.01,1.947129669189453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,1,16,power_law_1.01,0.1103116798400879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,1,16,power_law_1.01,0.07427072048187255
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,1,16,power_law_1.01,0.12083200454711913
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,1,16,power_law_1.01,0.12281855583190918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,1,16,power_law_1.01,0.11797760009765626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,1,16,power_law_1.01,0.1276211166381836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,1,16,power_law_1.01,0.128023042678833
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,1,16,power_law_1.01,0.13441023826599122
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,1,16,power_law_1.01,2.4238426208496096
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,1,16,balanced,0.28617984771728516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,1,16,power_law_1.01,0.13753600120544435
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,1,16,power_law_1.01,0.13510784149169922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,1,16,power_law_1.01,0.1338368034362793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,1,16,power_law_1.01,0.13819775581359864
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,1,16,power_law_1.01,0.13860735893249512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,1,16,power_law_1.01,0.15145983695983886
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,1,16,power_law_1.01,0.17113983154296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,1,16,power_law_1.01,0.215731201171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,1,16,power_law_1.01,0.2633459281921387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,1,16,power_law_1.01,0.31920896530151366
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,1,16,power_law_1.01,0.4187417602539062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,1,16,power_law_1.01,0.5230873489379884
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,1,16,power_law_1.01,0.0590169620513916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,1,16,power_law_1.01,0.7294425964355469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,16,1,power_law_1.2,3.216386413574219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,1,16,power_law_1.01,0.8871155548095704
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,1,16,power_law_1.01,0.06513152122497559
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,1,16,power_law_1.01,4.082867126464843
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,1,16,power_law_1.01,0.06530943870544434
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,1,16,power_law_1.01,0.0492083215713501
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,1,16,power_law_1.01,0.06441343784332275
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,1,16,power_law_1.01,0.06299903869628906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,1,16,power_law_1.01,0.06448256015777588
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,1,16,power_law_1.01,1.5464550781250002
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,1,16,power_law_1.01,0.06563968181610107
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,1,16,power_law_1.01,0.06465919971466064
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,1,16,power_law_1.01,0.06593408107757567
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,1,16,power_law_1.01,0.06517504215240479
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,1,16,power_law_1.01,0.06597504138946533
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,1,16,power_law_1.01,0.06752895832061767
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,1,16,power_law_1.01,1.8564915466308594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,1,16,power_law_1.01,0.06835840225219728
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,1,16,power_law_1.01,5.237031860351562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,1,16,power_law_1.01,0.0645849609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,1,16,power_law_1.01,0.07350783824920655
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,1,16,power_law_1.01,0.08921088218688965
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,1,16,power_law_1.01,0.07134592056274414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,1,16,power_law_1.01,0.11731455802917481
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,1,16,power_law_1.01,0.08235008239746093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,1,16,power_law_1.01,0.16163583755493166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,1,16,power_law_1.2,0.1397606372833252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,1,16,power_law_1.2,0.17281280517578126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,1,16,power_law_1.2,0.16604543685913087
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,1,16,power_law_1.01,0.18971391677856447
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,1,16,power_law_1.2,0.1887808036804199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,1,16,power_law_1.2,0.19781631469726563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,1,16,power_law_1.2,0.22447231292724606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,1,16,power_law_1.2,0.2288972854614258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,1,16,power_law_1.2,0.2374336051940918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,1,16,power_law_1.2,0.23516799926757814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,1,16,power_law_1.2,0.22637952804565428
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,1,16,power_law_1.2,0.25697664260864256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,1,16,power_law_1.2,0.27180288314819334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,1,16,power_law_1.01,0.10627584457397461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,1,16,power_law_1.2,0.26008832931518555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,1,16,power_law_1.2,0.28711551666259766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,1,16,power_law_1.2,0.33227649688720706
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,1,16,power_law_1.2,0.34593406677246097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,1,16,power_law_1.2,0.3696588897705078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,1,16,power_law_1.2,0.4295846557617187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,1,16,power_law_1.2,0.5160627365112305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,1,16,power_law_1.2,0.6732121276855468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,1,16,power_law_1.2,0.8721920013427734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,1,16,power_law_1.2,1.1041126251220703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,1,16,power_law_1.2,1.5319001770019531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,1,16,power_law_1.2,2.1612403869628904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,1,16,power_law_1.2,0.10234368324279786
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,1,16,power_law_1.2,0.09439104080200196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,1,16,power_law_1.2,0.10993791580200196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,1,16,power_law_1.2,2.9649688720703127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,1,16,power_law_1.2,0.1012377643585205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,1,16,power_law_1.2,0.10877696037292481
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,1,16,power_law_1.2,0.12276864051818848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,1,16,power_law_1.2,0.1161945629119873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,1,16,power_law_1.2,0.12377728462219237
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,1,16,power_law_1.2,0.12810367584228516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,1,16,power_law_1.2,0.1298303985595703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,1,16,power_law_1.2,0.1354099178314209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,1,16,power_law_1.2,0.14025471687316896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,1,16,power_law_1.2,0.13577728271484374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,1,16,power_law_1.2,0.13829504013061525
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,1,16,power_law_1.2,0.14031488418579102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,1,16,power_law_1.2,0.14702207565307618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,1,16,power_law_1.2,0.16057472229003905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,1,16,power_law_1.2,0.18764928817749021
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,1,16,power_law_1.2,5.3361315917968755
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,1,16,power_law_1.2,0.2224025535583496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,1,16,power_law_1.2,0.26124927520751956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,1,16,power_law_1.2,0.35399551391601564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,1,16,power_law_1.2,0.5204544067382812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,1,16,power_law_1.2,0.5830182266235352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,1,16,power_law_1.2,0.8986918640136718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,1,16,power_law_1.2,1.0933350372314454
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,1,16,power_law_1.2,0.05867263793945312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,1,16,power_law_1.2,0.057542400360107424
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,1,16,power_law_1.2,0.06451200008392334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,1,16,power_law_1.2,0.05961088180541992
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,1,16,power_law_1.2,0.04974720001220703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,1,16,power_law_1.2,0.06284416198730469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,1,16,power_law_1.2,0.062277121543884276
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,1,16,power_law_1.2,1.7798208618164062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,1,16,power_law_1.2,0.063438081741333
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,1,16,power_law_1.2,0.06326272010803222
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,1,16,power_law_1.2,0.06349567890167236
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,1,16,power_law_1.2,0.06522496223449707
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,1,16,power_law_1.2,0.0649945592880249
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,1,16,power_law_1.2,0.06771584033966065
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,1,16,power_law_1.2,0.0675865602493286
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,1,16,power_law_1.2,0.06842624187469483
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,1,16,power_law_1.2,2.583842468261719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,1,16,power_law_1.2,0.08702079772949219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,1,16,power_law_1.2,7.068230590820312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,1,16,power_law_1.01,0.26362112045288083
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,1,16,power_law_1.01,0.36229248046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,1,16,power_law_1.01,0.6046668624877929
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,1,16,power_law_1.01,0.7488755035400391
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,1,16,power_law_1.2,0.07095168113708497
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,1,16,power_law_1.2,0.078023681640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,1,16,power_law_1.2,0.09704192161560059
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,1,16,power_law_1.2,0.11903743743896485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,2,8,balanced,0.03535615921020508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,2,8,balanced,0.05954048156738281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,2,8,balanced,0.07147776126861573
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,1,16,power_law_1.2,0.13998720169067383
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,1,16,power_law_1.2,0.20941951751708981
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,2,8,balanced,0.16586496353149413
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,2,8,balanced,0.20745983123779296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,2,8,balanced,0.21548160552978515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,2,8,balanced,0.22112255096435546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,2,8,balanced,0.22331775665283202
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,1,16,power_law_1.2,0.23722368240356445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,2,8,balanced,0.22491008758544923
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,2,8,balanced,0.22492288589477538
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,2,8,balanced,0.22529279708862307
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,2,8,balanced,0.23312768936157227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,2,8,balanced,0.2421670341491699
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,2,8,balanced,0.2445017623901367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,2,8,balanced,0.2544742393493652
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,2,8,balanced,0.2784639930725098
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,2,8,balanced,0.3066790390014648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,2,8,balanced,0.3520127868652344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,2,8,balanced,0.5016665649414063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,2,8,balanced,0.09196031570434571
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,2,8,balanced,0.5666521453857423
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,2,8,balanced,0.7949644470214844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,1,16,power_law_1.2,0.4044300842285156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,2,8,balanced,0.9862425231933594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,2,8,balanced,0.03437439918518066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,2,8,balanced,0.03580800056457519
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,2,8,balanced,1.3461734008789061
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,2,8,balanced,0.04582015991210937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,2,8,balanced,0.06428671836853027
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,2,8,balanced,0.10850560188293457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,2,8,balanced,1.7843788146972657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,2,8,balanced,0.13036416053771974
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,2,8,balanced,0.13465472221374514
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,2,8,balanced,0.13622528076171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,2,8,balanced,0.14034048080444336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,2,8,balanced,0.14365951538085936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,2,8,balanced,0.13285759925842283
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,2,8,balanced,0.1357759952545166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,2,8,balanced,0.13771648406982423
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,2,8,balanced,0.14336128234863282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,2,8,balanced,0.14313088417053224
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,2,8,balanced,0.14751232147216797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,2,8,balanced,2.578236083984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,2,8,balanced,0.16569856643676756
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,2,8,balanced,0.185930233001709
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,2,8,balanced,0.20539392471313475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,2,8,balanced,0.2454118347167969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,2,8,balanced,0.28788864135742187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,2,8,balanced,0.400371208190918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,2,8,balanced,0.48316287994384766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,2,8,balanced,0.6889778900146484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,2,8,balanced,3.5504702758789066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,2,8,balanced,0.8861721801757814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,2,8,balanced,1.2441343688964843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,2,8,balanced,1.6483865356445313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,1,16,power_law_1.2,0.7759040069580079
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,1,16,power_law_1.2,0.5156851196289063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,1,16,power_law_1.2,1.0602432250976563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,2,8,balanced,0.022320640087127686
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,2,8,balanced,0.0673689603805542
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,2,8,balanced,0.06830080032348633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,2,8,balanced,0.06793727874755859
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,2,8,balanced,0.022751998901367188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,2,8,balanced,0.06830463886260987
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,2,8,balanced,0.02524415969848633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,2,8,balanced,0.0413324785232544
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,2,8,balanced,0.06837503910064698
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,2,8,balanced,0.06882559776306152
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,2,8,balanced,0.06950016021728515
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,2,8,balanced,0.07064191818237305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,2,8,balanced,0.07153664112091065
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,2,8,balanced,0.07343359947204589
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,2,8,balanced,0.0762943983078003
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,2,8,balanced,0.08060544013977052
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,2,8,balanced,0.060732159614562985
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,2,8,balanced,0.0685158395767212
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,2,8,balanced,0.06768640041351318
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,2,8,balanced,0.06709887981414794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,2,8,power_law_1.01,0.0902899169921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,2,8,power_law_1.01,0.14216320037841795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,2,8,power_law_1.01,0.11463040351867675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,2,8,power_law_1.01,0.14745087623596193
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,2,8,power_law_1.01,0.1622835159301758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,2,8,power_law_1.01,0.20221696853637697
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,2,8,power_law_1.01,0.20537343978881836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,2,8,power_law_1.01,0.2086988830566406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,2,8,power_law_1.01,0.2168230438232422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,2,8,power_law_1.01,0.2166220855712891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,2,8,power_law_1.01,0.22730880737304687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,2,8,balanced,0.06744063854217529
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,2,8,power_law_1.01,0.22148351669311522
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,2,8,power_law_1.01,0.22099584579467774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,2,8,power_law_1.01,0.23455871582031249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,2,8,power_law_1.01,0.26692352294921873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,2,8,power_law_1.01,0.2809087944030762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,2,8,power_law_1.01,0.31759616851806644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,2,8,power_law_1.01,0.373359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,2,8,power_law_1.01,0.43164928436279293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,2,8,power_law_1.01,0.5084198379516602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,2,8,power_law_1.01,0.6450521850585937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,2,8,power_law_1.01,0.8476211547851562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,2,8,balanced,0.10947072029113769
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,2,8,power_law_1.01,1.1286310577392578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,2,8,balanced,0.15441151618957522
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,2,8,balanced,0.19630079269409179
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,2,8,balanced,0.11766016006469726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,2,8,power_law_1.01,0.061607680320739745
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,2,8,power_law_1.01,0.08739328384399414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,2,8,balanced,0.25239679336547854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,2,8,power_law_1.01,0.08230015754699707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,2,8,power_law_1.01,0.08981247901916503
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,2,8,power_law_1.01,0.09478272438049316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,2,8,power_law_1.01,1.6066905212402343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,2,8,power_law_1.01,0.12087679862976075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,2,8,power_law_1.01,0.12043392181396484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,2,8,power_law_1.01,0.12428288459777832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,2,8,power_law_1.01,0.13531007766723632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,2,8,power_law_1.01,0.13341952323913575
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,2,8,power_law_1.01,0.13393407821655273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,2,8,power_law_1.01,0.14059136390686036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,2,8,power_law_1.01,0.13811200141906738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,2,8,power_law_1.01,0.133372163772583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,2,8,power_law_1.01,2.2583935546875002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,2,8,power_law_1.01,0.14098560333251953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,2,8,power_law_1.01,0.14326656341552735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,2,8,power_law_1.01,0.1517734432220459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,2,8,power_law_1.01,0.17854719161987304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,2,8,power_law_1.01,0.20739839553833006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,2,8,power_law_1.01,0.25084159851074217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,2,8,power_law_1.01,0.288033275604248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,2,8,power_law_1.01,0.38268543243408204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,2,8,power_law_1.01,0.5312537765502929
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,2,8,power_law_1.01,3.1810943603515627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,2,8,power_law_1.01,0.6974400329589844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,2,8,power_law_1.01,0.8822541046142579
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,2,8,power_law_1.01,0.05813504219055175
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,2,8,power_law_1.01,0.05059455871582032
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,2,8,power_law_1.01,0.05443840026855469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,2,8,power_law_1.01,0.06335103988647461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,2,8,power_law_1.01,3.739554443359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,2,8,power_law_1.01,0.0642751979827881
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,2,8,power_law_1.01,0.06598144054412843
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,2,8,power_law_1.01,0.06523903846740722
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,2,8,power_law_1.01,1.3916607666015626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,2,8,power_law_1.01,0.06597504138946533
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,2,8,power_law_1.01,0.06654079914093017
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,2,8,power_law_1.01,0.06598656177520752
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,2,8,power_law_1.01,0.06793983936309814
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,2,8,power_law_1.01,0.06820096015930176
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,2,8,power_law_1.01,0.06957824230194092
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,2,8,power_law_1.01,0.07082240104675293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,2,8,power_law_1.01,0.04348159790039062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,2,8,power_law_1.01,0.07280000209808349
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,2,8,power_law_1.01,1.8596377563476563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,2,8,balanced,0.33936256408691406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,2,8,power_law_1.01,0.09589632034301758
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,2,8,power_law_1.01,0.1784601593017578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,2,8,power_law_1.01,0.11098624229431153
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,2,8,power_law_1.01,0.08705408096313476
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,2,8,power_law_1.01,0.07559552192687988
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,2,8,power_law_1.01,0.23037567138671874
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,2,8,power_law_1.01,0.14243968009948732
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,2,8,power_law_1.2,0.09961983680725098
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,2,8,power_law_1.2,0.12287872314453126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,2,8,power_law_1.2,0.13418111801147461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,2,8,power_law_1.2,0.16529792785644531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,2,8,power_law_1.2,0.20128128051757815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,2,8,power_law_1.2,0.20160640716552733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,2,8,power_law_1.2,0.2126246452331543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,2,8,power_law_1.2,0.21279104232788085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,2,8,power_law_1.2,0.21474559783935546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,2,8,power_law_1.2,0.22489343643188478
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,2,8,power_law_1.2,0.09004799842834472
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,2,8,power_law_1.2,0.22204160690307617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,2,8,power_law_1.2,0.22712575912475588
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,2,8,power_law_1.2,0.2576780891418457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,2,8,power_law_1.2,0.2690380859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,2,8,power_law_1.2,0.30082815170288085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,2,8,power_law_1.2,0.3203327941894531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,2,8,power_law_1.2,0.3805055999755859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,2,8,power_law_1.2,0.4454399871826172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,2,8,power_law_1.2,0.5544473648071289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,2,8,power_law_1.2,0.6786624145507812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,2,8,power_law_1.2,0.9304537963867187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,2,8,power_law_1.2,1.192231674194336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,2,8,power_law_1.2,2.2456204223632814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,2,8,power_law_1.2,0.06151296138763428
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,2,8,power_law_1.2,0.06416639804840088
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,2,8,power_law_1.2,1.7350758361816407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,2,8,power_law_1.2,0.08098688125610351
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,2,8,power_law_1.2,3.6380224609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,2,8,power_law_1.2,0.10058752059936524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,2,8,power_law_1.2,0.11134592056274414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,2,8,power_law_1.2,0.12220800399780274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,2,8,power_law_1.2,0.12434559822082521
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,2,8,power_law_1.2,0.12767744064331055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,2,8,power_law_1.2,0.13468159675598143
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,2,8,power_law_1.2,0.13692031860351564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,2,8,power_law_1.2,0.1428275203704834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,2,8,power_law_1.2,0.13974143981933593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,2,8,power_law_1.2,0.13393664360046387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,2,8,power_law_1.2,0.14120448112487793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,2,8,power_law_1.2,0.1429708766937256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,2,8,power_law_1.2,0.1471116828918457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,2,8,power_law_1.2,0.15251968383789063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,2,8,power_law_1.2,0.18399232864379883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,2,8,power_law_1.2,0.21841152191162108
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,2,8,power_law_1.2,0.2779532814025879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,2,8,power_law_1.2,0.31564031600952147
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,2,8,power_law_1.2,0.4468646240234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,2,8,power_law_1.2,0.5959167861938477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,2,8,power_law_1.2,0.8196736145019532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,2,8,power_law_1.2,1.0035135650634766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,2,8,power_law_1.2,0.04322688102722168
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,2,8,power_law_1.01,0.3021183967590332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,2,8,power_law_1.2,0.04674431800842285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,2,8,power_law_1.2,0.05138944149017334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,2,8,power_law_1.2,0.061944317817687986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,2,8,power_law_1.2,1.5733197021484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,2,8,power_law_1.2,0.0625113582611084
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,2,8,power_law_1.2,0.06535168170928955
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,2,8,power_law_1.2,0.06425216197967529
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,2,8,power_law_1.2,0.06561535835266114
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,2,8,power_law_1.2,0.06551167964935303
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,2,8,power_law_1.2,0.06645887851715088
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,2,8,power_law_1.2,0.06692351818084716
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,2,8,power_law_1.2,0.06857344150543213
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,2,8,power_law_1.2,0.06826240062713623
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,2,8,power_law_1.2,0.0692416000366211
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,2,8,power_law_1.2,0.07709824085235596
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,2,8,power_law_1.2,2.059278106689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,2,8,power_law_1.2,4.484341735839844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,2,8,power_law_1.01,0.34684928894042966
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,2,8,power_law_1.01,0.5861017608642578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,2,8,power_law_1.01,0.8064934539794922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,2,8,power_law_1.2,0.08652288436889648
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,2,8,power_law_1.2,0.09266048431396484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,2,8,power_law_1.2,0.0734169578552246
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,2,8,power_law_1.2,0.1088319969177246
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,2,8,power_law_1.2,0.13031423568725586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,4,4,balanced,0.03418751955032349
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,4,4,balanced,0.05777664184570312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,4,4,balanced,0.07316607952117919
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,4,4,balanced,0.09304832458496094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,4,4,balanced,0.14883071899414063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,4,4,balanced,0.18172159194946289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,4,4,balanced,0.2093734359741211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,4,4,balanced,0.21279232025146486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,4,4,balanced,0.21540735244750975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,4,4,balanced,0.2199027252197266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,4,4,balanced,0.21401983261108398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,4,4,balanced,0.21760896682739256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,4,4,balanced,0.2221900749206543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,4,4,balanced,0.22914688110351564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,4,4,balanced,0.2327577590942383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,4,4,balanced,0.24116735458374022
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,4,4,balanced,0.25510784149169924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,4,4,balanced,0.2847001647949219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,4,4,balanced,0.31691392898559567
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,2,8,power_law_1.2,0.1773516845703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,4,4,balanced,0.4642483139038086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,2,8,power_law_1.2,0.2128281593322754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,4,4,balanced,0.522366714477539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,4,4,balanced,0.7337049865722657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,2,8,power_law_1.2,0.3265625762939453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,4,4,balanced,0.912642593383789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,4,4,balanced,1.2379788970947265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,4,4,balanced,0.03455104112625122
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,4,4,balanced,0.03709439992904663
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,4,4,balanced,1.6393792724609377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,4,4,balanced,0.05132031917572022
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,4,4,balanced,0.07247488021850586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,4,4,balanced,0.11503616333007813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,4,4,balanced,0.13108736038208008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,4,4,balanced,0.14484095573425293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,4,4,balanced,2.3687525939941407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,4,4,balanced,0.1481126403808594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,4,4,balanced,0.1495142364501953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,4,4,balanced,0.15213695526123047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,4,4,balanced,0.1429683208465576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,4,4,balanced,0.14584832191467284
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,4,4,balanced,0.1473689556121826
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,4,4,balanced,0.1519808006286621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,4,4,balanced,0.15296511650085448
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,4,4,balanced,0.15835519790649416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,4,4,balanced,3.1070233154296876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,4,4,balanced,0.17070592880249022
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,4,4,balanced,0.19003007888793946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,4,4,balanced,0.20464511871337893
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,4,4,balanced,0.2432076835632324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,4,4,balanced,0.28593408584594726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,4,4,balanced,0.3940518569946289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,4,4,balanced,0.47019134521484374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,4,4,balanced,0.6790169525146484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,4,4,balanced,0.8672844696044922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,4,4,balanced,1.201908493041992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,4,4,balanced,1.5918284606933593
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,2,8,power_law_1.2,0.4101145553588867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,2,8,power_law_1.2,0.4969369506835937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,2,8,power_law_1.2,0.8320588684082031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,4,4,balanced,0.025169920921325684
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,4,4,balanced,0.07606783866882325
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,4,4,balanced,0.026085119247436523
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,4,4,balanced,0.07696767807006835
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,4,4,balanced,0.07741568088531495
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,4,4,balanced,0.07790463924407959
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,4,4,balanced,0.07815296173095702
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,4,4,balanced,0.07904895782470703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,4,4,balanced,0.0801420783996582
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,4,4,balanced,0.08020223617553711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,4,4,balanced,0.04794367790222168
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,4,4,balanced,0.07673344135284424
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,4,4,balanced,0.08241791725158691
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,4,4,balanced,0.0845248031616211
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,4,4,balanced,0.08754560470581055
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,4,4,balanced,0.09363455772399902
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,4,4,balanced,0.09939583778381347
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,4,4,balanced,0.02994431972503662
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,2,8,power_law_1.2,1.2369395446777343
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,4,4,balanced,0.06867712020874024
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,4,4,balanced,0.07647999763488769
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,4,4,balanced,0.07651711940765381
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,4,4,power_law_1.01,0.07153791904449462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,4,4,power_law_1.01,0.08723711967468262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,4,4,power_law_1.01,0.0886464023590088
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,4,4,power_law_1.01,0.11426048278808594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,4,4,power_law_1.01,0.13682559967041014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,4,4,power_law_1.01,0.16694015502929688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,4,4,power_law_1.01,0.1972697639465332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,4,4,power_law_1.01,0.20095359802246096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,4,4,power_law_1.01,0.20672256469726563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,4,4,power_law_1.01,0.20317312240600588
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,4,4,power_law_1.01,0.21583871841430663
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,4,4,power_law_1.01,0.20628351211547852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,4,4,power_law_1.01,0.20767616271972655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,4,4,power_law_1.01,0.2224076843261719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,4,4,power_law_1.01,0.23140735626220704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,4,4,power_law_1.01,0.24503551483154298
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,4,4,power_law_1.01,0.2760204887390137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,4,4,power_law_1.01,0.3653363037109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,4,4,power_law_1.01,0.39735294342041017
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,4,4,power_law_1.01,0.4737241744995117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,4,4,power_law_1.01,0.5775859069824218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,4,4,power_law_1.01,0.7660800170898437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,4,4,power_law_1.01,1.0043673706054688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,4,4,power_law_1.01,1.370146484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,4,4,power_law_1.01,1.867653045654297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,4,4,power_law_1.01,2.614202880859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,4,4,power_law_1.01,0.04812032222747803
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,4,4,power_law_1.01,0.061835517883300786
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,4,4,power_law_1.01,0.06554111957550049
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,4,4,power_law_1.01,0.07316480159759522
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,4,4,power_law_1.01,0.0853593635559082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,4,4,power_law_1.01,3.711191101074219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,4,4,power_law_1.01,0.12080256462097168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,4,4,power_law_1.01,0.13981823921203612
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,4,4,power_law_1.01,0.14065919876098634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,4,4,power_law_1.01,0.14361472129821778
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,4,4,power_law_1.01,0.14578944206237793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,4,4,power_law_1.01,0.14918272018432616
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,4,4,power_law_1.01,0.14203776359558104
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,4,4,power_law_1.01,0.14417792320251466
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,4,4,power_law_1.01,0.14720767974853516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,4,4,power_law_1.01,0.14934528350830079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,4,4,power_law_1.01,0.15422847747802734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,4,4,power_law_1.01,0.16700927734374998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,4,4,power_law_1.01,0.1918720054626465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,4,4,power_law_1.01,0.21119615554809573
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,4,4,power_law_1.01,0.2744204711914063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,4,4,power_law_1.01,0.3117132759094238
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,4,4,power_law_1.01,0.40073089599609374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,4,4,power_law_1.01,0.5063948822021485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,4,4,power_law_1.01,0.7143103790283203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,4,4,power_law_1.01,0.8817011260986328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,4,4,power_law_1.01,1.300299530029297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,4,4,power_law_1.01,0.029410560131073
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,4,4,power_law_1.01,0.04337408065795899
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,4,4,power_law_1.01,1.6804595947265626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,4,4,power_law_1.01,0.046522879600524904
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,4,4,power_law_1.01,0.046337280273437496
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,4,4,power_law_1.01,0.05435008049011231
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,4,4,power_law_1.01,0.06941311836242677
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,4,4,balanced,0.21741567611694337
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,4,4,power_law_1.01,0.07085311889648438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,4,4,power_law_1.01,0.07344255924224855
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,4,4,power_law_1.01,0.07381120204925537
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,4,4,power_law_1.01,0.07377024173736572
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,4,4,power_law_1.01,0.07532544136047363
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,4,4,power_law_1.01,0.0760319995880127
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,4,4,power_law_1.01,0.07703551769256592
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,4,4,power_law_1.01,0.08137855529785157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,4,4,power_law_1.01,0.08349696159362793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,4,4,power_law_1.01,0.08584447860717773
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,4,4,balanced,0.14305919647216797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,4,4,balanced,0.15648256301879884
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,4,4,balanced,0.2764505577087403
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,4,4,balanced,0.3662771224975586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,4,4,balanced,0.48790527343749995
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,4,4,power_law_1.01,0.09177599906921387
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,4,4,power_law_1.01,0.1488499164581299
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,4,4,power_law_1.01,0.1250867176055908
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,4,4,power_law_1.01,0.10488191604614258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,4,4,power_law_1.2,0.07118847846984863
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,4,4,power_law_1.01,0.16810880661010744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,4,4,power_law_1.2,0.08197504043579101
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,4,4,power_law_1.2,0.09095680236816406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,4,4,power_law_1.2,0.11270912170410156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,4,4,power_law_1.2,0.13296256065368653
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,4,4,power_law_1.01,0.2401203155517578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,4,4,power_law_1.2,0.1788172721862793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,4,4,power_law_1.2,0.20187519073486326
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,4,4,power_law_1.01,0.29503231048583983
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,4,4,power_law_1.2,0.19501567840576173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,4,4,power_law_1.2,0.20155391693115235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,4,4,power_law_1.2,0.20803327560424806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,4,4,power_law_1.2,0.20861696243286132
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,4,4,power_law_1.2,0.20966272354125975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,4,4,power_law_1.2,0.2141209602355957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,4,4,power_law_1.2,0.23439615249633788
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,4,4,power_law_1.2,0.2467840003967285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,4,4,power_law_1.2,0.25500160217285156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,4,4,power_law_1.2,0.27876607894897465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,4,4,power_law_1.2,0.3594265747070312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,4,4,power_law_1.2,0.39991424560546873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,4,4,power_law_1.2,0.48588417053222654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,4,4,power_law_1.2,0.6060595321655273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,4,4,power_law_1.2,0.805713882446289
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,4,4,power_law_1.01,0.4422719955444336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,4,4,power_law_1.2,1.0360192108154298
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,4,4,power_law_1.2,0.04897280216217041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,4,4,power_law_1.2,0.05261568069458008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,4,4,power_law_1.2,1.4845811462402343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,4,4,power_law_1.2,0.0664908790588379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,4,4,power_law_1.2,0.08143232345581056
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,4,4,power_law_1.2,0.07606143951416015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,4,4,power_law_1.2,0.12080256462097168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,4,4,power_law_1.2,0.13697407722473146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,4,4,power_law_1.2,0.14072575569152831
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,4,4,power_law_1.2,2.9483328247070313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,4,4,power_law_1.2,0.1455795192718506
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,4,4,power_law_1.2,0.1462335968017578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,4,4,power_law_1.2,0.1500876808166504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,4,4,power_law_1.2,1.8628402709960938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,4,4,power_law_1.2,0.14223615646362303
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,4,4,power_law_1.2,0.1456281566619873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,4,4,power_law_1.2,0.15200639724731446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,4,4,power_law_1.2,0.15401344299316405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,4,4,power_law_1.2,0.15853311538696288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,4,4,power_law_1.2,0.17271039962768556
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,4,4,power_law_1.2,0.1935206413269043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,4,4,power_law_1.2,0.21702783584594726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,4,4,power_law_1.2,0.28254079818725586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,4,4,power_law_1.2,0.321952018737793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,4,4,power_law_1.2,0.4301043319702148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,4,4,power_law_1.2,0.5554931259155274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,4,4,power_law_1.2,0.7282598114013672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,4,4,power_law_1.2,0.0301529598236084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,4,4,power_law_1.2,0.9301055908203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,4,4,power_law_1.2,0.035069439411163326
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,4,4,power_law_1.2,0.053183999061584476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,4,4,power_law_1.2,1.424785919189453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,4,4,power_law_1.2,0.05507328033447265
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,4,4,power_law_1.2,0.06941311836242677
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,4,4,power_law_1.2,0.06924543857574464
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,4,4,power_law_1.2,0.07227519989013673
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,4,4,power_law_1.2,0.07319039821624755
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,4,4,power_law_1.2,0.07581823825836181
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,4,4,power_law_1.2,0.079137282371521
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,4,4,power_law_1.2,0.0790067195892334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,4,4,power_law_1.2,0.08341888427734376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,4,4,power_law_1.2,0.07382527828216554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,4,4,power_law_1.2,1.886360321044922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,4,4,power_law_1.2,3.7361227416992184
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,4,4,power_law_1.01,0.5650227355957032
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,4,4,power_law_1.01,0.8534976196289061
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,4,4,power_law_1.01,1.1033984374999999
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,4,4,power_law_1.2,0.04442240238189697
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,4,4,power_law_1.2,0.08652031898498536
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,4,4,power_law_1.2,0.08971136093139649
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,4,4,power_law_1.2,0.10311296463012695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,4,4,power_law_1.2,0.1154419231414795
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,4,4,power_law_1.2,0.13498880386352538
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,4,4,power_law_1.2,0.17105791091918945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,8,2,balanced,0.03221888065338135
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,4,4,power_law_1.2,0.23805952072143555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,8,2,balanced,0.05672832012176514
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,8,2,balanced,0.093438720703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,8,2,balanced,0.07147136211395264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,8,2,balanced,0.14909695625305175
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,8,2,balanced,0.17915264129638672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,8,2,balanced,0.18085887908935547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,8,2,balanced,0.18451967239379882
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,8,2,balanced,0.18599935531616213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,8,2,balanced,0.21276159286499025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,8,2,balanced,0.2152115249633789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,8,2,balanced,0.22081024169921876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,8,2,balanced,0.22223615646362305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,8,2,balanced,0.22833791732788086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,8,2,balanced,0.23280000686645508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,8,2,balanced,0.24016895294189453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,8,2,balanced,0.25349376678466795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,8,2,balanced,0.28296319961547856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,8,2,balanced,0.3129804801940918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,8,2,balanced,0.45015296936035154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,8,2,balanced,0.5054041671752929
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,8,2,balanced,0.7091404724121093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,8,2,balanced,0.8859980773925781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,8,2,balanced,1.1911961364746095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,8,2,balanced,1.5717669677734374
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,4,4,power_law_1.2,0.295548152923584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,8,2,balanced,2.2659059143066407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,8,2,balanced,0.03417855978012085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,8,2,balanced,0.0381657600402832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,8,2,balanced,0.052761597633361815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,8,2,balanced,0.08031104087829591
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,8,2,balanced,0.13444735527038573
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,8,2,balanced,0.14986111640930175
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,8,2,balanced,0.15154687881469725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,8,2,balanced,0.1518131160736084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,8,2,balanced,0.1541324806213379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,8,2,balanced,0.17683456420898438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,8,2,balanced,0.17781503677368166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,8,2,balanced,2.9747955322265627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,8,2,balanced,0.18085376739501952
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,8,2,balanced,0.18229631423950193
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,8,2,balanced,0.18584064483642576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,8,2,balanced,0.18529024124145507
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,8,2,balanced,0.1860825538635254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,8,2,balanced,0.19506559371948243
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,4,4,power_law_1.2,0.40304512023925787
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,8,2,balanced,0.21675647735595702
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,8,2,balanced,0.23374847412109373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,8,2,balanced,0.26036224365234373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,8,2,balanced,0.3010572814941406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,8,2,balanced,0.4568921661376953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,8,2,balanced,0.524142074584961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,8,2,balanced,0.7327155303955079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,8,2,balanced,0.9587238311767579
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,4,4,power_law_1.2,0.6511654663085937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,8,2,balanced,1.3220761108398438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,8,2,balanced,1.7537843322753908
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,4,4,power_law_1.2,0.9728205108642578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,4,4,power_law_1.2,1.3234136962890626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,4,4,power_law_1.2,1.785233917236328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,8,2,balanced,0.03496959924697876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,8,2,balanced,0.03343487977981567
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,8,2,balanced,0.09790207862854003
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,8,2,balanced,0.0990937614440918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,8,2,balanced,0.09967103958129883
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,8,2,balanced,0.060058879852294925
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,8,2,balanced,0.10051584243774414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,8,2,balanced,0.04174208164215088
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,8,2,balanced,0.10213631629943846
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,8,2,balanced,0.10270208358764647
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,8,2,balanced,0.08634112358093263
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,8,2,balanced,0.10429823875427247
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,8,2,balanced,0.10807295799255372
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,8,2,balanced,0.1126579189300537
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,8,2,balanced,0.09854207992553711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,8,2,balanced,0.11753600120544434
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,8,2,balanced,0.12814208030700686
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,8,2,balanced,0.13812352180480958
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,8,2,balanced,0.09749504089355468
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,8,2,balanced,0.09739263534545899
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,8,2,power_law_1.01,0.0570636796951294
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,8,2,power_law_1.01,0.07127935886383056
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,8,2,power_law_1.01,0.07817471981048583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,8,2,power_law_1.01,0.09943807601928711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,8,2,power_law_1.01,0.12157952308654785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,8,2,power_law_1.01,0.15747712135314942
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,8,2,power_law_1.01,0.15801471710205078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,8,2,power_law_1.01,0.16415487289428712
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,8,2,power_law_1.01,0.17118207931518553
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,8,2,power_law_1.01,0.20062847137451173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,8,2,power_law_1.01,0.21030656814575197
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,8,2,power_law_1.01,0.2124403190612793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,8,2,power_law_1.01,0.21393407821655272
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,8,2,power_law_1.01,0.22562816619873044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,8,2,power_law_1.01,0.2307967948913574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,8,2,power_law_1.01,0.2409868812561035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,8,2,power_law_1.01,0.2615603256225586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,8,2,power_law_1.01,0.34222976684570316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,8,2,power_law_1.01,0.3916902542114258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,8,2,power_law_1.01,0.45382911682128907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,8,2,power_law_1.01,0.5469043350219727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,8,2,power_law_1.01,0.7288703918457031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,8,2,power_law_1.01,0.9132377624511718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,8,2,power_law_1.01,1.2881701660156248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,8,2,power_law_1.01,1.6724134826660155
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,8,2,balanced,0.0977996826171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,8,2,power_law_1.01,2.395878448486328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,8,2,balanced,0.20724735260009766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,8,2,balanced,0.23039232254028322
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,8,2,power_law_1.01,0.04077311992645264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,8,2,power_law_1.01,0.04940415859222412
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,8,2,balanced,0.5639360046386719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,8,2,power_law_1.01,0.06060800075531005
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,8,2,balanced,0.3264550399780274
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,8,2,balanced,0.4186521530151367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,8,2,power_law_1.01,0.07794047832489014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,8,2,power_law_1.01,0.13225215911865235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,8,2,power_law_1.01,0.09290623664855957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,8,2,power_law_1.01,0.13875840187072755
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,8,2,power_law_1.01,0.14386303901672365
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,8,2,power_law_1.01,0.14538368225097656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,8,2,power_law_1.01,0.16780799865722656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,8,2,power_law_1.01,0.17061504364013672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,8,2,power_law_1.01,0.1749875259399414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,8,2,power_law_1.01,0.17868032455444335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,8,2,power_law_1.01,0.18306047439575196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,8,2,power_law_1.01,0.18290559768676756
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,8,2,power_law_1.01,0.18527103424072267
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,8,2,power_law_1.01,0.19736831665039062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,8,2,power_law_1.01,0.22013439178466795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,8,2,power_law_1.01,0.24076927185058591
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,8,2,power_law_1.01,0.302740478515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,8,2,power_law_1.01,0.34799617767333985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,8,2,power_law_1.01,0.45598720550537114
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,8,2,power_law_1.01,0.5716492843627929
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,8,2,power_law_1.01,0.03545216083526611
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,8,2,power_law_1.01,0.04203392028808593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,8,2,power_law_1.01,3.1720538330078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,8,2,power_law_1.01,0.04618624210357666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,8,2,power_law_1.01,0.7730726623535157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,8,2,power_law_1.01,0.05903488159179687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,8,2,power_law_1.01,0.9783731079101562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,8,2,power_law_1.01,0.08510080337524414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,8,2,power_law_1.01,0.08903552055358886
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,8,2,balanced,0.7577715301513672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,8,2,power_law_1.01,0.089617919921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,8,2,power_law_1.01,0.09286144256591797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,8,2,power_law_1.01,0.09448575973510742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,8,2,power_law_1.01,0.09636608123779297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,8,2,power_law_1.01,0.09898624420166016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,8,2,power_law_1.01,1.418329620361328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,8,2,power_law_1.01,0.10050047874450682
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,8,2,power_law_1.01,0.10452223777770997
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,8,2,power_law_1.01,0.1085580825805664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,8,2,power_law_1.01,0.11339776039123535
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,8,2,power_law_1.01,0.15152640342712403
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,8,2,power_law_1.01,0.16835199356079103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,8,2,power_law_1.01,1.8111871337890626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,8,2,power_law_1.01,0.07180799961090087
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,8,2,power_law_1.01,0.2648908805847168
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,8,2,power_law_1.01,0.12655360221862794
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,8,2,power_law_1.01,0.21862016677856447
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,8,2,power_law_1.2,0.05642240047454834
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,8,2,power_law_1.01,0.4406105422973633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,8,2,power_law_1.01,0.642814712524414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,8,2,power_law_1.2,0.07808512210845947
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,8,2,power_law_1.2,0.065765118598938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,8,2,power_law_1.2,0.08869759559631347
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,8,2,power_law_1.2,0.10299519538879394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,8,2,power_law_1.2,0.16023551940917968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,8,2,power_law_1.2,0.15999487876892088
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,8,2,power_law_1.2,0.17402368545532226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,8,2,power_law_1.2,0.20276863098144532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,8,2,power_law_1.2,0.2060799980163574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,8,2,power_law_1.2,0.21119743347167969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,8,2,power_law_1.2,0.21646207809448242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,8,2,power_law_1.01,1.3749452209472657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,8,2,power_law_1.01,0.37518463134765623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,8,2,power_law_1.2,0.22841600418090818
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,8,2,power_law_1.2,0.23384960174560546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,8,2,power_law_1.2,0.2415616035461426
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,8,2,power_law_1.01,0.8428953552246095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,8,2,power_law_1.2,0.2766694450378418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,8,2,power_law_1.2,0.35867263793945314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,8,2,power_law_1.2,0.39159423828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,8,2,power_law_1.2,0.46220928192138666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,8,2,power_law_1.2,0.5584588623046874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,8,2,power_law_1.2,0.7363212585449219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,8,2,power_law_1.2,0.16153472900390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,8,2,power_law_1.2,0.04059135913848877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,8,2,power_law_1.2,0.04821504116058349
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,8,2,power_law_1.2,0.06090879917144776
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,8,2,power_law_1.2,0.0774783992767334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,8,2,power_law_1.2,0.08996864318847655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,8,2,power_law_1.2,1.7306918334960937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,8,2,power_law_1.2,0.9254528045654297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,8,2,power_law_1.2,0.13617024421691895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,8,2,power_law_1.2,0.1395353603363037
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,8,2,power_law_1.2,0.14283391952514649
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,8,2,power_law_1.2,0.14637951850891112
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,8,2,power_law_1.2,0.17118335723876954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,8,2,power_law_1.2,1.3067173767089844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,8,2,power_law_1.2,0.1742361640930176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,8,2,power_law_1.2,0.17805440902709962
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,8,2,power_law_1.2,0.18080127716064454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,8,2,power_law_1.2,0.18557695388793943
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,8,2,power_law_1.2,0.1839411163330078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,8,2,power_law_1.2,2.455036163330078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,8,2,power_law_1.2,0.18804096221923827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,8,2,power_law_1.2,0.20220672607421877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,8,2,power_law_1.2,0.22390527725219728
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,8,2,power_law_1.2,0.2594009590148926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,8,2,power_law_1.2,0.30610687255859376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,8,2,power_law_1.2,0.36044288635253907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,8,2,power_law_1.2,0.4621683120727539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,8,2,power_law_1.2,0.5804569625854492
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,8,2,power_law_1.2,0.035828480720520015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,8,2,power_law_1.2,0.04038015842437744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,8,2,power_law_1.2,1.0026892852783202
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,8,2,power_law_1.2,0.045966081619262696
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,8,2,power_law_1.2,0.055125761032104495
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,8,2,power_law_1.2,0.7854386901855468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,8,2,power_law_1.2,1.4577958679199219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,8,2,power_law_1.2,0.0627558422088623
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,8,2,power_law_1.2,0.08415360450744629
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,8,2,power_law_1.2,0.09009535789489746
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,8,2,power_law_1.2,0.09256447792053223
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,8,2,power_law_1.2,0.09314175605773925
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,8,2,power_law_1.2,0.09457792282104492
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,8,2,power_law_1.2,0.09777791976928711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,8,2,power_law_1.2,0.10018048286437989
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,8,2,power_law_1.2,0.10414079666137695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,8,2,power_law_1.2,0.10854016304016115
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,8,2,power_law_1.2,0.11400959968566896
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,8,2,power_law_1.2,0.1209881591796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,8,2,power_law_1.2,0.16618751525878905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,8,2,power_law_1.2,3.247818298339844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,8,2,power_law_1.2,1.87617919921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,8,2,power_law_1.01,1.8053823852539062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,8,2,power_law_1.2,0.1376959991455078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,8,2,power_law_1.2,0.19260671615600586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,8,2,power_law_1.2,0.31755647659301756
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,8,2,power_law_1.2,0.2714803123474121
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,8,2,power_law_1.2,0.5380198287963867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,16,1,balanced,0.0702233600616455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,16,1,balanced,0.08856896400451661
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,16,1,balanced,0.11639488220214844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,16,1,balanced,0.180447998046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,16,1,balanced,0.22054719924926758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,16,1,balanced,0.22238655090332032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,16,1,balanced,0.22374784469604495
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,16,1,balanced,0.04994495868682861
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,8,2,power_law_1.2,0.639477767944336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,16,1,balanced,0.22859264373779298
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,16,1,balanced,0.23536895751953124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,16,1,balanced,0.2825164794921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,16,1,balanced,0.2223360061645508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,16,1,balanced,0.2863500785827637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,16,1,balanced,0.29412288665771485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,16,1,balanced,0.2956902313232422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,8,2,power_law_1.2,1.0347122955322265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,16,1,balanced,0.30950847625732425
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,16,1,balanced,0.3406496047973633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,16,1,balanced,0.3633932876586914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,16,1,balanced,0.23200000762939452
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,16,1,balanced,0.5137369537353516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,16,1,balanced,0.5667225646972656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,16,1,balanced,0.7787232208251953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,16,1,balanced,0.9580754852294922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,16,1,balanced,0.045920000076293946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,16,1,balanced,0.05255551815032959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,16,1,balanced,1.2721651458740235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,16,1,balanced,0.07258111953735351
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,16,1,balanced,0.10972543716430665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,16,1,balanced,0.1857651138305664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,16,1,balanced,0.21564416885375975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,16,1,balanced,1.6667225646972657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,16,1,balanced,0.2165273666381836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,16,1,balanced,0.21974720001220702
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,16,1,balanced,0.22068799972534178
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,16,1,balanced,0.22090623855590819
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,16,1,balanced,0.22471872329711914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,16,1,balanced,0.2277395248413086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,16,1,balanced,2.379239044189453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,16,1,balanced,0.27949567794799807
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,16,1,balanced,0.2834144020080566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,16,1,balanced,0.2827199935913086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,16,1,balanced,0.28608575820922855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,16,1,balanced,0.2950943946838379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,16,1,balanced,0.31130815505981446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,16,1,balanced,0.32534912109374997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,16,1,balanced,0.3609535980224609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,16,1,balanced,0.39346752166748045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,16,1,balanced,3.10621826171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,16,1,balanced,0.6090681457519531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,16,1,balanced,0.6733606719970704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,16,1,balanced,0.9586630249023438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,16,1,balanced,1.246402587890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,16,1,balanced,1.668876190185547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,16,1,balanced,2.235308837890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,8,2,power_law_1.2,1.2135501098632813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,8,2,power_law_1.2,2.9474777221679687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,8,2,power_law_1.2,2.100537567138672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,16,1,balanced,0.05048831939697266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,16,1,balanced,0.05399807929992676
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,16,1,balanced,0.1571833610534668
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,16,1,balanced,0.06482048034667968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,16,1,balanced,0.08796159744262695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,16,1,balanced,0.16080896377563475
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,16,1,balanced,0.15727871894836426
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,16,1,balanced,0.15868224143981932
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,16,1,balanced,0.1301459217071533
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,16,1,balanced,0.16524415969848633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,16,1,balanced,0.16333887100219727
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,16,1,balanced,0.16586175918579102
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,16,1,balanced,0.16972991943359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,16,1,balanced,0.17742528915405273
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,16,1,balanced,0.18177087783813478
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,16,1,balanced,0.1917087936401367
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,16,1,balanced,0.21450815200805665
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,16,1,balanced,0.23843776702880862
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,16,1,balanced,0.15591360092163087
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,16,1,balanced,0.15703743934631348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,16,1,power_law_1.01,0.045528960227966306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,16,1,power_law_1.01,0.07138175964355468
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,16,1,balanced,0.15938112258911133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,16,1,power_law_1.01,0.08865216255187988
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,16,1,power_law_1.01,0.10749055862426757
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,16,1,power_law_1.01,0.13179583549499513
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,16,1,power_law_1.01,0.18602304458618163
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,16,1,power_law_1.01,0.1940390396118164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,16,1,power_law_1.01,0.20181568145751952
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,16,1,power_law_1.01,0.20601343154907226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,16,1,power_law_1.01,0.21719999313354493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,16,1,power_law_1.01,0.22743104934692382
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,16,1,power_law_1.01,0.25576128005981447
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,16,1,power_law_1.01,0.2715897560119629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,16,1,power_law_1.01,0.28244672775268553
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,16,1,power_law_1.01,0.2862873649597168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,16,1,power_law_1.01,0.2932953643798828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,16,1,power_law_1.01,0.31449600219726564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,16,1,power_law_1.01,0.37148609161376955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,16,1,power_law_1.01,0.4484735870361328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,16,1,power_law_1.01,0.5113548660278321
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,16,1,power_law_1.01,0.5945644760131836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,16,1,balanced,0.4053209686279297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,16,1,balanced,0.36052928924560546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,16,1,balanced,0.5804032135009766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,16,1,power_law_1.01,0.782279052734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,16,1,balanced,0.745409927368164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,16,1,power_law_1.01,0.05075456142425537
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,16,1,power_law_1.01,0.9557817840576173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,16,1,power_law_1.01,0.05620287895202637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,16,1,power_law_1.01,0.07514944076538085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,16,1,power_law_1.01,1.3163052368164063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,16,1,balanced,1.0198834991455077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,16,1,power_law_1.01,0.09940223693847657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,16,1,power_law_1.01,0.13134400367736815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,16,1,power_law_1.01,1.6823385620117186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,16,1,power_law_1.01,0.19150272369384766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,16,1,power_law_1.01,0.19672063827514646
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,16,1,power_law_1.01,0.2020479965209961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,16,1,power_law_1.01,0.2056083106994629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,16,1,power_law_1.01,0.20797376632690429
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,16,1,power_law_1.01,0.21364927291870117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,16,1,power_law_1.01,0.21906303405761718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,16,1,power_law_1.01,0.2569273567199707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,16,1,power_law_1.01,0.27235200881958005
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,16,1,power_law_1.01,0.2724985694885254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,16,1,power_law_1.01,0.27816320419311524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,16,1,power_law_1.01,2.405364532470703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,16,1,power_law_1.01,0.2927577590942383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,16,1,power_law_1.01,0.31865152359008786
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,16,1,power_law_1.01,0.34019454956054684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,16,1,power_law_1.01,0.39203136444091796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,16,1,power_law_1.01,0.4714668655395508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,16,1,power_law_1.01,0.5989132690429687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,16,1,power_law_1.01,0.7135072326660156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,16,1,power_law_1.01,3.1264743041992187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,16,1,power_law_1.01,0.9682150268554688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,16,1,power_law_1.01,0.05135551929473877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,16,1,power_law_1.01,1.2192447662353516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,16,1,power_law_1.01,0.057955198287963874
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,16,1,power_law_1.01,0.06598976135253906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,16,1,power_law_1.01,0.08123392105102539
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,16,1,balanced,1.3587110900878907
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,16,1,power_law_1.01,0.09362688064575195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,16,1,power_law_1.01,1.715235137939453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,16,1,power_law_1.01,0.1294553565979004
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,16,1,power_law_1.01,0.1398406410217285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,16,1,power_law_1.01,0.14545472145080568
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,16,1,power_law_1.01,2.2336729431152342
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,16,1,power_law_1.01,0.1471884822845459
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,16,1,power_law_1.01,0.1579520034790039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,16,1,power_law_1.01,0.15276864051818848
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,16,1,power_law_1.01,0.16448320388793944
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,16,1,power_law_1.01,0.1745913505554199
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,16,1,power_law_1.01,0.13634943962097168
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,16,1,power_law_1.01,0.18272512435913085
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,16,1,power_law_1.01,0.18882623672485352
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,16,1,power_law_1.01,0.20957632064819337
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,16,1,power_law_1.01,0.29357376098632815
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,16,1,power_law_1.01,0.2516646385192871
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,16,1,power_law_1.01,0.4769740676879882
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,16,1,power_law_1.01,0.37602111816406253
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,16,1,power_law_1.2,0.045170559883117675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,16,1,power_law_1.2,0.07196928024291992
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,16,1,power_law_1.01,0.8576070404052734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,16,1,power_law_1.01,0.6574540710449218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,16,1,power_law_1.2,0.08735296249389649
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,16,1,power_law_1.2,0.18390207290649413
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,16,1,power_law_1.01,1.2591353607177733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,16,1,power_law_1.2,0.1948249626159668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,16,1,power_law_1.2,0.20206655502319334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,16,1,power_law_1.2,0.2141663932800293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,16,1,power_law_1.2,0.22188352584838866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,16,1,power_law_1.2,0.23630464553833005
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,16,1,power_law_1.2,0.10278592109680176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,16,1,power_law_1.2,0.26761087417602536
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,16,1,power_law_1.2,0.27585344314575194
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,16,1,power_law_1.2,0.12556672096252441
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,16,1,power_law_1.2,0.2823814392089844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,16,1,power_law_1.2,0.2889894485473633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,16,1,power_law_1.2,0.298670711517334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,16,1,power_law_1.2,0.3203168106079102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,16,1,power_law_1.2,0.3753081512451172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,16,1,power_law_1.2,0.448287353515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,16,1,power_law_1.2,0.5156825637817383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,16,1,power_law_1.2,0.5986099243164062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,16,1,power_law_1.2,0.7866445159912109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,16,1,power_law_1.2,0.960040283203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,16,1,power_law_1.2,0.047641601562500004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,16,1,power_law_1.2,0.05988863945007324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,16,1,power_law_1.2,1.3231347656249999
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,16,1,power_law_1.2,0.07397632122039795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,16,1,power_law_1.2,0.09382719993591308
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,16,1,power_law_1.2,0.12692288398742674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,16,1,power_law_1.2,0.19054719924926758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,16,1,power_law_1.2,0.19561855316162108
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,16,1,power_law_1.2,1.6839801025390624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,16,1,power_law_1.2,0.20206079483032227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,16,1,power_law_1.2,0.20749120712280272
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,16,1,power_law_1.2,0.2115519905090332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,16,1,power_law_1.2,0.216628475189209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,16,1,power_law_1.2,0.2204979133605957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,16,1,power_law_1.2,0.27357120513916017
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,16,1,power_law_1.2,0.27719680786132816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,16,1,power_law_1.2,0.2815795135498047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,16,1,power_law_1.2,0.28940288543701176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,16,1,power_law_1.2,0.3019865608215332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,16,1,power_law_1.2,0.3217728042602539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,16,1,power_law_1.2,0.3482592010498047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,16,1,power_law_1.2,0.4186291122436524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,16,1,power_law_1.2,0.47645118713378903
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,16,1,power_law_1.2,0.608202896118164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,16,1,power_law_1.2,0.7318284606933594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,16,1,power_law_1.2,0.9745536041259765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,16,1,power_law_1.2,1.2236390686035157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,16,1,power_law_1.2,2.410695037841797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,16,1,power_law_1.2,0.05171135902404785
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,16,1,power_law_1.2,0.057796478271484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,16,1,power_law_1.2,0.06592703819274902
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,16,1,power_law_1.2,0.07520576000213623
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,16,1,power_law_1.2,0.09111488342285157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,16,1,power_law_1.2,0.1284895992279053
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,16,1,power_law_1.2,0.13389951705932618
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,16,1,power_law_1.2,0.1441958427429199
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,16,1,power_law_1.2,0.14625856399536133
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,16,1,power_law_1.2,0.15330559730529786
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,16,1,power_law_1.2,0.1596217632293701
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,16,1,power_law_1.2,1.7250962829589844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,16,1,power_law_1.2,0.17105472564697266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,16,1,power_law_1.2,0.16502464294433594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,16,1,power_law_1.2,0.18634559631347655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,16,1,power_law_1.2,3.144200439453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,16,1,power_law_1.01,2.52845947265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,16,1,power_law_1.01,3.397107849121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,16,1,power_law_1.2,2.243388824462891
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,16,1,power_law_1.2,0.2105632019042969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,16,1,power_law_1.2,0.19795391082763672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,16,1,power_law_1.2,0.23535615921020508
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,16,1,power_law_1.2,0.29203136444091793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,16,1,power_law_1.2,0.35377025604248047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,16,1,power_law_1.2,0.48525566101074225
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,16,1,power_law_1.2,0.6291219329833984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,16,balanced,0.06863872051239014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,16,balanced,0.06843647956848145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,16,balanced,0.07981567859649659
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,16,balanced,0.09715456008911133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,16,balanced,0.14541055679321288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,16,balanced,0.2988019180297852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,16,balanced,0.38816638946533205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,16,balanced,0.4529510498046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,16,balanced,0.45646846771240235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,16,balanced,0.4577484893798828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,16,balanced,0.4631206512451172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,16,balanced,0.46908031463623046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,16,balanced,0.47130622863769533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,16,balanced,0.4686604690551758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,16,balanced,0.47638526916503904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,16,balanced,0.4792051315307617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,16,balanced,0.49290496826171876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,16,balanced,0.5148185729980469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,16,balanced,0.5388531112670899
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,16,balanced,0.5808512115478516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,16,balanced,0.6243417739868165
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,16,1,power_law_1.2,0.9310387420654298
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,16,balanced,0.7008000183105468
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,16,1,power_law_1.2,1.2584121704101563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,16,balanced,0.9734041595458984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,16,balanced,1.110412826538086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,16,balanced,1.4951437377929688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,16,balanced,0.042398719787597655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,16,balanced,0.042389760017395015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,16,balanced,1.9806565856933596
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,16,1,power_law_1.2,1.9639334106445312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,16,balanced,0.04644608020782471
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,16,balanced,0.06325376033782959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,16,balanced,0.08788864135742187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,16,balanced,0.1508249568939209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,16,balanced,0.21273984909057617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,16,balanced,0.23007360458374024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,16,balanced,0.23170688629150388
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,16,balanced,0.23177600860595704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,16,balanced,0.2361996841430664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,16,balanced,2.878309020996094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,16,balanced,0.2389324760437012
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,16,balanced,0.23629695892333985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,16,balanced,0.23209600448608397
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,16,balanced,0.2413836860656738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,16,balanced,0.23441152572631835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,16,balanced,0.23761152267456054
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,16,balanced,0.2488153648376465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,16,balanced,0.274334716796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,16,balanced,0.30359296798706054
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,16,balanced,0.32023681640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,16,balanced,0.3698905563354492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,16,balanced,0.42028030395507815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,16,balanced,0.5224588775634766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,16,balanced,0.7087615966796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,16,1,power_law_1.2,2.6977389526367186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,16,balanced,0.8872998046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,16,balanced,1.1818201446533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,16,1,power_law_1.2,4.160956115722657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,16,1,power_law_1.2,5.650264892578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,1,16,balanced,0.02758271932601929
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,1,16,balanced,0.024138240814208983
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,1,16,balanced,0.03560703992843628
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,1,16,balanced,0.023671040534973146
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,1,16,balanced,0.12531840324401855
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,1,16,balanced,0.12546688079833984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,1,16,balanced,0.12658687591552736
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,1,16,balanced,0.12698495864868165
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,1,16,balanced,0.12626175880432128
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,1,16,balanced,0.12737664222717285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,1,16,balanced,0.1279142379760742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,1,16,balanced,0.12797951698303223
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,1,16,balanced,0.13005824089050294
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,1,16,balanced,0.1306175994873047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,1,16,balanced,0.05481984138488769
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,1,16,balanced,0.13278976440429688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,1,16,balanced,0.12501248359680175
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,1,16,balanced,0.13400959968566895
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,1,16,balanced,0.13927295684814453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,1,16,balanced,0.14349568367004395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,1,16,balanced,0.15160320281982423
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,1,16,balanced,0.1276262378692627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,16,power_law_1.01,0.1447500801086426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,16,power_law_1.01,0.2915660858154297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,16,power_law_1.01,0.38177406311035156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,16,power_law_1.01,0.3738905715942383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,16,power_law_1.01,0.3852275085449219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,16,power_law_1.01,0.3878847885131836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,16,power_law_1.01,0.4048115158081055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,16,power_law_1.01,0.41741184234619144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,16,power_law_1.01,0.41045246124267576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,16,power_law_1.01,0.4143398284912109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,16,power_law_1.01,0.4324313735961914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,16,power_law_1.01,0.43214080810546873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,16,power_law_1.01,0.44644096374511716
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,16,power_law_1.01,0.4378841781616211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,16,power_law_1.01,0.44891521453857425
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,16,power_law_1.01,0.460711669921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,16,power_law_1.01,0.4794636917114258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,16,power_law_1.01,0.5085260772705078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,16,power_law_1.01,0.5807897567749023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,16,power_law_1.01,0.6968601226806641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,16,power_law_1.01,0.8005248260498046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,16,power_law_1.01,1.0080255889892578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,1,16,balanced,0.12040703773498536
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,16,power_law_1.01,1.2763801574707032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,16,power_law_1.01,1.658146514892578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,16,power_law_1.01,2.032471008300781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,16,power_law_1.01,0.08630399703979492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,16,power_law_1.01,0.14868351936340332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,16,power_law_1.01,0.20662656784057618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,16,power_law_1.01,3.413480834960937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,16,power_law_1.01,0.19877248764038086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,16,power_law_1.01,0.210064640045166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,16,power_law_1.01,0.21516672134399414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,16,power_law_1.01,0.2060633659362793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,16,power_law_1.01,0.22280319213867186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,16,power_law_1.01,3.9623410034179685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,16,power_law_1.01,0.21875583648681643
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,16,power_law_1.01,0.20468992233276367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,16,power_law_1.01,0.21722496032714841
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,16,power_law_1.01,0.22367103576660155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,16,power_law_1.01,0.21651968002319336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,16,power_law_1.01,0.21542400360107422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,16,power_law_1.01,0.2189094352722168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,16,power_law_1.01,0.22213375091552737
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,16,power_law_1.01,0.2430847930908203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,16,power_law_1.01,0.2615398406982422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,16,power_law_1.01,0.29970304489135746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,16,power_law_1.01,0.35572864532470705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,16,power_law_1.01,0.43016193389892576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,16,power_law_1.01,0.5149977493286133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,16,power_law_1.01,0.6753855895996094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,16,power_law_1.01,0.8785420989990234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,1,16,power_law_1.01,0.05667200088500977
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,1,16,power_law_1.01,0.09099136352539063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,16,power_law_1.01,1.2072383880615234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,1,16,power_law_1.01,0.12508416175842285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,1,16,power_law_1.01,0.11788031578063965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,16,power_law_1.01,1.5394854736328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,1,16,power_law_1.01,0.12546815872192382
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,1,16,power_law_1.01,0.12147968292236329
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,1,16,power_law_1.01,0.11450752258300781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,1,16,power_law_1.01,0.12383999824523925
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,1,16,power_law_1.01,0.10712832450866698
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,1,16,power_law_1.01,0.10728192329406738
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,1,16,power_law_1.01,0.10634623527526857
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,1,16,power_law_1.01,0.10829440116882325
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,1,16,power_law_1.01,0.10911871910095214
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,1,16,power_law_1.01,0.11071231842041016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,1,16,power_law_1.01,0.12315263748168945
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,1,16,power_law_1.01,0.11859583854675293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,1,16,power_law_1.01,0.12504447937011717
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,1,16,power_law_1.01,0.12551039695739746
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,1,16,balanced,0.16632064819335937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,1,16,balanced,0.22594816207885743
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,1,16,balanced,0.3005491256713867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,16,power_law_1.01,0.19701120376586914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,1,16,balanced,0.20475648880004882
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,1,16,power_law_1.01,0.14711551666259765
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,1,16,power_law_1.01,0.15976192474365233
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,16,power_law_1.2,0.14730751991271973
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,1,16,power_law_1.01,0.12786687850952147
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,16,power_law_1.2,0.2943756866455078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,16,power_law_1.2,0.36162303924560546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,1,16,power_law_1.01,0.18595327377319334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,16,power_law_1.2,0.37415935516357424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,16,power_law_1.2,0.38489471435546874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,16,power_law_1.2,0.45007873535156245
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,16,power_law_1.2,0.38516864776611326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,16,power_law_1.2,0.4013478469848633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,16,power_law_1.2,0.40245632171630863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,16,power_law_1.2,0.4041190338134766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,16,power_law_1.2,0.4169919967651367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,16,power_law_1.2,0.43194625854492186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,16,power_law_1.2,0.44206977844238277
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,16,power_law_1.2,0.43533184051513674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,16,power_law_1.2,0.45179775238037106
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,16,power_law_1.2,0.46189311981201175
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,16,power_law_1.2,0.4737715148925782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,16,power_law_1.2,0.5398015975952148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,16,power_law_1.2,0.6159526443481445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,16,power_law_1.2,0.7368409729003906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,16,power_law_1.2,0.8184870147705079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,16,power_law_1.2,1.1307724761962892
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,16,power_law_1.2,1.4354969787597658
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,1,16,power_law_1.01,0.21564159393310547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,16,power_law_1.2,2.049759979248047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,16,power_law_1.2,0.08514176368713379
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,1,16,power_law_1.01,0.29026687622070313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,16,power_law_1.2,0.14632448196411132
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,16,power_law_1.2,2.417769012451172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,16,power_law_1.2,0.195993595123291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,16,power_law_1.2,0.19933183670043947
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,16,power_law_1.2,0.2091929626464844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,16,power_law_1.2,0.21196672439575198
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,16,power_law_1.2,0.20277631759643552
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,16,power_law_1.2,0.21871999740600584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,1,16,power_law_1.01,0.4932441711425781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,16,power_law_1.2,0.20115711212158205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,16,power_law_1.2,0.22089727401733397
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,16,power_law_1.2,0.21671936035156253
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,16,power_law_1.2,0.21361663818359372
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,16,power_law_1.2,0.206693115234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,16,power_law_1.2,0.20412031173706055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,16,power_law_1.2,3.629255676269531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,16,power_law_1.2,0.21716352462768557
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,16,power_law_1.2,0.21852287292480468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,16,power_law_1.2,0.22368127822875977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,16,power_law_1.2,0.2529740715026855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,16,power_law_1.2,0.2830067253112793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,16,power_law_1.2,0.3399628829956055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,16,power_law_1.2,0.37911167144775393
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,16,power_law_1.2,0.4819456100463867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,16,power_law_1.2,0.5904793548583984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,1,16,power_law_1.2,0.0562175989151001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,16,power_law_1.2,0.824352035522461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,16,power_law_1.2,4.583969421386719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,1,16,power_law_1.2,0.09123456001281738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,16,power_law_1.2,1.0425676727294921
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,1,16,power_law_1.2,0.11457152366638183
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,1,16,power_law_1.2,0.11976832389831542
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,1,16,power_law_1.01,0.34010623931884765
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,1,16,power_law_1.2,0.12444031715393067
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,1,16,power_law_1.2,0.12374527931213379
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,1,16,power_law_1.2,0.10184320449829101
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,16,power_law_1.2,1.629505310058594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,1,16,power_law_1.2,0.10066304206848145
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,1,16,power_law_1.2,0.10325887680053711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,1,16,power_law_1.2,0.10614656448364257
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,1,16,power_law_1.2,0.11294591903686524
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,1,16,power_law_1.2,0.10771072387695313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,16,power_law_1.2,1.9327488708496094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,1,16,power_law_1.2,0.10967040061950684
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,1,16,power_law_1.2,0.12131584167480469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,1,16,power_law_1.2,0.11972736358642577
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,1,16,power_law_1.2,0.12294272422790527
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,1,16,power_law_1.01,0.6117401504516602
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,1,16,power_law_1.2,0.10708992004394531
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,1,16,power_law_1.2,0.13782912254333496
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,1,16,power_law_1.2,0.16477184295654296
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,1,16,power_law_1.2,0.1274227237701416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,8,balanced,0.04038656234741211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,8,balanced,0.06451712131500245
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,1,16,power_law_1.2,0.22349952697753905
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,1,16,power_law_1.2,0.17142656326293945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,8,balanced,0.10078463554382325
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,1,16,power_law_1.2,0.2788057518005371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,8,balanced,0.1461094379425049
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,8,balanced,0.30549247741699215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,8,balanced,0.3921356964111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,8,balanced,0.4496371078491211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,8,balanced,0.45423744201660154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,8,balanced,0.07677440166473389
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,8,balanced,0.4575513458251953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,8,balanced,0.4539519882202148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,8,balanced,0.4609587097167969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,8,balanced,0.4629350280761718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,8,balanced,0.47796352386474605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,8,balanced,0.482165756225586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,8,balanced,0.49315582275390624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,8,balanced,0.5088959884643555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,8,balanced,0.5487654495239258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,8,balanced,0.5802841567993164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,8,balanced,0.6462464141845703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,8,balanced,0.7140684509277344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,8,balanced,0.8587980651855469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,8,balanced,1.132140808105469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,8,balanced,1.4323814392089844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,8,balanced,0.038227200508117676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,8,balanced,0.040721921920776366
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,8,balanced,1.8872050476074218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,8,balanced,0.05335807800292969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,8,balanced,0.07320831775665283
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,8,balanced,0.10207360267639161
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,8,balanced,0.16211967468261718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,8,balanced,0.22303871154785154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,8,balanced,0.24094463348388673
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,8,balanced,2.6291903686523437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,8,balanced,0.24342016220092771
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,8,balanced,0.2466854476928711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,8,balanced,0.23789567947387696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,8,balanced,0.24164480209350586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,8,balanced,0.2426777648925781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,8,balanced,0.2479385566711426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,8,balanced,0.2525107192993164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,8,balanced,0.25900672912597655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,8,balanced,0.27839616775512693
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,8,balanced,0.2969817543029785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,8,balanced,0.3194649505615234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,8,balanced,3.8301516723632814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,8,balanced,0.3628761672973633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,8,balanced,0.3984576034545898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,8,balanced,0.4854604721069336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,8,balanced,0.5753523254394531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,8,balanced,0.7291788482666016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,8,balanced,1.0003263854980469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,8,balanced,1.337154541015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,8,balanced,1.7569587707519532
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,1,16,power_law_1.2,0.37563903808593746
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,1,16,power_law_1.2,0.5497663879394532
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,1,16,power_law_1.2,0.7366259002685547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,1,16,power_law_1.2,0.9321113586425781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,2,8,balanced,0.025955839157104494
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,2,8,balanced,0.13373184204101562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,2,8,balanced,0.13387264251708983
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,2,8,balanced,0.02714751958847046
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,2,8,balanced,0.13369215965270995
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,2,8,balanced,0.13363712310791015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,2,8,balanced,0.13409791946411134
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,2,8,balanced,0.1342732810974121
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,2,8,balanced,0.1349516773223877
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,2,8,balanced,0.1356710433959961
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,2,8,balanced,0.13714176177978515
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,2,8,balanced,0.13902848243713378
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,2,8,balanced,0.030108160972595215
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,2,8,balanced,0.14155903816223145
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,2,8,balanced,0.14423935890197753
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,2,8,balanced,0.03783168077468872
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,2,8,balanced,0.14711039543151855
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,2,8,balanced,0.15589247703552248
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,2,8,balanced,0.1616473579406738
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,2,8,balanced,0.17557376861572266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,8,power_law_1.01,0.09819904327392578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,8,power_law_1.01,0.14073984146118163
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,8,power_law_1.01,0.29693824768066407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,2,8,balanced,0.05828479766845703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,8,power_law_1.01,0.2298726463317871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,8,power_law_1.01,0.31840511322021486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,8,power_law_1.01,0.36825984954833985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,8,power_law_1.01,0.38456321716308595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,8,power_law_1.01,0.4304166412353515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,8,power_law_1.01,0.43303039550781247
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,8,power_law_1.01,0.415973129272461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,8,power_law_1.01,0.42819839477539057
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,8,power_law_1.01,0.4068940734863281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,8,power_law_1.01,0.4146521759033203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,8,power_law_1.01,0.4262335968017578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,8,power_law_1.01,0.44535934448242187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,8,power_law_1.01,0.4484902572631836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,8,power_law_1.01,0.48255615234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,8,power_law_1.01,0.5081727981567383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,8,power_law_1.01,0.5631513595581055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,8,power_law_1.01,0.668853759765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,8,power_law_1.01,0.7602803039550781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,8,power_law_1.01,0.907031021118164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,8,power_law_1.01,1.128106231689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,8,power_law_1.01,1.5544627380371092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,8,power_law_1.01,1.9018162536621095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,8,power_law_1.01,2.8878988647460937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,8,power_law_1.01,0.06163712024688721
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,8,power_law_1.01,0.08959232330322266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,2,8,balanced,0.09481727600097656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,8,power_law_1.01,0.15119359970092774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,8,power_law_1.01,3.6497805786132815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,8,power_law_1.01,0.17302528381347657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,8,power_law_1.01,0.20336639404296877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,8,power_law_1.01,0.18175104141235351
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,2,8,balanced,0.13320704460144045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,8,power_law_1.01,0.14459775924682616
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,8,power_law_1.01,0.20569471359252928
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,8,power_law_1.01,0.2214784049987793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,8,power_law_1.01,0.2022092819213867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,8,power_law_1.01,0.2345305633544922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,8,power_law_1.01,0.22533632278442384
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,8,power_law_1.01,0.2080191993713379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,8,power_law_1.01,0.22054655075073243
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,2,8,balanced,0.13286272048950196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,8,power_law_1.01,0.22940927505493164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,8,power_law_1.01,0.23129087448120117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,8,power_law_1.01,0.25298688888549803
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,8,power_law_1.01,0.27218687057495117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,8,power_law_1.01,0.2896601676940918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,8,power_law_1.01,0.3406758499145508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,8,power_law_1.01,0.3682457733154297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,8,power_law_1.01,0.4797119903564453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,8,power_law_1.01,0.5319283294677735
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,2,8,power_law_1.01,0.04235648155212403
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,8,power_law_1.01,0.7376525115966797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,8,power_law_1.01,0.8911039733886719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,2,8,power_law_1.01,0.06046207904815674
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,2,8,power_law_1.01,0.0973964786529541
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,2,8,power_law_1.01,0.08556032180786133
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,2,8,power_law_1.01,0.10279040336608887
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,2,8,power_law_1.01,0.1187929630279541
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,2,8,power_law_1.01,0.1144435214996338
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,2,8,power_law_1.01,0.1109990406036377
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,2,8,power_law_1.01,0.11073408126831055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,8,power_law_1.01,1.7685043334960937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,2,8,power_law_1.01,0.10765567779541016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,2,8,power_law_1.01,0.11235456466674805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,2,8,power_law_1.01,0.11155584335327148
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,2,8,power_law_1.01,0.11203583717346191
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,8,power_law_1.01,1.2960166931152344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,2,8,power_law_1.01,0.11843199729919433
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,2,8,power_law_1.01,0.13601920127868652
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,2,8,power_law_1.01,0.11410688400268554
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,2,8,power_law_1.01,0.1133631992340088
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,2,8,power_law_1.01,0.13287296295166015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,2,8,balanced,0.23842304229736327
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,2,8,balanced,0.2731379127502441
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,2,8,balanced,0.3643097686767578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,2,8,power_law_1.01,0.14524160385131837
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,2,8,power_law_1.01,0.15548800468444823
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,8,power_law_1.2,0.0976460838317871
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,2,8,power_law_1.01,0.1814259147644043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,8,power_law_1.2,0.13948543548583986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,8,power_law_1.2,0.20537984848022459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,8,power_law_1.2,0.23186815261840818
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,2,8,power_law_1.01,0.2092812728881836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,8,power_law_1.2,0.32457344055175785
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,2,8,power_law_1.01,0.27025407791137696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,8,power_law_1.2,0.3794073486328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,8,power_law_1.2,0.3792972946166992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,8,power_law_1.2,0.429939193725586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,8,power_law_1.2,0.4040486526489258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,8,power_law_1.2,0.396497917175293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,8,power_law_1.2,0.39322368621826176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,8,power_law_1.2,0.39708160400390624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,8,power_law_1.2,0.4268825531005859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,8,power_law_1.2,0.41202560424804685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,8,power_law_1.2,0.4463641738891601
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,8,power_law_1.2,0.46408321380615236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,8,power_law_1.2,0.5033689498901367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,8,power_law_1.2,0.43582977294921876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,8,power_law_1.2,0.615101432800293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,8,power_law_1.2,0.6976882934570312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,2,8,power_law_1.01,0.33229694366455076
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,8,power_law_1.2,0.7735679626464844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,8,power_law_1.2,1.0735897827148437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,2,8,power_law_1.01,0.4301055908203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,8,power_law_1.2,1.2400985717773438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,8,power_law_1.2,0.06237696170806885
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,8,power_law_1.2,0.0901030445098877
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,2,8,power_law_1.01,0.6370124816894531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,8,power_law_1.2,1.7284364318847658
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,8,power_law_1.2,0.12764160156249998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,8,power_law_1.2,0.14000767707824707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,8,power_law_1.2,0.18199424743652343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,8,power_law_1.2,2.2106597900390623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,8,power_law_1.2,0.18042879104614257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,8,power_law_1.2,0.21078784942626952
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,8,power_law_1.2,0.20990207672119138
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,8,power_law_1.2,0.22383615493774417
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,8,power_law_1.2,0.20982656478881836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,8,power_law_1.2,0.21843072891235354
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,8,power_law_1.2,0.21140863418579103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,8,power_law_1.2,0.21055744171142576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,8,power_law_1.2,0.21041151046752932
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,8,power_law_1.2,0.20696960449218751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,8,power_law_1.2,3.073388671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,8,power_law_1.2,0.2088217544555664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,8,power_law_1.2,0.2540326309204102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,8,power_law_1.2,0.26767616271972655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,8,power_law_1.2,0.28649215698242186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,8,power_law_1.2,0.35248512268066406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,8,power_law_1.2,0.39261825561523434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,8,power_law_1.2,0.5013888168334961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,8,power_law_1.2,0.5825164794921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,2,8,power_law_1.2,0.04253312110900879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,8,power_law_1.2,0.793702392578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,2,8,power_law_1.2,0.060568318367004395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,2,8,power_law_1.2,0.07512191772460938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,2,8,power_law_1.2,0.08502911567687989
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,8,power_law_1.2,0.993222427368164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,2,8,power_law_1.2,0.10157055854797363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,8,power_law_1.2,4.383815612792969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,2,8,power_law_1.2,0.10459263801574707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,8,power_law_1.2,1.3514764404296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,2,8,power_law_1.2,0.10344832420349122
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,2,8,power_law_1.2,0.1131328010559082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,2,8,power_law_1.2,0.10591232299804687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,2,8,power_law_1.2,0.11671039581298828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,2,8,power_law_1.2,0.11032575607299804
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,2,8,power_law_1.2,0.11502719879150392
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,2,8,power_law_1.2,0.11698304176330568
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,2,8,power_law_1.2,0.12352255821228027
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,2,8,power_law_1.2,0.12399744033813476
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,2,8,power_law_1.2,0.13559807777404784
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,2,8,power_law_1.2,0.14428671836853027
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,8,power_law_1.2,1.984620819091797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,2,8,power_law_1.2,0.1124505615234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,2,8,power_law_1.01,0.7820390319824219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,2,8,power_law_1.2,0.1828339195251465
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,2,8,power_law_1.2,0.4961433410644531
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,2,8,power_law_1.2,0.2195699119567871
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,2,8,power_law_1.2,0.1596006393432617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,4,balanced,0.06160128116607666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,4,balanced,0.07541759967803954
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,2,8,power_law_1.2,0.2837516784667969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,4,4,balanced,0.034942719936370846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,4,balanced,0.14426752090454104
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,4,balanced,0.30031103134155274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,4,balanced,0.3864499282836914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,4,balanced,0.3880716705322266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,2,8,power_law_1.2,0.6825369262695313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,4,balanced,0.38937854766845703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,4,balanced,0.39174144744873046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,4,balanced,0.09598976135253906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,4,balanced,0.445617904663086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,2,8,power_law_1.2,0.33852928161621093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,4,balanced,0.4512128067016602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,4,balanced,0.45638912200927734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,4,balanced,0.46271358489990233
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,4,balanced,0.4687155151367188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,4,balanced,0.47645694732666016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,4,balanced,0.4941875076293945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,4,balanced,0.528326416015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,4,balanced,0.5636953735351563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,4,balanced,0.6271308898925781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,4,balanced,0.6967078399658203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,4,balanced,0.8310809326171874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,4,balanced,0.0354585599899292
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,4,balanced,1.087831039428711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,4,balanced,0.03951616048812866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,4,balanced,0.05171711921691895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,4,balanced,0.07817728042602538
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,4,balanced,0.11725312232971193
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,4,balanced,0.1784716796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,4,balanced,0.24297088623046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,4,balanced,1.7909210205078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,4,balanced,0.24550527572631836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,4,balanced,0.2474150466918945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,4,balanced,0.24906751632690433
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,4,balanced,0.2761292839050293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,4,balanced,2.4907852172851563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,4,balanced,1.3647564697265626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,4,balanced,0.279422721862793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,4,balanced,0.28254720687866214
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,4,balanced,0.28871295928955076
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,4,balanced,0.29523712158203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,4,balanced,0.30527231216430667
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,4,balanced,0.3303731155395508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,4,balanced,0.3454028701782227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,4,balanced,3.5861569213867184
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,4,balanced,0.3869427108764648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,4,balanced,0.4310451126098632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,4,balanced,0.28724351882934573
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,4,balanced,0.4985331344604492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,4,balanced,0.5839667129516601
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,4,balanced,0.7386675262451172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,4,balanced,1.0118873596191407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,4,balanced,1.3218803405761719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,4,balanced,1.7442034912109374
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,2,8,power_law_1.2,0.9697689819335938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,2,8,power_law_1.2,1.2302745819091796
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,4,4,balanced,0.03548288106918335
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,4,4,balanced,0.15602687835693357
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,4,4,balanced,0.03923455953598022
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,4,4,balanced,0.03375744104385376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,4,4,balanced,0.15638527870178223
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,4,4,balanced,0.04964863777160645
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,4,4,balanced,0.15602560043334962
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,4,4,balanced,0.07180928230285645
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,4,4,balanced,0.15736960411071776
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,4,4,balanced,0.15717503547668457
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,4,4,balanced,0.1606835174560547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,4,4,balanced,0.1587942409515381
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,4,4,balanced,0.15838591575622557
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,4,4,balanced,0.16232959747314452
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,4,4,balanced,0.16373760223388673
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,4,4,balanced,0.1695577621459961
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,4,4,balanced,0.1740287971496582
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,4,4,balanced,0.1810201644897461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,4,4,balanced,0.19167104721069336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,4,4,balanced,0.20322816848754882
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,4,4,balanced,0.22708736419677736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,4,4,power_law_1.01,0.07430016040802002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,4,power_law_1.01,0.09372799873352052
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,4,4,balanced,0.11140607833862304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,4,power_law_1.01,0.13646847724914551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,4,power_law_1.01,0.14262784004211426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,4,power_law_1.01,0.23345663070678713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,4,power_law_1.01,0.2464499282836914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,4,power_law_1.01,0.2965555191040039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,4,power_law_1.01,0.3534604644775391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,4,power_law_1.01,0.3568972778320313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,4,power_law_1.01,0.35956478118896484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,4,power_law_1.01,0.3929164886474609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,4,power_law_1.01,0.42214015960693363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,4,power_law_1.01,0.4196006393432617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,4,power_law_1.01,0.42903934478759764
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,4,power_law_1.01,0.4377727890014649
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,4,power_law_1.01,0.4512089538574219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,4,power_law_1.01,0.4678860855102539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,4,power_law_1.01,0.5030783843994141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,4,power_law_1.01,0.5620223999023437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,4,power_law_1.01,0.6839564514160157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,4,power_law_1.01,0.7499110412597656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,4,power_law_1.01,0.9158898925781249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,4,power_law_1.01,1.0865484619140626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,4,power_law_1.01,1.50052734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,4,power_law_1.01,1.86134521484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,4,power_law_1.01,0.04434688091278076
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,4,power_law_1.01,0.06643968105316163
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,4,power_law_1.01,0.09688192367553711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,4,power_law_1.01,2.685759887695313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,4,power_law_1.01,0.10713727951049805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,4,power_law_1.01,0.15458175659179688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,4,power_law_1.01,0.16738048553466797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,4,4,balanced,0.15631360054016114
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,4,power_law_1.01,0.1761612892150879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,4,power_law_1.01,0.2115711975097656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,4,power_law_1.01,3.484346923828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,4,power_law_1.01,0.21613311767578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,4,power_law_1.01,0.19702783584594727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,4,power_law_1.01,0.23319936752319337
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,4,power_law_1.01,0.2310028839111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,4,power_law_1.01,0.2395315170288086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,4,power_law_1.01,0.2525760078430176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,4,power_law_1.01,0.26059263229370117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,4,power_law_1.01,0.2727142333984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,4,power_law_1.01,0.29858175277709964
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,4,power_law_1.01,0.28775552749633787
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,4,power_law_1.01,0.32629760742187497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,4,power_law_1.01,0.37788543701171873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,4,power_law_1.01,0.4056921768188476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,4,power_law_1.01,0.5159628677368164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,4,power_law_1.01,0.620294418334961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,4,power_law_1.01,0.7906610870361328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,4,power_law_1.01,0.9753279876708983
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,4,4,power_law_1.01,0.03990911960601807
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,4,power_law_1.01,1.4409829711914062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,4,4,power_law_1.01,0.053680639266967776
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,4,4,power_law_1.01,0.0749183988571167
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,4,power_law_1.01,1.841361846923828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,4,4,power_law_1.01,0.09967616081237793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,4,4,power_law_1.01,0.10163711547851562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,4,4,power_law_1.01,0.11966591835021974
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,4,4,power_law_1.01,0.14546688079833986
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,4,4,power_law_1.01,0.07743231773376465
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,4,4,power_law_1.01,0.13359359741210936
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,4,4,power_law_1.01,0.13711104393005372
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,4,4,power_law_1.01,0.13393664360046387
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,4,4,power_law_1.01,0.12303872108459472
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,4,4,power_law_1.01,0.13187456130981445
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,4,4,power_law_1.01,0.1341926383972168
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,4,4,power_law_1.01,0.1384332847595215
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,4,4,power_law_1.01,0.1452902412414551
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,4,4,power_law_1.01,0.15819519996643067
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,4,4,power_law_1.01,0.12558719635009766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,4,4,balanced,0.15614208221435547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,4,4,balanced,0.3670387268066406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,4,4,balanced,0.31416959762573243
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,4,4,balanced,0.4940582275390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,4,4,power_law_1.01,0.1701568031311035
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,4,4,power_law_1.01,0.20035327911376952
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,4,4,power_law_1.01,0.23118976593017576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,4,4,power_law_1.2,0.0748083209991455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,4,power_law_1.2,0.09371007919311523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,4,power_law_1.2,0.11351424217224122
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,4,power_law_1.2,0.14669440269470216
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,4,4,power_law_1.01,0.30066944122314454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,4,power_law_1.2,0.18442239761352539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,4,power_law_1.2,0.2675315284729004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,4,power_law_1.2,0.2625868797302246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,4,power_law_1.2,0.3460262298583984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,4,power_law_1.2,0.36083072662353516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,4,power_law_1.2,0.35491966247558593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,4,power_law_1.2,0.3849049758911133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,4,power_law_1.2,0.404186897277832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,4,power_law_1.2,0.4188108825683594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,4,power_law_1.2,0.4300211334228516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,4,power_law_1.2,0.4402803039550781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,4,power_law_1.2,0.44854270935058593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,4,power_law_1.2,0.46700672149658207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,4,power_law_1.2,0.5000396728515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,4,power_law_1.2,0.5707942581176757
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,4,4,power_law_1.01,0.3311360168457031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,4,power_law_1.2,0.6930828857421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,4,power_law_1.2,0.7568870544433594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,4,power_law_1.2,0.9520832061767578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,4,power_law_1.2,1.1663398742675781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,4,power_law_1.2,1.5718476867675781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,4,power_law_1.2,0.044337921142578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,4,power_law_1.2,0.06626304149627685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,4,power_law_1.2,1.9387174987792968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,4,power_law_1.2,0.08111488342285156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,4,power_law_1.2,0.1092454433441162
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,4,power_law_1.2,0.15003775596618651
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,4,power_law_1.2,0.16972288131713867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,4,power_law_1.2,2.7837274169921877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,4,power_law_1.2,0.16751743316650392
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,4,power_law_1.2,0.2198271942138672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,4,power_law_1.2,0.22514432907104492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,4,power_law_1.2,0.20605695724487302
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,4,power_law_1.2,0.21617664337158202
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,4,power_law_1.2,0.22542463302612306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,4,power_law_1.2,0.23563007354736326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,4,power_law_1.2,3.5663027954101567
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,4,power_law_1.2,0.24668031692504883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,4,power_law_1.2,0.26074752807617185
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,4,power_law_1.2,0.2680243110656738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,4,power_law_1.2,0.2829209518432617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,4,power_law_1.2,0.30757631301879884
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,4,4,power_law_1.01,0.5268390274047852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,4,power_law_1.2,0.33022590637207033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,4,power_law_1.2,0.3801356887817383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,4,power_law_1.2,0.4404684829711914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,4,power_law_1.2,0.5387609481811524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,4,power_law_1.2,0.6329894256591797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,4,power_law_1.2,0.8319590759277343
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,4,4,power_law_1.2,0.0386406397819519
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,4,power_law_1.2,1.0579468536376955
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,4,4,power_law_1.2,0.05392384052276611
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,4,4,power_law_1.2,0.06876416206359863
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,4,4,power_law_1.2,0.07115520000457763
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,4,4,power_law_1.01,0.6263872146606445
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,4,4,power_law_1.2,0.0951795196533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,4,power_law_1.2,1.440326385498047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,4,4,power_law_1.2,0.10880000114440917
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,4,4,power_law_1.2,0.11616640090942383
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,4,4,power_law_1.01,0.9162560272216798
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,4,4,power_law_1.2,0.1260108757019043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,4,4,power_law_1.2,0.13965312004089356
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,4,4,power_law_1.2,0.1256435203552246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,4,power_law_1.2,1.9099609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,4,4,power_law_1.2,0.142041597366333
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,4,4,power_law_1.2,0.13087488174438477
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,4,4,power_law_1.2,0.14214143753051758
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,4,4,power_law_1.2,0.13722496032714843
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,4,4,power_law_1.2,0.1322150421142578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,4,4,power_law_1.2,0.15934080123901367
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,4,4,power_law_1.2,0.18026111602783204
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,4,4,power_law_1.2,0.14924287796020508
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,4,4,power_law_1.01,1.1976882934570312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,4,4,power_law_1.2,0.24158208847045898
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,4,4,power_law_1.2,0.2826905632019043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,4,4,power_law_1.2,0.3869696044921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,4,4,power_law_1.2,0.2000230407714844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,8,2,balanced,0.037465600967407225
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,4,4,power_law_1.2,0.454595832824707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,8,2,balanced,0.06159103870391845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,8,2,balanced,0.07540863990783692
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,8,2,balanced,0.09796735763549805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,8,2,balanced,0.1513113594055176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,8,2,balanced,0.31012479782104496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,8,2,balanced,0.40153854370117187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,8,2,balanced,0.4028134536743164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,8,2,balanced,0.4043059158325195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,8,2,balanced,0.4046387100219727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,8,2,balanced,0.4074726486206055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,8,2,balanced,0.4117939376831055
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,4,4,power_law_1.2,0.6377830505371094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,8,2,balanced,0.4152140808105469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,8,2,balanced,0.49910911560058596
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,8,2,balanced,0.5045939254760742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,8,2,balanced,0.512350730895996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,8,2,balanced,0.52740478515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,8,2,balanced,0.5570086288452148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,8,2,balanced,0.5881036758422852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,8,2,balanced,0.6481407928466797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,8,2,balanced,0.7121663665771484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,8,2,balanced,0.842287368774414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,4,4,power_law_1.2,0.9002137756347656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,8,2,balanced,0.03953151941299439
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,8,2,balanced,1.1164825439453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,8,2,balanced,0.04073855876922607
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,8,2,balanced,0.056523518562316896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,8,2,balanced,1.372103729248047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,8,2,balanced,0.08448384284973146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,8,2,balanced,0.13737728118896483
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,8,2,balanced,0.23399423599243163
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,8,2,balanced,0.3131724739074707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,8,2,balanced,0.3145561599731445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,8,2,balanced,1.788179168701172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,8,2,balanced,0.31664384841918947
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,8,2,balanced,0.318525447845459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,8,2,balanced,0.3225139236450195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,8,2,balanced,0.3257625579833984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,8,2,balanced,0.3289766311645508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,8,2,balanced,0.37401088714599606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,8,2,balanced,2.4690010070800783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,8,2,balanced,0.3746368026733399
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,8,2,balanced,0.38026622772216795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,8,2,balanced,0.3873331069946289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,8,2,balanced,0.40376449584960933
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,8,2,balanced,0.4239616012573242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,8,2,balanced,0.45934719085693365
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,8,2,balanced,0.4980518341064453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,8,2,balanced,0.5678950500488281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,8,2,balanced,0.6368742370605469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,8,2,balanced,3.5785882568359377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,8,2,balanced,0.7846387481689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,8,2,balanced,1.1293682861328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,8,2,balanced,1.4255795288085937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,8,2,balanced,1.8895103454589843
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,4,4,power_law_1.2,1.472481231689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,4,4,power_law_1.2,1.9565632629394532
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,8,2,balanced,0.0517516803741455
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,8,2,balanced,0.1992870330810547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,8,2,balanced,0.1996940803527832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,8,2,balanced,0.053521919250488284
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,8,2,balanced,0.05932032108306885
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,8,2,balanced,0.0724441623687744
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,8,2,balanced,0.20090112686157227
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,8,2,balanced,0.20236032485961913
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,8,2,balanced,0.20280704498291016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,8,2,balanced,0.20543872833251955
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,8,2,balanced,0.20581375122070314
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,8,2,balanced,0.2072870445251465
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,8,2,balanced,0.21056255340576172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,8,2,balanced,0.2160563278198242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,8,2,balanced,0.22191999435424803
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,8,2,balanced,0.23374208450317382
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,8,2,balanced,0.10002559661865233
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,8,2,balanced,0.26976512908935546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,8,2,balanced,0.2944960021972656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,8,2,balanced,0.14969728469848634
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,8,2,balanced,0.34583934783935544
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,8,2,power_law_1.01,0.05997056007385254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,8,2,power_law_1.01,0.07540095806121826
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,8,2,power_law_1.01,0.09600128173828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,8,2,power_law_1.01,0.11167103767395019
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,8,2,balanced,0.24569728851318356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,8,2,power_law_1.01,0.15855615615844726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,8,2,power_law_1.01,0.21016960144042968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,8,2,power_law_1.01,0.24981504440307617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,8,2,power_law_1.01,0.3376192092895508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,8,2,power_law_1.01,0.3404185485839844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,8,2,power_law_1.01,0.3533375930786133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,8,2,power_law_1.01,0.3693363189697266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,8,2,power_law_1.01,0.3586943817138672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,8,2,power_law_1.01,0.3662911987304688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,8,2,power_law_1.01,0.47323776245117183
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,8,2,power_law_1.01,0.48057472229003906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,8,2,power_law_1.01,0.48973056793212894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,8,2,power_law_1.01,0.509251823425293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,8,2,power_law_1.01,0.5394713592529297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,8,2,power_law_1.01,0.5860185623168945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,8,2,power_law_1.01,0.7195903778076171
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,8,2,power_law_1.01,0.7884467315673829
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,8,2,power_law_1.01,0.9296307373046876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,8,2,power_law_1.01,1.137991714477539
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,8,2,balanced,0.19966207504272462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,8,2,power_law_1.01,1.4933772277832031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,8,2,power_law_1.01,0.03993983983993531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,8,2,power_law_1.01,0.051569919586181644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,8,2,power_law_1.01,1.87001220703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,8,2,power_law_1.01,0.07661312103271485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,8,2,power_law_1.01,0.1476211166381836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,8,2,power_law_1.01,2.6489944458007812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,8,2,power_law_1.01,0.20492927551269532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,8,2,power_law_1.01,0.09417599678039551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,8,2,power_law_1.01,0.21542272567749024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,8,2,power_law_1.01,0.25142656326293944
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,8,2,power_law_1.01,3.397199401855469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,8,2,power_law_1.01,0.2743987274169922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,8,2,power_law_1.01,0.2681305694580078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,8,2,power_law_1.01,0.28628608703613284
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,8,2,power_law_1.01,0.2780697631835937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,8,2,power_law_1.01,0.2934169578552246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,8,2,power_law_1.01,0.33399936676025394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,8,2,power_law_1.01,0.33746559143066407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,8,2,power_law_1.01,0.3432998275756836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,8,2,power_law_1.01,0.3498278427124023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,8,2,power_law_1.01,0.3690867233276367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,8,2,power_law_1.01,0.3979084777832031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,8,2,power_law_1.01,0.44328063964843756
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,8,2,power_law_1.01,0.4988876724243164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,8,2,power_law_1.01,0.6020492935180665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,8,2,power_law_1.01,0.6996236419677734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,8,2,power_law_1.01,0.9093798065185548
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,8,2,power_law_1.01,0.05442815780639648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,8,2,power_law_1.01,1.1275302124023436
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,8,2,power_law_1.01,0.06124415874481202
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,8,2,power_law_1.01,0.0782489585876465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,8,2,power_law_1.01,1.5317861938476562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,8,2,power_law_1.01,0.0853593635559082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,8,2,power_law_1.01,0.1019711971282959
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,8,2,power_law_1.01,0.12444031715393067
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,8,2,power_law_1.01,0.1356172752380371
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,8,2,power_law_1.01,0.15753472328186036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,8,2,power_law_1.01,1.9645184326171876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,8,2,power_law_1.01,0.16163328170776367
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,8,2,power_law_1.01,0.16901760101318358
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,8,2,power_law_1.01,0.17502464294433592
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,8,2,power_law_1.01,0.17356416702270508
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,8,2,power_law_1.01,0.17805952072143555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,8,2,power_law_1.01,0.18379135131835939
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,8,2,power_law_1.01,0.19080703735351562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,8,2,power_law_1.01,0.19230464935302732
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,8,2,power_law_1.01,0.2278387260437012
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,8,2,power_law_1.01,0.20515968322753908
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,8,2,balanced,0.19893247604370118
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,8,2,balanced,0.8193766021728516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,8,2,balanced,0.5030988693237305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,8,2,balanced,0.6114380645751953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,8,2,power_law_1.01,0.25441919326782225
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,8,2,power_law_1.01,0.32969600677490235
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,8,2,power_law_1.01,0.28899839401245114
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,8,2,power_law_1.2,0.06163328170776368
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,8,2,power_law_1.2,0.07472896099090576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,8,2,power_law_1.2,0.09004416465759277
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,8,2,power_law_1.2,0.11061247825622558
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,8,2,power_law_1.2,0.1478809642791748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,8,2,power_law_1.2,0.19419776916503906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,8,2,power_law_1.2,0.22214143753051757
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,8,2,power_law_1.01,0.46546558380126957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,8,2,power_law_1.2,0.3245414352416992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,8,2,power_law_1.2,0.34564224243164066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,8,2,power_law_1.2,0.3252838516235351
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,8,2,power_law_1.2,0.3535923385620117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,8,2,power_law_1.2,0.3651839828491211
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,8,2,power_law_1.01,0.5367795181274414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,8,2,power_law_1.2,0.36686977386474606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,8,2,power_law_1.2,0.47258625030517576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,8,2,power_law_1.2,0.4789132690429687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,8,2,power_law_1.2,0.5062131118774414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,8,2,power_law_1.2,0.487375373840332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,8,2,power_law_1.2,0.5365184020996094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,8,2,power_law_1.2,0.6067865753173829
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,8,2,power_law_1.2,0.7128089904785156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,8,2,power_law_1.2,0.7989888000488281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,8,2,power_law_1.2,0.9536089324951172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,8,2,power_law_1.2,1.1631449890136718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,8,2,power_law_1.2,1.5171942138671874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,8,2,power_law_1.2,0.04039040088653564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,8,2,power_law_1.2,1.8663961791992185
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,8,2,power_law_1.2,0.05149439811706543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,8,2,power_law_1.2,0.07199872016906739
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,8,2,power_law_1.2,0.0995199966430664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,8,2,power_law_1.2,2.658161926269531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,8,2,power_law_1.2,0.14410367965698243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,8,2,power_law_1.2,0.1925324821472168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,8,2,power_law_1.2,0.2015065574645996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,8,2,power_law_1.2,0.2593715286254883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,8,2,power_law_1.2,0.2685055923461914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,8,2,power_law_1.2,3.4144128417968753
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,8,2,power_law_1.2,0.2716300773620605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,8,2,power_law_1.2,0.26978559494018556
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,8,2,power_law_1.2,0.28813695907592773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,8,2,power_law_1.2,0.29293695449829105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,8,2,power_law_1.2,0.3287654495239258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,8,2,power_law_1.2,0.3349747085571289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,8,2,power_law_1.2,0.33407230377197267
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,8,2,power_law_1.2,0.3507379150390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,8,2,power_law_1.2,0.3720345687866211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,8,2,power_law_1.2,0.4055462265014649
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,8,2,power_law_1.2,0.45245952606201173
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,8,2,power_law_1.01,0.7757836914062499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,8,2,power_law_1.2,0.5108160018920899
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,8,2,power_law_1.2,0.6235481643676758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,8,2,power_law_1.2,0.7281817626953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,8,2,power_law_1.2,0.05430399894714355
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,8,2,power_law_1.01,1.0819519805908204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,8,2,power_law_1.2,0.9510425567626953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,8,2,power_law_1.2,0.06089983940124512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,8,2,power_law_1.2,1.147479019165039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,8,2,power_law_1.2,0.08357631683349609
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,8,2,power_law_1.2,0.07351679801940918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,8,2,power_law_1.2,0.09764096260070801
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,8,2,power_law_1.2,0.12066559791564942
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,8,2,power_law_1.2,0.12854528427124023
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,8,2,power_law_1.2,0.1554918384552002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,8,2,power_law_1.2,1.57338623046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,8,2,power_law_1.2,0.16081151962280274
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,8,2,power_law_1.2,0.16491008758544923
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,8,2,power_law_1.2,0.17147520065307617
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,8,2,power_law_1.2,0.17810304641723634
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,8,2,power_law_1.2,0.18192768096923828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,8,2,power_law_1.2,0.18585472106933593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,8,2,power_law_1.2,2.030503692626953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,8,2,power_law_1.2,0.19334911346435546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,8,2,power_law_1.2,0.20059520721435548
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,8,2,power_law_1.2,0.21141632080078124
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,8,2,power_law_1.01,1.54887939453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,8,2,power_law_1.01,1.931745300292969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,8,2,power_law_1.2,0.28659711837768553
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,8,2,power_law_1.2,0.250067195892334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,8,2,power_law_1.2,0.42323455810546873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,16,1,balanced,0.05412032127380371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,16,1,balanced,0.07693056106567384
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,8,2,power_law_1.2,0.344474868774414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,16,1,balanced,0.09147711753845214
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,8,2,power_law_1.2,0.5854655838012695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,16,1,balanced,0.11524736404418945
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,8,2,power_law_1.2,0.7027686309814454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,16,1,balanced,0.17023616790771484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,16,1,balanced,0.34910911560058594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,16,1,balanced,0.45489856719970706
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,16,1,balanced,0.4565286254882812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,16,1,balanced,0.45657470703124997
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,8,2,power_law_1.2,1.0368870544433595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,16,1,balanced,0.4586342239379883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,16,1,balanced,0.4617555236816406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,16,1,balanced,0.46625217437744143
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,16,1,balanced,0.4688761520385742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,16,1,balanced,0.4800243377685547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,16,1,balanced,0.487438735961914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,16,1,balanced,0.6643750762939453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,16,1,balanced,0.6948057556152344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,16,1,balanced,0.7255443572998047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,16,1,balanced,0.7888108825683593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,16,1,balanced,0.4772294235229492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,16,1,balanced,0.8522163391113281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,16,1,balanced,0.9881478118896485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,16,1,balanced,1.2861798095703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,16,1,balanced,0.04961152076721191
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,16,1,balanced,1.5392160034179687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,16,1,balanced,0.05780672073364258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,16,1,balanced,0.07609792232513428
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,16,1,balanced,0.11283072471618652
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,16,1,balanced,1.99943359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,16,1,balanced,0.17889280319213868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,16,1,balanced,0.3255673599243164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,16,1,balanced,0.46393409729003904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,16,1,balanced,0.46521472930908203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,16,1,balanced,0.4670348739624023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,16,1,balanced,2.7099954223632814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,16,1,balanced,0.46950527191162106
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,16,1,balanced,0.47372673034667967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,16,1,balanced,0.47965312957763667
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,16,1,balanced,0.4845375823974609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,16,1,balanced,0.49061439514160154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,16,1,balanced,0.49370368957519534
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,16,1,balanced,0.500513916015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,16,1,balanced,0.5603622436523438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,16,1,balanced,0.5780960083007812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,16,1,balanced,3.7955474853515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,16,1,balanced,0.5934380722045899
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,16,1,balanced,0.632092170715332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,16,1,balanced,0.6690592193603516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,16,1,balanced,0.7425561523437499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,16,1,balanced,0.8198502349853516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,16,1,balanced,0.969528350830078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,16,1,balanced,1.4484632873535157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,16,1,balanced,1.747685089111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,16,1,balanced,2.3744134521484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,8,2,power_law_1.2,1.5098739624023438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,8,2,power_law_1.2,2.271653137207031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,8,2,power_law_1.2,3.2298483276367187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,16,1,balanced,0.08369024276733399
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,16,1,balanced,0.08328512191772461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,16,1,balanced,0.08445759773254394
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,16,1,balanced,0.08615807533264161
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,16,1,balanced,0.11686016082763673
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,16,1,balanced,0.12331904411315917
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,16,1,balanced,0.1372640037536621
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,16,1,balanced,0.08975232124328614
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,16,1,balanced,0.09605376243591308
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,16,1,balanced,0.10294079780578613
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,16,1,balanced,0.10996735572814942
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,16,1,balanced,0.1713209533691406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,16,1,balanced,0.24006399154663086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,16,1,balanced,0.19118719100952147
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,16,1,balanced,0.33591743469238283
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,16,1,balanced,0.294333438873291
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,16,1,balanced,0.42513854980468746
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,16,1,balanced,0.5964108657836914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,16,1,balanced,0.7746726226806641
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,16,1,balanced,1.133807373046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,16,1,power_law_1.01,0.04764927864074707
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,16,1,balanced,1.4762471008300782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,16,1,power_law_1.01,0.06771647930145264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,16,1,power_law_1.01,0.11627136230468751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,16,1,power_law_1.01,0.0921280002593994
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,16,1,power_law_1.01,0.16315519332885742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,16,1,balanced,2.181095733642578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,16,1,power_law_1.01,0.22902143478393552
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,16,1,power_law_1.01,0.266625919342041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,16,1,power_law_1.01,0.3752409744262695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,16,1,power_law_1.01,0.3559987258911133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,16,1,power_law_1.01,0.3890777587890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,16,1,power_law_1.01,0.39474113464355465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,16,1,power_law_1.01,0.40625728607177736
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,16,1,balanced,2.8802706909179685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,16,1,power_law_1.01,0.4124518585205078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,16,1,power_law_1.01,0.4270803070068359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,16,1,power_law_1.01,0.44139648437499995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,16,1,power_law_1.01,0.6445401763916015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,16,1,power_law_1.01,0.47319808959960935
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,16,1,power_law_1.01,0.6724921417236328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,16,1,power_law_1.01,0.7126464080810547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,16,1,power_law_1.01,0.8520800018310547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,16,1,balanced,4.294075622558593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,16,1,power_law_1.01,0.9258086395263672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,16,1,power_law_1.01,1.0748786926269531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,16,1,power_law_1.01,1.2686605072021484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,16,1,power_law_1.01,0.04964352130889892
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,16,1,power_law_1.01,0.05744383811950684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,16,1,power_law_1.01,1.6592037963867188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,16,1,power_law_1.01,0.07185152053833008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,16,1,power_law_1.01,0.11269439697265624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,16,1,power_law_1.01,2.0404217529296877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,16,1,power_law_1.01,0.1714963150024414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,16,1,power_law_1.01,0.2621299171447754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,16,1,power_law_1.01,2.8083889770507815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,16,1,power_law_1.01,0.31519744873046873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,16,1,power_law_1.01,0.3575398254394531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,16,1,power_law_1.01,0.3700806427001953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,16,1,power_law_1.01,0.37963840484619144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,16,1,power_law_1.01,0.3929158401489258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,16,1,power_law_1.01,3.560061950683594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,16,1,power_law_1.01,0.3971571350097656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,16,1,power_law_1.01,0.4174457550048828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,16,1,power_law_1.01,0.43124225616455075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,16,1,power_law_1.01,0.4525606536865235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,16,1,power_law_1.01,0.5002758407592773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,16,1,power_law_1.01,0.5261523056030273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,16,1,power_law_1.01,0.5523974227905273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,16,1,power_law_1.01,0.6119174575805664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,16,1,power_law_1.01,0.6683103942871094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,16,1,power_law_1.01,0.4351033782958984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,16,1,power_law_1.01,0.8038489532470703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,16,1,power_law_1.01,0.9225651550292969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,16,1,power_law_1.01,1.1784896087646484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,16,1,power_law_1.01,1.4271987915039062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,16,1,power_law_1.01,1.923585205078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,16,1,power_law_1.01,0.08347200393676758
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,16,1,power_law_1.01,0.08323967933654786
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,16,1,power_law_1.01,2.4331922912597657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,16,1,power_law_1.01,0.08425536155700683
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,16,1,power_law_1.01,0.08587648391723632
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,16,1,power_law_1.01,0.08935551643371582
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,16,1,power_law_1.01,0.09603327751159668
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,16,1,power_law_1.01,0.10338815689086914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,16,1,power_law_1.01,0.10985024452209473
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,16,1,power_law_1.01,0.11687423706054687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,16,1,power_law_1.01,0.13688447952270508
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,16,1,power_law_1.01,0.12358783721923827
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,16,1,power_law_1.01,0.18193792343139648
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,16,1,power_law_1.01,0.23965440750122072
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,16,1,power_law_1.01,0.293311996459961
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,16,1,power_law_1.01,0.197194881439209
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,16,1,power_law_1.01,0.33609279632568356
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,16,1,power_law_1.01,0.4245312118530274
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,16,1,power_law_1.01,0.7714579010009766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,16,1,power_law_1.01,0.5990054321289062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,16,1,power_law_1.01,2.1808268737792966
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,16,1,power_law_1.01,2.8810699462890623
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,16,1,power_law_1.01,1.1249913787841797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,16,1,power_law_1.01,1.4833509826660156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,16,1,power_law_1.2,0.04770688056945801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,16,1,power_law_1.2,0.06975743770599366
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,16,1,power_law_1.2,0.09192000389099121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,16,1,power_law_1.2,0.11629311561584472
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,16,1,power_law_1.2,0.15626239776611328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,16,1,power_law_1.2,0.2141414451599121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,16,1,power_law_1.2,0.25307584762573243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,16,1,power_law_1.2,0.3504064178466797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,16,1,power_law_1.2,0.36721153259277345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,16,1,power_law_1.2,0.37788928985595704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,16,1,power_law_1.2,0.3941107177734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,16,1,power_law_1.2,0.4012876892089844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,16,1,power_law_1.2,0.4088409423828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,16,1,power_law_1.2,0.42479614257812504
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,16,1,power_law_1.01,4.306385803222656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,16,1,power_law_1.2,0.4466854476928711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,16,1,power_law_1.2,0.4763801574707031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,16,1,power_law_1.2,0.6417823791503906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,16,1,power_law_1.2,0.6743885040283203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,16,1,power_law_1.2,0.7284095764160157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,16,1,power_law_1.2,0.8577324676513672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,16,1,power_law_1.2,0.9326943969726562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,16,1,power_law_1.2,1.0877126312255858
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,16,1,power_law_1.2,1.2868531799316405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,16,1,power_law_1.2,1.666972198486328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,16,1,power_law_1.2,2.0587493896484377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,16,1,power_law_1.2,2.8158636474609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,16,1,power_law_1.2,0.058453760147094726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,16,1,power_law_1.2,0.07992640018463135
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,16,1,power_law_1.2,0.1123417568206787
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,16,1,power_law_1.2,3.5959674072265626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,16,1,power_law_1.2,0.16213247299194336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,16,1,power_law_1.2,0.24772031784057616
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,16,1,power_law_1.2,0.305731201171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,16,1,power_law_1.2,0.3542828750610352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,16,1,power_law_1.2,0.3651897430419922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,16,1,power_law_1.2,0.3718777465820312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,16,1,power_law_1.2,0.3834086227416992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,16,1,power_law_1.2,0.395285758972168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,16,1,power_law_1.2,0.4104755020141601
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,16,1,power_law_1.2,0.4256633758544922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,16,1,power_law_1.2,0.4358060836791992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,16,1,power_law_1.2,0.4512006378173828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,16,1,power_law_1.2,0.5041324615478515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,16,1,power_law_1.2,0.6269254302978515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,16,1,power_law_1.2,0.04684224128723145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,16,1,power_law_1.2,0.5654719924926758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,16,1,power_law_1.2,0.6909030151367188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,16,1,power_law_1.2,0.8284038543701172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,16,1,power_law_1.2,0.95229248046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,16,1,power_law_1.2,1.2052563476562501
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,16,1,power_law_1.2,0.08344639778137207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,16,1,power_law_1.2,0.530442886352539
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,16,1,power_law_1.2,0.08348159790039063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,16,1,power_law_1.2,1.9637132263183594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,16,1,power_law_1.2,2.464671325683594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,16,1,power_law_1.2,0.085731840133667
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,16,1,power_law_1.2,0.09728960037231446
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,16,1,power_law_1.2,0.10285759925842286
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,16,1,power_law_1.2,0.11209407806396485
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,16,1,power_law_1.2,0.11676223754882813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,16,1,power_law_1.2,1.4526170349121093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,16,1,power_law_1.2,0.12568063735961915
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,16,1,power_law_1.2,0.13690367698669434
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,16,1,power_law_1.2,0.19694463729858397
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,16,1,power_law_1.2,0.2401113510131836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,16,1,power_law_1.2,0.2953945541381836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,16,1,power_law_1.2,0.08985535621643066
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,16,1,power_law_1.2,0.08369024276733399
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,16,1,power_law_1.2,0.16957183837890627
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,16,1,power_law_1.2,0.33697086334228515
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,16,1,power_law_1.2,0.4268601608276367
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,16,1,power_law_1.2,1.4768121337890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,16,1,power_law_1.2,0.5995891189575195
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,16,1,power_law_1.2,0.7696768188476563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,16,1,power_law_1.2,1.1263180541992188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,16,1,power_law_1.2,2.1905970764160156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,16,balanced,0.025177600383758547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,16,balanced,0.024404480457305908
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,16,balanced,0.028044800758361816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,16,balanced,0.06161280155181885
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,16,1,power_law_1.2,2.895721740722656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,16,balanced,0.11702272415161133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,16,balanced,0.12190208435058594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,16,balanced,0.11942655563354491
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,16,balanced,0.12171392440795899
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,16,balanced,0.12332799911499023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,16,balanced,0.12933247566223144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,16,balanced,0.13164159774780274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,16,balanced,0.12441216468811037
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,16,balanced,0.12757887840270996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,16,balanced,0.13187456130981445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,16,balanced,0.13902591705322265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,16,balanced,0.15110527992248535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,16,balanced,0.15890175819396973
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,16,balanced,0.20039424896240235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,16,balanced,0.21581312179565432
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,16,balanced,0.3135654449462891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,16,balanced,0.35949951171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,16,balanced,0.4878988647460938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,16,balanced,0.6098393630981446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,16,balanced,0.8810598754882812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,16,balanced,0.02846719980239868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,16,balanced,0.028465919494628907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,16,balanced,1.146762237548828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,16,balanced,0.029360640048980712
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,16,balanced,0.030087680816650392
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,16,balanced,0.04311935901641846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,16,balanced,0.07156735897064209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,16,balanced,0.08166912078857422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,16,balanced,0.0811417579650879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,16,balanced,0.08342399597167968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,16,balanced,0.08422143936157227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,16,balanced,0.08684288024902344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,16,balanced,0.08764287948608399
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,16,balanced,0.09190655708312988
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,16,balanced,0.08151935577392579
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,16,balanced,0.08323712348937988
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,16,balanced,0.08383232116699219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,16,balanced,0.08778240203857421
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,16,balanced,0.09005951881408691
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,16,balanced,0.10195839881896973
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,16,balanced,0.11741312026977539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,16,balanced,0.04672255992889404
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,16,balanced,0.13173631668090818
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,16,balanced,0.16409088134765626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,16,balanced,0.19078527450561525
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,16,balanced,0.2740224075317383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,16,balanced,0.33055614471435546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,16,balanced,0.4586111831665039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,16,balanced,0.6163788986206055
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,16,1,power_law_1.2,4.3043826293945315
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,1,16,balanced,0.019379199743270875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,1,16,balanced,0.019953919649124147
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,1,16,balanced,0.040757761001586915
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,1,16,balanced,0.04113535881042481
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,1,16,balanced,0.04135551929473877
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,1,16,balanced,0.04112639904022217
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,1,16,balanced,0.041265921592712404
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,1,16,balanced,0.04160511970520019
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,1,16,balanced,0.04148096084594727
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,1,16,balanced,0.041381120681762695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,1,16,balanced,0.041269760131835934
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,1,16,balanced,0.0421619176864624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,1,16,balanced,0.04196864128112793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,1,16,balanced,0.04224639892578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,1,16,balanced,0.04271872043609619
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,1,16,balanced,0.043576321601867675
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,1,16,balanced,0.044281601905822754
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,1,16,balanced,0.045401601791381835
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,1,16,balanced,0.04753280162811279
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,1,16,balanced,0.04962175846099853
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,1,16,balanced,0.019132159948349
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,1,16,balanced,0.020806400775909423
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,1,16,balanced,0.0685427188873291
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,1,16,balanced,0.026716160774230956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,16,power_law_1.01,0.06131968021392822
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,1,16,balanced,0.07130623817443847
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,1,16,balanced,0.0931980800628662
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,16,power_law_1.01,0.09049983978271484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,16,power_law_1.01,0.09666303634643555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,16,power_law_1.01,0.10141695976257323
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,16,power_law_1.01,0.11120896339416504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,16,power_law_1.01,0.11533951759338379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,16,power_law_1.01,0.12147711753845214
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,16,power_law_1.01,0.1201740837097168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,16,power_law_1.01,0.12448639869689941
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,16,power_law_1.01,0.1292505645751953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,16,power_law_1.01,0.13487232208251954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,16,power_law_1.01,0.12522368431091307
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,16,power_law_1.01,0.13123711585998535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,16,power_law_1.01,0.1357107162475586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,16,power_law_1.01,0.14916223526000977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,16,power_law_1.01,0.20071039199829102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,16,power_law_1.01,0.22162687301635745
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,16,power_law_1.01,0.26303359985351565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,16,power_law_1.01,0.2967795181274414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,16,power_law_1.01,0.09183232307434082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,16,power_law_1.01,0.4295820617675782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,16,power_law_1.01,0.5297536087036133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,16,power_law_1.01,0.77997314453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,16,power_law_1.01,0.8274649810791015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,16,power_law_1.01,0.043123202323913576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,16,power_law_1.01,1.2715878295898437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,16,power_law_1.01,0.06950655937194825
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,16,power_law_1.01,0.06999807834625245
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,16,power_law_1.01,0.0708838415145874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,16,power_law_1.01,0.07066624164581299
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,16,power_law_1.01,0.06822400093078614
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,16,power_law_1.01,1.749291534423828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,16,power_law_1.01,0.07606143951416015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,16,power_law_1.01,0.0760588788986206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,16,power_law_1.01,0.08327808380126953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,16,power_law_1.01,0.07898752212524414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,16,power_law_1.01,0.08355584144592285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,16,power_law_1.01,0.08584320068359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,16,power_law_1.01,0.08648447990417481
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,16,power_law_1.01,0.09040127754211426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,16,power_law_1.01,0.08078975677490234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,16,power_law_1.01,0.0819264030456543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,16,power_law_1.01,0.08350336074829101
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,16,power_law_1.01,0.09575167655944825
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,16,power_law_1.01,0.1120025634765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,16,power_law_1.01,0.08450048446655274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,16,power_law_1.01,0.13615615844726564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,16,power_law_1.01,0.16744192123413087
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,16,power_law_1.01,0.2021683120727539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,16,power_law_1.01,0.26911359786987304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,16,power_law_1.01,0.3633894348144531
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,1,16,power_law_1.01,0.02628479957580566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,16,power_law_1.01,0.4744422531127929
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,1,16,power_law_1.01,0.04086656093597412
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,1,16,balanced,0.11482368469238283
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,16,power_law_1.01,0.7376012420654297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,1,16,power_law_1.01,0.041414399147033695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,1,16,power_law_1.01,0.04081791877746582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,16,power_law_1.01,0.9096371459960938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,1,16,power_law_1.01,0.036858880519866945
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,1,16,power_law_1.01,0.03958271980285645
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,1,16,power_law_1.01,0.0409881591796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,1,16,power_law_1.01,0.040828161239624024
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,1,16,power_law_1.01,0.03892224073410035
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,1,16,power_law_1.01,0.04107391834259033
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,1,16,power_law_1.01,0.04104959964752197
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,1,16,power_law_1.01,0.04098432064056397
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,1,16,power_law_1.01,0.041843199729919435
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,1,16,power_law_1.01,0.04204160213470459
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,1,16,power_law_1.01,0.04197375774383545
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,1,16,power_law_1.01,0.043393278121948244
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,1,16,power_law_1.01,0.04870272159576416
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,1,16,power_law_1.01,0.03977344036102295
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,1,16,power_law_1.01,0.04568575859069825
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,1,16,power_law_1.01,0.056232957839965826
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,1,16,power_law_1.01,0.0631001615524292
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,1,16,power_law_1.01,0.09266559600830078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,1,16,power_law_1.01,0.07165567874908448
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,16,power_law_1.2,0.06136960029602051
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,16,power_law_1.2,0.09125120162963866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,16,power_law_1.2,0.08955519676208497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,16,power_law_1.2,0.09590911865234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,16,power_law_1.2,0.09789567947387695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,16,power_law_1.2,0.09969152450561523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,16,power_law_1.2,0.1102182388305664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,16,power_law_1.2,0.1164352035522461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,16,power_law_1.2,0.12240127563476562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,16,power_law_1.2,0.12473471641540528
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,16,power_law_1.2,0.12913408279418945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,16,power_law_1.2,0.12779647827148438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,16,power_law_1.2,0.12620800018310546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,16,power_law_1.2,0.13255040168762205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,16,power_law_1.2,0.13990655899047852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,16,power_law_1.2,0.1653299140930176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,16,power_law_1.2,0.2040115165710449
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,16,power_law_1.2,0.23138944625854493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,16,power_law_1.2,0.2683020782470703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,16,power_law_1.2,0.33545856475830077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,16,power_law_1.2,0.6803929901123047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,16,power_law_1.2,0.8508979034423827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,16,power_law_1.2,1.1450406646728515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,16,power_law_1.2,0.11362175941467285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,16,power_law_1.2,1.8566835021972659
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,16,power_law_1.2,0.04383999824523926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,16,power_law_1.2,0.06973055839538575
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,16,power_law_1.2,0.06114304065704346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,16,power_law_1.2,0.06982016086578369
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,16,power_law_1.2,0.07169407844543457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,16,power_law_1.2,0.0720255994796753
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,16,power_law_1.2,2.246307830810547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,16,power_law_1.2,0.07266816139221191
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,16,power_law_1.2,0.07965824127197266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,16,power_law_1.2,0.08088704109191894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,16,power_law_1.2,0.08169983863830567
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,16,power_law_1.2,0.09021439552307128
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,16,power_law_1.2,0.48700542449951173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,16,power_law_1.2,0.08039551734924316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,16,power_law_1.2,0.08075263977050781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,16,power_law_1.2,0.08142975807189942
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,16,power_law_1.2,0.0850111961364746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,16,power_law_1.2,0.086561279296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,16,power_law_1.2,0.10138879776000978
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,16,power_law_1.2,0.11889535903930663
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,16,power_law_1.2,0.1423423957824707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,16,power_law_1.2,0.18710527420043946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,16,power_law_1.2,0.2604953575134277
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,16,power_law_1.2,0.31446912765502927
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,16,power_law_1.2,0.4342809677124023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,16,power_law_1.2,0.5627699279785157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,1,16,power_law_1.2,0.0270796799659729
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,1,16,power_law_1.2,0.04074111938476562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,1,16,power_law_1.01,0.12402432441711427
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,1,16,power_law_1.01,0.15287551879882813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,16,power_law_1.2,0.9173145294189453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,1,16,power_law_1.01,0.2225587272644043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,1,16,power_law_1.2,0.04099584102630615
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,1,16,power_law_1.2,0.041218562126159666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,1,16,power_law_1.2,0.040757761001586915
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,1,16,power_law_1.2,0.040801281929016116
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,1,16,power_law_1.2,0.03639679908752441
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,16,power_law_1.2,1.0981517028808594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,1,16,power_law_1.2,0.04106880187988281
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,1,16,power_law_1.2,0.03952383995056152
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,1,16,power_law_1.2,0.04123519897460938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,1,16,power_law_1.2,0.04113279819488526
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,1,16,power_law_1.2,0.04165760040283203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,1,16,power_law_1.2,0.042191362380981444
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,1,16,power_law_1.2,0.04982783794403076
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,1,16,power_law_1.2,0.0529472017288208
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,1,16,power_law_1.2,0.040459518432617185
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,1,16,power_law_1.01,0.25167999267578123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,16,power_law_1.2,0.06757887840270996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,1,16,power_law_1.2,0.04425600051879883
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,1,16,power_law_1.2,0.07559296131134033
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,1,16,power_law_1.2,0.06399487972259521
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,1,16,power_law_1.2,0.09775487899780273
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,1,16,power_law_1.2,0.03315711975097656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,1,16,power_law_1.2,0.040989441871643065
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,1,16,power_law_1.2,0.11057279586791993
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,2,8,balanced,0.04661119937896728
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,2,8,balanced,0.026462719440460202
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,2,8,balanced,0.06260479927062988
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,2,8,balanced,0.12016511917114259
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,2,8,balanced,0.1221887969970703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,2,8,balanced,0.1258790397644043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,2,8,balanced,0.12940799713134765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,2,8,balanced,0.10859264373779295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,2,8,balanced,0.12039551734924317
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,2,8,balanced,0.02646912097930908
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,2,8,balanced,0.12382464408874512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,2,8,balanced,0.1265510368347168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,2,8,balanced,0.13121919631958007
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,2,8,balanced,0.14098560333251953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,2,8,balanced,0.13603455543518067
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,2,8,balanced,0.14859135627746584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,2,8,balanced,0.16880640029907226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,2,8,balanced,0.1889664077758789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,2,8,balanced,0.22537471771240236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,2,8,balanced,0.2656806373596191
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,2,8,balanced,0.3843609619140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,2,8,balanced,0.4565785598754883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,2,8,balanced,0.6367539215087891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,2,8,balanced,0.8127667236328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,2,8,balanced,0.02789247989654541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,2,8,balanced,1.188321304321289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,2,8,balanced,0.029277439117431643
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,2,8,balanced,0.03570303916931152
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,2,8,balanced,0.050104317665100095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,2,8,balanced,1.5623513793945312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,2,8,balanced,0.08410880088806152
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,2,8,balanced,0.08648192405700683
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,2,8,balanced,0.028728320598602298
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,2,8,balanced,0.09197183609008788
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,2,8,balanced,0.09340288162231444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,2,8,balanced,0.02957184076309204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,2,8,balanced,0.08367103576660156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,2,8,balanced,0.0849407958984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,2,8,balanced,0.08559871673583984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,2,8,balanced,0.08853376388549805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,2,8,balanced,0.08967679977416992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,2,8,balanced,0.09199872016906738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,2,8,balanced,0.1017024040222168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,2,8,balanced,0.11125247955322266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,2,8,balanced,0.12479999542236328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,2,8,balanced,0.14932095527648925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,2,8,balanced,0.0885747241973877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,2,8,balanced,0.17542144775390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,2,8,balanced,0.22617471694946287
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,2,8,balanced,0.27706495285034183
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,2,8,balanced,0.39156734466552734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,2,8,balanced,0.4869286346435547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,2,8,balanced,0.6951654052734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,2,8,balanced,0.9193510437011719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,1,16,power_law_1.2,0.1557900810241699
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,1,16,power_law_1.2,0.3064998435974121
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,1,16,power_law_1.2,0.19262847900390626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,1,16,power_law_1.2,0.4710400009155274
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,2,8,balanced,0.02052736043930054
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,2,8,balanced,0.04503039836883545
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,2,8,balanced,0.021107199192047118
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,2,8,balanced,0.02166271924972534
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,2,8,balanced,0.0448089599609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,2,8,balanced,0.04500095844268799
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,2,8,balanced,0.04549759864807129
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,2,8,balanced,0.04537471771240235
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,2,8,balanced,0.046014719009399414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,2,8,balanced,0.04618112087249756
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,2,8,balanced,0.04660736083984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,2,8,balanced,0.046762242317199706
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,2,8,balanced,0.02367487907409668
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,2,8,balanced,0.047640318870544436
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,2,8,balanced,0.04940032005310059
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,2,8,balanced,0.050228481292724614
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,2,8,balanced,0.05205887794494629
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,2,8,balanced,0.055068159103393556
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,2,8,balanced,0.05900288105010987
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,2,8,balanced,0.029208319187164305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,2,8,power_law_1.01,0.046175999641418455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,2,8,power_law_1.01,0.05929599761962891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,2,8,power_law_1.01,0.06920063972473145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,2,8,power_law_1.01,0.07852928161621094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,2,8,power_law_1.01,0.08776063919067383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,2,8,power_law_1.01,0.09369088172912597
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,2,8,power_law_1.01,0.11363583564758302
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,2,8,power_law_1.01,0.11190015792846679
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,2,8,power_law_1.01,0.11317760467529296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,2,8,power_law_1.01,0.11554304122924805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,2,8,power_law_1.01,0.12415743827819825
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,2,8,power_law_1.01,0.11847935676574708
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,2,8,balanced,0.04473983764648438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,2,8,power_law_1.01,0.1159334373474121
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,2,8,balanced,0.045185279846191403
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,2,8,power_law_1.01,0.12049407958984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,2,8,power_law_1.01,0.12556415557861328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,2,8,power_law_1.01,0.14141183853149414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,2,8,power_law_1.01,0.12637951850891113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,2,8,power_law_1.01,0.1655615997314453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,2,8,power_law_1.01,0.19757440567016601
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,2,8,power_law_1.01,0.24334848403930662
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,2,8,power_law_1.01,0.2866815948486328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,2,8,power_law_1.01,0.3722227096557617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,2,8,power_law_1.01,0.4851302337646485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,2,8,power_law_1.01,0.6668646240234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,2,8,balanced,0.044925441741943364
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,2,8,power_law_1.01,0.8547443389892578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,2,8,power_law_1.01,0.03214848041534424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,2,8,power_law_1.01,0.04673791885375976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,2,8,power_law_1.01,0.05139840126037598
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,2,8,power_law_1.01,1.2204978942871094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,2,8,power_law_1.01,0.05613696098327636
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,2,8,power_law_1.01,0.06833792209625245
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,2,8,power_law_1.01,0.07739264011383057
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,2,8,power_law_1.01,0.07333631992340088
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,2,8,power_law_1.01,1.5724223327636717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,2,8,power_law_1.01,0.07399551868438721
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,2,8,power_law_1.01,0.07981567859649659
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,2,8,power_law_1.01,0.08997504234313965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,2,8,power_law_1.01,0.0847871971130371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,2,8,power_law_1.01,0.08015232086181641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,2,8,power_law_1.01,0.08367103576660156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,2,8,power_law_1.01,0.0864140796661377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,2,8,power_law_1.01,0.08799360275268556
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,2,8,power_law_1.01,0.09298815727233886
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,2,8,power_law_1.01,0.10290944099426269
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,2,8,power_law_1.01,0.1130508804321289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,2,8,power_law_1.01,0.1422707176208496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,2,8,power_law_1.01,0.16181119918823242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,2,8,power_law_1.01,0.22589183807373048
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,2,8,power_law_1.01,0.2690278434753418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,2,8,power_law_1.01,0.3752345657348633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,2,8,power_law_1.01,0.4644992065429688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,2,8,power_law_1.01,0.023470079898834227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,2,8,power_law_1.01,0.6651174163818359
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,2,8,power_law_1.01,0.030117120742797855
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,2,8,power_law_1.01,0.03870464086532593
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,2,8,power_law_1.01,0.03589632034301758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,2,8,power_law_1.01,0.8768473815917968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,2,8,power_law_1.01,0.042293758392333985
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,2,8,power_law_1.01,0.04161920070648194
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,2,8,power_law_1.01,0.042375679016113284
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,2,8,power_law_1.01,0.04011392116546631
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,2,8,power_law_1.01,0.04278783798217774
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,2,8,power_law_1.01,0.04175487995147705
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,2,8,power_law_1.01,0.04246016025543213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,2,8,power_law_1.01,0.07067647933959961
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,2,8,power_law_1.01,0.04389247894287109
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,2,8,power_law_1.01,0.043624958992004394
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,2,8,power_law_1.01,0.04492288112640381
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,2,8,power_law_1.01,0.0455244779586792
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,2,8,power_law_1.01,0.04566143989562989
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,2,8,power_law_1.01,0.04663936138153076
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,2,8,balanced,0.08021632194519043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,2,8,balanced,0.08711935997009276
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,2,8,balanced,0.11297280311584472
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,2,8,balanced,0.14226048469543456
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,2,8,power_law_1.01,0.07130112171173095
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,2,8,power_law_1.01,0.05350016117095947
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,2,8,power_law_1.01,0.050173439979553223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,2,8,power_law_1.2,0.04654848098754883
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,2,8,power_law_1.01,0.0651033592224121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,2,8,power_law_1.2,0.059335680007934564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,2,8,power_law_1.2,0.06409471988677978
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,2,8,power_law_1.2,0.07448959827423096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,2,8,power_law_1.2,0.0873100757598877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,2,8,power_law_1.2,0.09726079940795898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,2,8,power_law_1.2,0.11726847648620606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,2,8,power_law_1.2,0.11374719619750977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,2,8,power_law_1.2,0.11160320281982421
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,2,8,power_law_1.2,0.116244478225708
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,2,8,power_law_1.01,0.09179903984069823
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,2,8,power_law_1.01,0.10370047569274903
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,2,8,power_law_1.2,0.12296704292297363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,2,8,power_law_1.2,0.12099583625793457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,2,8,power_law_1.2,0.12059904098510743
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,2,8,power_law_1.2,0.112989444732666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,2,8,power_law_1.2,0.12445823669433594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,2,8,power_law_1.2,0.1303974437713623
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,2,8,power_law_1.01,0.18940160751342774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,2,8,power_law_1.2,0.13861760139465332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,2,8,power_law_1.2,0.17198463439941408
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,2,8,power_law_1.2,0.20855680465698243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,2,8,power_law_1.2,0.256496639251709
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,2,8,power_law_1.2,0.30518911361694334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,2,8,power_law_1.01,0.13565312385559083
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,2,8,power_law_1.2,0.4031974411010742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,2,8,power_law_1.2,0.49711486816406253
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,2,8,power_law_1.2,0.04591104030609131
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,2,8,power_law_1.2,0.726960678100586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,2,8,power_law_1.2,0.04368896007537842
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,2,8,power_law_1.2,0.9130278778076171
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,2,8,power_law_1.2,0.05682047843933106
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,2,8,power_law_1.2,0.030161919593811037
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,2,8,power_law_1.2,0.07614848136901856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,2,8,power_law_1.2,0.07621632099151612
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,2,8,power_law_1.2,1.3576882934570311
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,2,8,power_law_1.2,0.06873983860015868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,2,8,power_law_1.2,0.07219583988189697
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,2,8,power_law_1.2,0.0771123218536377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,2,8,power_law_1.2,0.08923392295837403
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,2,8,power_law_1.2,0.0785152006149292
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,2,8,power_law_1.2,1.7136793518066404
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,2,8,power_law_1.2,0.08471424102783202
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,2,8,power_law_1.2,0.08577280044555664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,2,8,power_law_1.2,0.08822400093078613
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,2,8,power_law_1.2,0.09151103973388672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,2,8,power_law_1.2,0.107258882522583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,2,8,power_law_1.2,0.1184934425354004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,2,8,power_law_1.2,0.1500160026550293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,2,8,power_law_1.2,0.06869887828826904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,2,8,power_law_1.2,0.16993663787841798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,2,8,power_law_1.2,0.22781183242797853
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,2,8,power_law_1.2,0.29083135604858396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,2,8,power_law_1.2,0.3876300811767578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,2,8,power_law_1.2,0.023486719131469727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,2,8,power_law_1.2,0.08029696464538574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,2,8,power_law_1.2,0.5127884674072265
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,2,8,power_law_1.2,0.029678080081939694
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,2,8,power_law_1.2,0.03011199951171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,2,8,power_law_1.2,0.7348671722412109
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,2,8,power_law_1.2,0.04145919799804688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,2,8,power_law_1.2,0.04068607807159424
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,2,8,power_law_1.2,0.04248064041137695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,2,8,power_law_1.2,0.038255360126495365
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,2,8,power_law_1.2,0.9763494110107421
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,2,8,power_law_1.2,0.04106239795684814
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,2,8,power_law_1.2,0.042302718162536626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,2,8,power_law_1.2,0.04297215938568115
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,2,8,power_law_1.2,0.04324863910675049
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,2,8,power_law_1.2,0.04387455940246582
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,2,8,power_law_1.2,0.04474624156951904
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,2,8,power_law_1.2,0.04605567932128906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,2,8,power_law_1.2,0.04713215827941895
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,2,8,power_law_1.2,0.06819968223571778
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,2,8,power_law_1.2,0.08005248069763184
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,2,8,power_law_1.2,0.03756799936294556
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,2,8,power_law_1.01,0.3233817672729492
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,2,8,power_law_1.01,0.26619136810302735
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,2,8,power_law_1.2,0.0526259183883667
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,2,8,power_law_1.2,0.048725762367248536
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,2,8,power_law_1.2,0.11614975929260254
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,2,8,power_law_1.2,0.14590720176696775
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,4,4,balanced,0.022310400009155275
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,4,4,balanced,0.024707839488983155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,4,4,balanced,0.04477183818817139
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,4,4,balanced,0.027308800220489503
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,4,4,balanced,0.06380928039550782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,4,4,balanced,0.09812095642089844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,4,4,balanced,0.09815679550170898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,4,4,balanced,0.09923199653625489
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,4,4,balanced,0.11484928131103515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,4,4,balanced,0.11418880462646483
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,4,4,balanced,0.11667712211608887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,4,4,balanced,0.11955967903137207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,4,4,balanced,0.12249600410461425
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,2,8,power_law_1.2,0.21619327545166014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,4,4,balanced,0.12453248023986815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,4,4,balanced,0.12975104331970216
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,4,4,balanced,0.13259903907775877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,4,4,balanced,0.1377280044555664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,4,4,balanced,0.15340160369873046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,4,4,balanced,0.17294208526611327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,4,4,balanced,0.20695680618286133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,4,4,balanced,0.24184831619262695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,4,4,balanced,0.3469747161865234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,4,4,balanced,0.41582977294921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,4,4,balanced,0.5833152008056641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,4,4,balanced,0.7431846618652344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,4,4,balanced,0.02565376043319702
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,4,4,balanced,0.026876161098480223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,4,4,balanced,1.0842073822021485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,4,4,balanced,0.03005824089050293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,4,4,balanced,0.05047808170318604
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,2,8,power_law_1.2,0.05793407917022705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,4,4,balanced,0.08326272010803223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,4,4,balanced,0.08293503761291504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,4,4,balanced,0.08376704216003418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,4,4,balanced,0.09403776168823243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,4,4,balanced,0.09453951835632324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,4,4,balanced,0.09538304328918457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,4,4,balanced,0.09682047843933106
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,4,4,balanced,0.0364352011680603
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,4,4,balanced,0.09867648124694825
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,4,4,balanced,0.09948543548583985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,4,4,balanced,0.10121600151062012
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,4,4,balanced,0.10351231575012207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,4,4,balanced,0.10789119720458984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,4,4,balanced,0.11986432075500489
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,4,4,balanced,0.13052800178527832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,4,4,balanced,0.15341055870056153
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,4,4,balanced,0.174388484954834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,4,4,balanced,0.21333887100219728
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,4,4,balanced,0.2605964851379395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,4,4,balanced,0.3754240036010742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,4,4,balanced,0.4595788955688477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,4,4,balanced,0.6567654418945312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,4,4,balanced,0.8679180908203126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,4,4,balanced,1.4186419677734374
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,2,8,power_law_1.2,0.27817344665527344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,2,8,power_law_1.2,0.4548339080810547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,2,8,power_law_1.2,0.5162124633789062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,4,4,balanced,0.026859519481658933
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,4,4,balanced,0.05664000034332275
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,4,4,balanced,0.05645311832427978
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,4,4,balanced,0.03058432102203369
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,4,4,balanced,0.027927041053771973
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,4,4,balanced,0.05722623825073242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,4,4,balanced,0.05696256160736084
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,4,4,balanced,0.028487679958343504
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,4,4,balanced,0.03976959943771362
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,4,4,balanced,0.057774081230163574
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,4,4,balanced,0.057681918144226074
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,4,4,balanced,0.05823616027832031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,4,4,balanced,0.05951871871948242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,4,4,balanced,0.060906238555908196
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,4,4,balanced,0.06284671783447265
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,4,4,balanced,0.06540671825408936
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,4,4,balanced,0.06880383968353272
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,4,4,balanced,0.07384575843811035
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,4,4,balanced,0.08068991661071777
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,4,4,balanced,0.0568064022064209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,4,4,power_law_1.01,0.025628159046173094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,4,4,power_law_1.01,0.04510848045349121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,4,4,power_law_1.01,0.050576639175415036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,4,4,power_law_1.01,0.05898623943328858
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,4,4,power_law_1.01,0.0678003215789795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,4,4,power_law_1.01,0.07350783824920655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,4,4,power_law_1.01,0.08946816444396972
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,4,4,power_law_1.01,0.08911999702453613
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,4,4,power_law_1.01,0.1086796760559082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,4,4,power_law_1.01,0.10399616241455079
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,4,4,balanced,0.056308479309082025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,4,4,power_law_1.01,0.10761728286743164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,4,4,power_law_1.01,0.1132364845275879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,4,4,power_law_1.01,0.10895487785339356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,4,4,power_law_1.01,0.11943039894104004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,4,4,power_law_1.01,0.12148351669311523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,4,4,power_law_1.01,0.12432127952575685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,4,4,power_law_1.01,0.13458175659179689
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,4,4,power_law_1.01,0.15574655532836915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,4,4,power_law_1.01,0.18118272781372072
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,4,4,power_law_1.01,0.23927423477172854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,4,4,power_law_1.01,0.27608575820922854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,4,4,power_law_1.01,0.35784576416015623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,4,4,power_law_1.01,0.4399014282226562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,4,4,power_law_1.01,0.6201369476318359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,4,4,power_law_1.01,0.7917708587646485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,4,4,power_law_1.01,1.17417724609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,4,4,power_law_1.01,0.027258880138397217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,4,4,power_law_1.01,1.5145869445800781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,4,4,power_law_1.01,0.03191551923751831
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,4,4,power_law_1.01,0.039188480377197264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,4,4,power_law_1.01,0.04829311847686767
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,4,4,power_law_1.01,0.05672448158264161
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,4,4,power_law_1.01,0.062382078170776366
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,4,4,power_law_1.01,0.07983359813690186
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,4,4,balanced,0.05645311832427978
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,4,4,power_law_1.01,0.07630591869354249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,4,4,power_law_1.01,0.08267135620117187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,4,4,power_law_1.01,0.08351743698120118
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,4,4,power_law_1.01,0.08321151733398438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,4,4,power_law_1.01,0.08804736137390137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,4,4,power_law_1.01,0.09185407638549806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,4,4,power_law_1.01,0.09373311996459961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,4,4,power_law_1.01,0.0981619167327881
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,4,4,power_law_1.01,0.09950719833374023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,4,4,power_law_1.01,0.10429311752319337
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,4,4,power_law_1.01,0.11376255989074706
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,4,4,power_law_1.01,0.1283340835571289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,4,4,power_law_1.01,0.1537715244293213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,4,4,power_law_1.01,0.17667455673217775
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,4,4,balanced,0.05663487911224365
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,4,4,power_law_1.01,0.23400320053100584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,4,4,power_law_1.01,0.2847219276428222
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,4,4,power_law_1.01,0.38474239349365236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,4,4,power_law_1.01,0.5007231903076172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,4,4,power_law_1.01,0.0288588809967041
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,4,4,power_law_1.01,0.030807039737701415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,4,4,power_law_1.01,0.6898162841796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,4,4,power_law_1.01,0.03472127914428711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,4,4,power_law_1.01,0.037868800163269045
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,4,4,power_law_1.01,0.0442137622833252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,4,4,power_law_1.01,0.919304962158203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,4,4,power_law_1.01,0.04706943988800048
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,4,4,power_law_1.01,0.05279744148254395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,4,4,power_law_1.01,0.052862720489501955
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,4,4,power_law_1.01,0.05268095970153809
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,4,4,power_law_1.01,0.052933120727539064
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,4,4,power_law_1.01,0.051814398765563964
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,4,4,power_law_1.01,0.05331711769104004
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,4,4,power_law_1.01,0.05351168155670166
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,4,4,power_law_1.01,0.055105280876159665
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,4,4,power_law_1.01,0.055641598701477044
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,4,4,power_law_1.01,0.05762303829193115
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,4,4,power_law_1.01,0.06017407894134521
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,4,4,balanced,0.12589183807373047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,4,4,balanced,0.1691980743408203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,4,4,balanced,0.11339903831481934
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,4,4,balanced,0.21837951660156252
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,4,4,power_law_1.01,0.06615935802459717
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,4,4,power_law_1.01,0.09831423759460449
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,4,4,power_law_1.01,0.07363711833953858
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,4,4,power_law_1.01,0.0833676815032959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,4,4,power_law_1.2,0.02566015958786011
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,4,4,power_law_1.2,0.04359295845031738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,4,4,power_law_1.2,0.049419522285461426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,4,4,power_law_1.2,0.055928320884704584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,4,4,power_law_1.2,0.06438399791717529
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,4,4,power_law_1.2,0.07304192066192627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,4,4,power_law_1.2,0.0936729621887207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,4,4,power_law_1.2,0.09757823944091797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,4,4,power_law_1.2,0.10419712066650391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,4,4,power_law_1.2,0.10747520446777344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,4,4,power_law_1.2,0.10606080055236818
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,4,4,power_law_1.2,0.10709376335144043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,4,4,power_law_1.2,0.11297408103942871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,4,4,power_law_1.2,0.11993344306945801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,4,4,power_law_1.2,0.12157055854797363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,4,4,power_law_1.2,0.1256486415863037
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,4,4,power_law_1.2,0.13384832382202147
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,4,4,power_law_1.2,0.16198783874511719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,4,4,power_law_1.2,0.1846988868713379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,4,4,power_law_1.2,0.245662727355957
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,4,4,power_law_1.01,0.12855551719665528
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,4,4,power_law_1.01,0.15194496154785156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,4,4,power_law_1.2,0.2790220832824707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,4,4,power_law_1.2,0.36877185821533204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,4,4,power_law_1.2,0.47344001770019534
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,4,4,power_law_1.2,0.6487296295166016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,4,4,power_law_1.01,0.2117478370666504
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,4,4,power_law_1.01,0.27483264923095707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,4,4,power_law_1.2,0.8362624359130859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,4,4,power_law_1.2,0.032561919689178466
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,4,4,power_law_1.2,0.03727999925613403
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,4,4,power_law_1.2,0.04779007911682129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,4,4,power_law_1.2,0.054928641319274905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,4,4,power_law_1.2,1.228953628540039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,4,4,power_law_1.2,0.02764672040939331
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,4,4,power_law_1.2,0.05937535762786865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,4,4,power_law_1.2,0.0745638418197632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,4,4,power_law_1.2,0.07715072154998778
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,4,4,power_law_1.2,0.08246272087097169
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,4,4,power_law_1.2,0.08321791648864746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,4,4,power_law_1.2,0.08104191780090332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,4,4,power_law_1.2,0.08750335693359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,4,4,power_law_1.2,0.0915993595123291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,4,4,power_law_1.2,0.09553664207458497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,4,4,power_law_1.2,0.09765888214111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,4,4,power_law_1.2,0.10049152374267578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,4,4,power_law_1.2,0.10462719917297363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,4,4,power_law_1.2,0.11730943679809572
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,4,4,power_law_1.2,0.12922240257263184
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,4,4,power_law_1.2,0.15651328086853028
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,4,4,power_law_1.2,0.18428031921386717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,4,4,power_law_1.2,0.2456959915161133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,4,4,power_law_1.2,0.2888857650756836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,4,4,power_law_1.2,1.616575927734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,4,4,power_law_1.2,0.4139507293701172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,4,4,power_law_1.2,0.02908672094345093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,4,4,power_law_1.2,0.5207347106933594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,4,4,power_law_1.2,0.03053440093994141
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,4,4,power_law_1.2,0.03245311975479126
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,4,4,power_law_1.2,0.03624703884124756
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,4,4,power_law_1.2,0.728691177368164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,4,4,power_law_1.2,0.04264959812164307
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,4,4,power_law_1.2,0.04492288112640381
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,4,4,power_law_1.2,0.052878079414367674
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,4,4,power_law_1.2,0.053812479972839354
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,4,4,power_law_1.2,0.9675520324707032
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,4,4,power_law_1.2,0.050658559799194335
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,4,4,power_law_1.2,0.05315711975097657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,4,4,power_law_1.2,0.05552256107330322
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,4,4,power_law_1.2,0.055182080268859866
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,4,4,power_law_1.2,0.057918720245361335
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,4,4,power_law_1.2,0.059982080459594724
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,4,4,power_law_1.2,0.06393983840942383
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,4,4,power_law_1.2,0.050658559799194335
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,4,4,power_law_1.2,0.04852863788604737
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,4,4,power_law_1.01,0.5294323348999024
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,4,4,power_law_1.01,0.39117183685302737
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,4,4,power_law_1.2,0.08622207641601562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,4,4,power_law_1.2,0.07249023914337158
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,4,4,power_law_1.2,0.10552448272705077
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,4,4,power_law_1.2,0.11918848037719727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,8,2,balanced,0.02514175891876221
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,8,2,balanced,0.024832000732421873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,8,2,balanced,0.02681472063064575
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,8,2,balanced,0.047667198181152344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,8,2,balanced,0.0632371187210083
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,8,2,balanced,0.10699263572692871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,8,2,balanced,0.09971455574035645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,8,2,balanced,0.10088191986083986
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,4,4,power_law_1.2,0.17332735061645507
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,8,2,balanced,0.10112128257751465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,8,2,balanced,0.10577792167663573
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,8,2,balanced,0.13256192207336426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,8,2,balanced,0.10505727767944335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,8,2,balanced,0.1333887958526611
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,8,2,balanced,0.13533439636230468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,8,2,balanced,0.13803263664245607
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,8,2,balanced,0.14190336227416994
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,8,2,balanced,0.14827263832092286
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,8,2,balanced,0.1630886459350586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,4,4,power_law_1.2,0.22363264083862305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,8,2,balanced,0.1797875213623047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,8,2,balanced,0.20804607391357424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,8,2,balanced,0.2398361587524414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,8,2,balanced,0.3429644775390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,8,2,balanced,0.40562175750732427
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,8,2,balanced,0.5628492736816406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,8,2,balanced,0.027178239822387696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,8,2,balanced,0.7177664184570313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,8,2,balanced,0.028066558837890627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,8,2,balanced,0.031357440948486325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,8,2,balanced,0.041588478088378907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,8,2,balanced,0.06681856155395507
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,8,2,balanced,1.0407078552246094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,8,2,balanced,0.10396032333374024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,8,2,balanced,0.10460927963256836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,8,2,balanced,0.10568063735961915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,8,2,balanced,0.1060979175567627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,8,2,balanced,1.3581733703613281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,8,2,balanced,0.10700032234191895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,8,2,balanced,0.10797311782836913
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,8,2,balanced,0.12347904205322266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,8,2,balanced,0.12403200149536134
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,8,2,balanced,0.12631296157836913
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,8,2,balanced,0.1272537612915039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,8,2,balanced,0.1327347183227539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,8,2,balanced,0.13959808349609376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,8,2,balanced,0.1506009578704834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,8,2,balanced,0.16915456771850584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,8,2,balanced,0.1901055908203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,8,2,balanced,0.228720645904541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,8,2,balanced,0.2682393646240234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,8,2,balanced,0.4067238235473633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,8,2,balanced,0.48467582702636725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,8,2,balanced,0.12878975868225098
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,8,2,balanced,0.7008895874023438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,8,2,balanced,0.9113330841064453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,4,4,power_law_1.2,0.28802431106567383
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,4,4,power_law_1.2,0.46866687774658206
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,4,4,power_law_1.2,0.6717900848388672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,4,4,power_law_1.2,0.8431436920166016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,8,2,balanced,0.042004480361938476
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,8,2,balanced,0.03996543884277344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,8,2,balanced,0.07871103763580323
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,8,2,balanced,0.07890431880950928
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,8,2,balanced,0.04119040012359619
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,8,2,balanced,0.0787827205657959
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,8,2,balanced,0.07940351963043213
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,8,2,balanced,0.04593664169311523
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,8,2,balanced,0.08003711700439453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,8,2,balanced,0.08109439849853516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,8,2,balanced,0.081212158203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,8,2,balanced,0.08190208435058594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,8,2,balanced,0.08302207946777343
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,8,2,balanced,0.08930944442749024
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,8,2,balanced,0.09363712310791016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,8,2,balanced,0.09913215637207032
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,8,2,balanced,0.10971263885498046
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,8,2,balanced,0.1206118392944336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,8,2,balanced,0.058348798751831056
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,8,2,balanced,0.07840256214141846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,8,2,power_law_1.01,0.023613440990447997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,8,2,power_law_1.01,0.02682751893997192
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,8,2,power_law_1.01,0.04236159801483154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,8,2,power_law_1.01,0.05180799961090088
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,8,2,power_law_1.01,0.06033535957336426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,8,2,power_law_1.01,0.06803584098815918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,8,2,power_law_1.01,0.08321791648864746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,8,2,power_law_1.01,0.08431360244750977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,8,2,power_law_1.01,0.08702336311340332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,8,2,power_law_1.01,0.0873472023010254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,8,2,power_law_1.01,0.09221119880676269
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,8,2,power_law_1.01,0.1231987190246582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,8,2,power_law_1.01,0.1254092788696289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,8,2,power_law_1.01,0.13064703941345215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,8,2,power_law_1.01,0.1337113571166992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,8,2,power_law_1.01,0.1372428798675537
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,8,2,power_law_1.01,0.1458035182952881
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,8,2,power_law_1.01,0.1641164779663086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,8,2,power_law_1.01,0.18176000595092773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,8,2,power_law_1.01,0.23614463806152344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,8,2,power_law_1.01,0.2796428871154785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,8,2,power_law_1.01,0.3516147232055664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,8,2,balanced,0.07841536045074463
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,8,2,power_law_1.01,0.4335539245605469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,8,2,power_law_1.01,0.601399040222168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,8,2,power_law_1.01,0.7682393646240234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,8,2,power_law_1.01,1.0917094421386717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,8,2,power_law_1.01,0.026908159255981445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,8,2,power_law_1.01,1.4160716247558593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,8,2,power_law_1.01,0.03134720087051392
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,8,2,power_law_1.01,0.038846719264984134
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,8,2,power_law_1.01,0.04731391906738282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,8,2,power_law_1.01,0.0638476800918579
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,8,2,power_law_1.01,0.07105535984039306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,8,2,power_law_1.01,0.08945023536682128
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,8,2,power_law_1.01,0.08939135551452637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,8,2,power_law_1.01,0.09256959915161132
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,8,2,power_law_1.01,0.09311488151550293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,8,2,power_law_1.01,0.09693440437316894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,8,2,power_law_1.01,0.11222271919250489
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,8,2,power_law_1.01,0.11640576362609863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,8,2,power_law_1.01,0.11714943885803222
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,8,2,power_law_1.01,0.11967103958129884
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,8,2,balanced,0.19804672241210936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,8,2,power_law_1.01,0.11280256271362306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,8,2,power_law_1.01,0.12492927551269531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,8,2,power_law_1.01,0.14737664222717287
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,8,2,balanced,0.17520128250122072
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,8,2,power_law_1.01,0.1746188735961914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,8,2,power_law_1.01,0.2079539108276367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,8,2,power_law_1.01,0.3176780891418457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,8,2,power_law_1.01,0.25986944198608397
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,8,2,power_law_1.01,0.4310412979125976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,8,2,power_law_1.01,0.544395523071289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,8,2,power_law_1.01,0.13631360054016112
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,8,2,power_law_1.01,0.7523072052001953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,8,2,power_law_1.01,0.042617602348327635
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,8,2,power_law_1.01,0.041221117973327635
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,8,2,power_law_1.01,0.046545920372009275
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,8,2,balanced,0.2754060745239258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,8,2,balanced,0.35314815521240234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,8,2,power_law_1.01,0.04975615978240967
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,8,2,power_law_1.01,0.05514624118804932
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,8,2,power_law_1.01,0.06027391910552978
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,8,2,power_law_1.01,0.06869631767272949
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,8,2,power_law_1.01,0.07024511814117432
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,8,2,power_law_1.01,0.9852569580078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,8,2,power_law_1.01,0.07210112094879151
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,8,2,power_law_1.01,0.07409408092498779
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,8,2,power_law_1.01,0.07660416126251221
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,8,2,power_law_1.01,0.07583231925964355
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,8,2,power_law_1.01,0.07916287899017334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,8,2,power_law_1.01,0.081976318359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,8,2,power_law_1.01,0.084017915725708
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,8,2,power_law_1.01,0.08881407737731933
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,8,2,power_law_1.01,0.06932991981506348
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,8,2,power_law_1.01,0.0991487979888916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,8,2,power_law_1.01,0.13585280418395995
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,8,2,power_law_1.01,0.20141183853149416
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,8,2,power_law_1.01,0.10951935768127441
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,8,2,power_law_1.01,0.16032255172729493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,8,2,power_law_1.2,0.02312704086303711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,8,2,power_law_1.2,0.026044158935546874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,8,2,power_law_1.2,0.040245761871337896
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,8,2,power_law_1.01,0.2556812858581543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,8,2,power_law_1.2,0.049687042236328124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,8,2,power_law_1.2,0.058325757980346685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,8,2,power_law_1.2,0.06490240097045899
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,8,2,power_law_1.2,0.08193920135498048
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,8,2,power_law_1.2,0.0867903995513916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,8,2,power_law_1.2,0.08423680305480957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,8,2,power_law_1.2,0.08865792274475097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,8,2,power_law_1.2,0.0916198444366455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,8,2,power_law_1.2,0.12263808250427247
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,8,2,power_law_1.2,0.12580224037170412
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,8,2,power_law_1.2,0.13144319534301757
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,8,2,power_law_1.2,0.1343564796447754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,8,2,power_law_1.2,0.13866751670837402
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,8,2,power_law_1.2,0.14669055938720704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,8,2,power_law_1.2,0.1649817657470703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,8,2,power_law_1.2,0.18519552230834962
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,8,2,power_law_1.2,0.246297607421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,8,2,power_law_1.2,0.2808243179321289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,8,2,power_law_1.2,0.3547660827636719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,8,2,power_law_1.2,0.4454975891113281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,8,2,power_law_1.2,0.6033689498901367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,8,2,power_law_1.2,0.7791410827636719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,8,2,power_law_1.2,1.1095577239990235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,8,2,power_law_1.2,0.0286297607421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,8,2,power_law_1.2,0.03182463884353638
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,8,2,power_law_1.2,0.03630592107772827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,8,2,power_law_1.2,0.04622719764709472
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,8,2,power_law_1.2,0.06162816047668457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,8,2,power_law_1.2,1.4607180786132812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,8,2,power_law_1.2,0.06721920013427735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,8,2,power_law_1.2,0.08755583763122558
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,8,2,power_law_1.2,0.08666751861572267
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,8,2,power_law_1.2,0.09010944366455079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,8,2,power_law_1.2,0.09413120269775391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,8,2,power_law_1.2,0.11182975769042969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,8,2,power_law_1.2,0.11413887977600097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,8,2,power_law_1.2,0.11591679573059081
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,8,2,power_law_1.2,0.11660544395446777
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,8,2,power_law_1.2,0.12070015907287597
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,8,2,power_law_1.2,0.12782719612121582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,8,2,power_law_1.2,0.13985280036926268
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,8,2,power_law_1.2,0.15666175842285157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,8,2,power_law_1.01,0.3357555389404297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,8,2,power_law_1.2,0.19054079055786133
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,8,2,power_law_1.01,0.4514751815795899
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,8,2,power_law_1.2,0.274202880859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,8,2,power_law_1.2,0.21769983291625977
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,8,2,power_law_1.01,0.6772083282470703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,8,2,power_law_1.2,0.32939647674560546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,8,2,power_law_1.2,0.4474905776977539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,8,2,power_law_1.2,0.5608025741577148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,8,2,power_law_1.2,0.0961023998260498
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,8,2,power_law_1.2,0.04115456104278564
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,8,2,power_law_1.2,0.042488322257995606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,8,2,power_law_1.2,0.7715968322753907
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,8,2,power_law_1.2,0.04534143924713135
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,8,2,power_law_1.2,0.05465472221374512
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,8,2,power_law_1.2,0.05868031978607178
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,8,2,power_law_1.2,0.06786943912506103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,8,2,power_law_1.2,1.0052595520019532
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,8,2,power_law_1.01,0.9833689880371093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,8,2,power_law_1.2,0.06956160068511963
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,8,2,power_law_1.2,0.06966911792755127
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,8,2,power_law_1.2,0.07145088195800782
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,8,2,power_law_1.2,0.07456639766693116
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,8,2,power_law_1.2,0.07769216060638427
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,8,2,power_law_1.2,0.08089088439941407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,8,2,power_law_1.2,0.08435839653015137
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,8,2,power_law_1.2,0.08758272171020508
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,8,2,power_law_1.2,0.09585023880004882
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,8,2,power_law_1.2,0.11101951599121093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,8,2,power_law_1.2,0.048531198501586915
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,8,2,power_law_1.2,0.07695871829986572
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,8,2,power_law_1.2,0.16514175415039062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,8,2,power_law_1.2,0.2520921516418457
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,8,2,power_law_1.2,0.12485247611999513
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,8,2,power_law_1.2,0.19470848083496092
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,8,2,power_law_1.2,0.36853248596191407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,16,1,balanced,0.034359040260314944
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,16,1,balanced,0.046945919990539556
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,8,2,power_law_1.2,0.5432294464111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,16,1,balanced,0.05997759819030761
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,16,1,balanced,0.08270272254943847
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,16,1,balanced,0.13250240325927734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,16,1,balanced,0.1316044807434082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,16,1,balanced,0.1357996845245361
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,16,1,balanced,0.13479424476623536
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,16,1,balanced,0.13694592475891115
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,16,1,balanced,0.13732031822204588
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,16,1,balanced,0.1413663959503174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,16,1,balanced,0.03292479991912842
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,16,1,balanced,0.14249792098999023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,16,1,balanced,0.1476332759857178
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,16,1,balanced,0.18655807495117188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,16,1,balanced,0.18974464416503906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,16,1,balanced,0.19543039321899414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,16,1,balanced,0.20900032043457034
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,16,1,balanced,0.2236479949951172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,16,1,balanced,0.24995136260986328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,16,1,balanced,0.2798886489868164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,16,1,balanced,0.3827814483642578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,16,1,balanced,0.4408671951293945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,16,1,balanced,0.5982559967041016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,16,1,balanced,0.7519168090820313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,16,1,balanced,0.03410048007965088
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,16,1,balanced,0.03859904050827027
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,16,1,balanced,1.0721868896484374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,16,1,balanced,0.04390399932861328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,16,1,balanced,0.0570304012298584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,16,1,balanced,0.0875494384765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,16,1,balanced,0.15031423568725585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,16,1,balanced,1.3910208129882813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,16,1,balanced,0.15088831901550293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,16,1,balanced,0.15244928359985352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,16,1,balanced,0.15295167922973635
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,16,1,balanced,0.1534444808959961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,16,1,balanced,0.15517056465148926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,16,1,balanced,0.15760895729064942
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,16,1,balanced,0.15907391548156738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,16,1,balanced,0.16282175064086915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,16,1,balanced,0.18179071426391602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,16,1,balanced,0.18317184448242188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,16,1,balanced,0.1878995132446289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,16,1,balanced,0.19467071533203123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,16,1,balanced,0.20388608932495117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,16,1,balanced,0.2189983940124512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,16,1,balanced,0.23676223754882814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,16,1,balanced,0.2709702491760254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,16,1,balanced,0.3061702346801758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,16,1,balanced,0.47509502410888677
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,16,1,balanced,0.5441836929321289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,16,1,balanced,0.7820972442626953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,16,1,balanced,1.0172185516357422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,8,2,power_law_1.2,0.7627238464355469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,8,2,power_law_1.2,1.2960115051269532
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,8,2,power_law_1.2,1.5844659423828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,16,1,balanced,0.06682688236236572
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,16,1,balanced,0.06826303958892822
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,16,1,balanced,0.1293881607055664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,16,1,balanced,0.0705721616744995
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,16,1,balanced,0.1294489574432373
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,16,1,balanced,0.1306265640258789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,16,1,balanced,0.13021120071411132
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,16,1,balanced,0.13178303718566894
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,16,1,balanced,0.13324031829833985
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,16,1,balanced,0.1329811191558838
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,16,1,balanced,0.13480256080627442
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,16,1,balanced,0.1367366409301758
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,16,1,balanced,0.13838656425476076
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,16,1,balanced,0.14643327713012694
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,16,1,balanced,0.14983360290527342
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,16,1,balanced,0.16006336212158204
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,16,1,balanced,0.07652991771697998
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,16,1,balanced,0.17116416931152342
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,16,1,balanced,0.19291776657104492
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,16,1,balanced,0.2140518379211426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,16,1,power_law_1.01,0.031226239204406737
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,16,1,balanced,0.09524928092956543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,16,1,power_law_1.01,0.03368256092071533
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,16,1,balanced,0.12889280319213867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,16,1,power_law_1.01,0.047599358558654783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,16,1,power_law_1.01,0.061076478958129884
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,16,1,power_law_1.01,0.07390592098236085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,16,1,power_law_1.01,0.09152768135070802
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,16,1,power_law_1.01,0.10852160453796386
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,16,1,power_law_1.01,0.113755521774292
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,16,1,power_law_1.01,0.11807616233825684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,16,1,power_law_1.01,0.11786944389343261
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,16,1,power_law_1.01,0.12451775550842285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,16,1,power_law_1.01,0.13892736434936523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,16,1,power_law_1.01,0.1343174362182617
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,16,1,balanced,0.12948415756225587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,16,1,power_law_1.01,0.14861760139465333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,16,1,power_law_1.01,0.17987007141113281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,16,1,power_law_1.01,0.18434560775756836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,16,1,power_law_1.01,0.19167552947998048
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,16,1,power_law_1.01,0.20756736755371094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,16,1,power_law_1.01,0.22702400207519532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,16,1,power_law_1.01,0.2712819290161133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,16,1,power_law_1.01,0.32265281677246094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,16,1,power_law_1.01,0.38666431427001957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,16,1,power_law_1.01,0.4714406585693359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,16,1,power_law_1.01,0.6285529708862304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,16,1,power_law_1.01,0.7907762908935547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,16,1,power_law_1.01,0.034508800506591795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,16,1,power_law_1.01,1.1050214385986328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,16,1,power_law_1.01,0.03880831956863403
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,16,1,power_law_1.01,0.045272960662841796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,16,1,power_law_1.01,0.060706558227539065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,16,1,power_law_1.01,0.07811391830444336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,16,1,power_law_1.01,0.10589887619018554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,16,1,power_law_1.01,1.424962615966797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,16,1,power_law_1.01,0.13131839752197266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,16,1,power_law_1.01,0.13487744331359863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,16,1,power_law_1.01,0.13755200386047364
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,16,1,power_law_1.01,0.1379360008239746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,16,1,power_law_1.01,0.14177727699279785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,16,1,power_law_1.01,0.14583807945251465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,16,1,power_law_1.01,0.1483129596710205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,16,1,power_law_1.01,0.1515782356262207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,16,1,power_law_1.01,0.17129024505615234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,16,1,power_law_1.01,0.17230911254882814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,16,1,power_law_1.01,0.17803583145141602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,16,1,power_law_1.01,0.18801919937133788
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,16,1,power_law_1.01,0.2035955238342285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,16,1,power_law_1.01,0.23040960311889647
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,16,1,power_law_1.01,0.26396799087524414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,16,1,power_law_1.01,0.3201055908203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,16,1,power_law_1.01,0.3753126525878906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,16,1,power_law_1.01,0.4917728042602539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,16,1,power_law_1.01,0.6206547164916992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,16,1,power_law_1.01,0.8593478393554687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,16,1,power_law_1.01,1.0975475311279297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,16,1,power_law_1.01,0.06684288024902343
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,16,1,power_law_1.01,0.06806911945343017
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,16,1,power_law_1.01,0.07382080078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,16,1,power_law_1.01,0.08906304359436035
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,16,1,power_law_1.01,0.09644415855407715
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,16,1,power_law_1.01,0.07888000011444092
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,16,1,power_law_1.01,0.11103360176086426
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,16,1,power_law_1.01,0.1142240047454834
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,16,1,power_law_1.01,0.11621055603027344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,16,1,power_law_1.01,0.11765567779541015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,16,1,power_law_1.01,0.12205696105957031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,16,1,power_law_1.01,0.12452287673950195
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,16,1,power_law_1.01,0.1265983963012695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,16,1,power_law_1.01,0.13036928176879883
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,16,1,power_law_1.01,0.13701184272766115
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,16,1,power_law_1.01,0.15186495780944825
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,16,1,power_law_1.01,0.14124863624572753
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,16,1,balanced,0.3696588897705078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,16,1,balanced,0.6872064208984374
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,16,1,balanced,0.5297343826293945
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,16,1,power_law_1.01,0.17182144165039062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,16,1,power_law_1.01,0.23863296508789061
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,16,1,power_law_1.01,0.1933228874206543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,16,1,power_law_1.01,0.2838297653198242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,16,1,power_law_1.2,0.028487679958343504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,16,1,power_law_1.2,0.033697919845581056
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,16,1,power_law_1.01,0.48443519592285156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,16,1,power_law_1.2,0.04888000011444092
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,16,1,power_law_1.01,0.38110591888427736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,16,1,power_law_1.2,0.07019711971282959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,16,1,power_law_1.2,0.08794048309326172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,16,1,power_law_1.2,0.10834303855895996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,16,1,power_law_1.2,0.11139072418212889
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,16,1,power_law_1.2,0.11477824211120606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,16,1,power_law_1.2,0.11922304153442383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,16,1,power_law_1.2,0.12644288063049317
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,16,1,power_law_1.2,0.059936637878417964
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,16,1,power_law_1.2,0.13452799797058107
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,16,1,power_law_1.2,0.13996095657348634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,16,1,power_law_1.2,0.18193088531494142
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,16,1,power_law_1.2,0.1851091194152832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,16,1,power_law_1.2,0.19420927047729492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,16,1,power_law_1.2,0.21089920043945312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,16,1,power_law_1.2,0.2302118492126465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,16,1,power_law_1.2,0.29248191833496096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,16,1,power_law_1.2,0.3255852890014649
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,16,1,power_law_1.2,0.15058431625366211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,16,1,power_law_1.2,0.39100414276123047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,16,1,power_law_1.2,0.4771942520141602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,16,1,power_law_1.2,0.6315744018554688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,16,1,power_law_1.2,0.7960332489013672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,16,1,power_law_1.01,0.6924288177490234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,16,1,power_law_1.2,0.03409984111785889
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,16,1,power_law_1.2,0.0456601619720459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,16,1,power_law_1.2,1.1103692626953126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,16,1,power_law_1.2,0.05807168006896972
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,16,1,power_law_1.2,0.07624256134033203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,16,1,power_law_1.2,1.4315359497070312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,16,1,power_law_1.2,0.10140159606933594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,16,1,power_law_1.2,0.13038528442382813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,16,1,power_law_1.2,0.13326527595520019
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,16,1,power_law_1.2,0.13520447731018065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,16,1,power_law_1.2,0.13881535530090333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,16,1,power_law_1.2,0.1410604763031006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,16,1,power_law_1.2,0.1450534439086914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,16,1,power_law_1.2,0.1485855960845947
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,16,1,power_law_1.2,0.1532812786102295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,16,1,power_law_1.2,0.17193023681640623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,16,1,power_law_1.2,0.17595136642456055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,16,1,power_law_1.2,0.1808255958557129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,16,1,power_law_1.2,0.195231990814209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,16,1,power_law_1.2,0.20856447219848634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,16,1,power_law_1.2,0.2432307243347168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,16,1,power_law_1.2,0.2709625625610351
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,16,1,power_law_1.2,0.3283865737915039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,16,1,power_law_1.2,0.39585151672363283
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,16,1,power_law_1.2,0.4982553482055664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,16,1,power_law_1.2,0.6286937713623046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,16,1,power_law_1.2,0.8667628479003906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,16,1,power_law_1.2,0.06609280109405517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,16,1,power_law_1.2,1.108035202026367
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,16,1,power_law_1.2,0.06849792003631591
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,16,1,power_law_1.2,0.07320447921752929
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,16,1,power_law_1.2,0.07802688121795655
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,16,1,power_law_1.2,0.08650303840637207
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,16,1,power_law_1.2,0.09493120193481445
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,16,1,power_law_1.2,0.11026944160461424
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,16,1,power_law_1.2,0.11364095687866212
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,16,1,power_law_1.01,0.9059916687011718
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,16,1,power_law_1.2,0.11546751976013184
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,16,1,power_law_1.2,0.11843775749206542
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,16,1,power_law_1.2,0.12182463645935057
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,16,1,power_law_1.2,0.12597248077392578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,16,1,power_law_1.2,0.13064640045166015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,16,1,power_law_1.2,0.13671680450439455
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,16,1,power_law_1.01,1.330055694580078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,16,1,power_law_1.2,0.14266367912292482
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,16,1,power_law_1.2,0.03664576053619385
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,16,1,power_law_1.2,0.15080448150634768
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,16,1,power_law_1.2,0.16536767959594728
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,16,1,power_law_1.01,1.8248268127441407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,16,1,power_law_1.2,0.2292608070373535
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,16,1,power_law_1.2,0.1968204879760742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,16,1,power_law_1.2,0.30302528381347654
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,16,1,power_law_1.2,0.5486003112792969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,16,1,power_law_1.2,0.3768960189819336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,16,1,power_law_1.2,0.720165786743164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,16,1,power_law_1.2,1.109741439819336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,16,1,power_law_1.2,1.5265580749511718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,16,balanced,0.02753920078277588
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,16,balanced,0.02563711881637573
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,16,balanced,0.02716032028198242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,16,balanced,0.050335998535156246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,16,balanced,0.05771135807037353
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,16,balanced,0.09191807746887207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,16,balanced,0.10003968238830567
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,16,balanced,0.09515135765075684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,16,balanced,0.10324224472045898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,16,balanced,0.1026585578918457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,16,balanced,0.10626303672790527
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,16,balanced,0.11298560142517089
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,16,balanced,0.11745920181274414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,16,balanced,0.11059200286865234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,16,balanced,0.117326078414917
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,16,balanced,0.09430784225463867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,16,balanced,0.1509823989868164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,16,balanced,0.17264511108398437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,16,balanced,0.09439359664916992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,16,balanced,0.21954559326171874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,16,balanced,0.24429439544677733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,16,balanced,0.34178176879882816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,16,balanced,0.4128895950317383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,16,balanced,0.5846041488647461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,16,balanced,0.09843839645385742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,16,balanced,0.7578688049316407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,16,balanced,0.10660351753234862
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,16,1,power_law_1.2,2.4173120117187503
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,16,1,power_law_1.2,3.392023620605469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,1,16,balanced,0.01950719952583313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,1,16,balanced,0.019837440252304078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,1,16,balanced,0.02803584098815918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,1,16,balanced,0.019856640100479127
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,1,16,balanced,0.02848639965057373
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,1,16,balanced,0.027601919174194335
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,1,16,balanced,0.020627200603485107
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,1,16,balanced,0.027955200672149662
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,1,16,balanced,0.028728320598602298
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,1,16,balanced,0.02847232103347778
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,1,16,balanced,0.02851455926895142
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,1,16,balanced,0.028456959724426266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,1,16,balanced,0.028944640159606932
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,1,16,balanced,0.029282560348510744
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,1,16,balanced,0.029040639400482175
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,1,16,balanced,0.03028736114501953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,1,16,balanced,0.03123456001281738
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,1,16,balanced,0.024021759033203124
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,1,16,balanced,0.029772799015045166
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,1,16,balanced,0.029527039527893068
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,16,power_law_1.01,0.05452159881591797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,1,16,balanced,0.02678656101226807
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,1,16,balanced,0.045932798385620116
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,1,16,balanced,0.06279168128967286
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,1,16,balanced,0.04947199821472168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,16,power_law_1.01,0.059631361961364746
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,1,16,balanced,0.05860352039337158
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,16,power_law_1.01,0.0623744010925293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,16,power_law_1.01,0.08498815536499024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,16,power_law_1.01,0.09274368286132813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,16,power_law_1.01,0.08994688034057617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,16,power_law_1.01,0.09922047615051269
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,16,power_law_1.01,0.09694208145141602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,16,power_law_1.01,0.10377087593078613
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,16,power_law_1.01,0.10287232398986816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,16,power_law_1.01,0.10552319526672363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,16,power_law_1.01,0.11127679824829102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,16,power_law_1.01,0.11306367874145509
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,16,power_law_1.01,0.11326080322265626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,16,power_law_1.01,0.11330944061279298
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,16,power_law_1.01,0.1500057601928711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,16,power_law_1.01,0.1682022476196289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,16,power_law_1.01,0.21260671615600585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,16,power_law_1.01,0.25690879821777346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,16,power_law_1.01,0.3479654312133789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,16,power_law_1.01,0.09544447898864747
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,16,power_law_1.01,0.4886515045166016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,16,power_law_1.01,0.6824230194091797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,1,16,balanced,0.10272383689880371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,16,power_law_1.01,0.827447052001953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,1,16,power_law_1.01,0.02764672040939331
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,1,16,balanced,0.08437760353088379
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,1,16,power_law_1.01,0.027174398899078366
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,1,16,power_law_1.01,0.02395008087158203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,1,16,power_law_1.01,0.026997759342193606
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,1,16,power_law_1.01,0.026847999095916748
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,1,16,power_law_1.01,0.026392319202423097
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,1,16,power_law_1.01,0.027534079551696778
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,1,16,power_law_1.01,0.026009600162506103
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,1,16,power_law_1.01,0.028090879917144772
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,1,16,power_law_1.01,0.027709441184997557
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,1,16,power_law_1.01,0.027896320819854735
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,1,16,power_law_1.01,0.027957758903503417
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,1,16,power_law_1.01,0.024394240379333496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,16,power_law_1.01,1.630135040283203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,1,16,power_law_1.01,0.028028159141540526
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,1,16,power_law_1.01,0.028816640377044678
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,1,16,power_law_1.01,0.029040639400482175
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,1,16,power_law_1.01,0.030982398986816408
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,1,16,power_law_1.01,0.03615744113922119
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,1,16,power_law_1.01,0.046799359321594236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,16,power_law_1.01,0.05679488182067871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,16,power_law_1.01,1.3003826904296876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,1,16,power_law_1.01,0.03895807981491089
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,1,16,power_law_1.01,0.05075712203979492
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,1,16,power_law_1.01,0.06189184188842774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,16,power_law_1.2,0.059269118309021
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,1,16,power_law_1.01,0.11167743682861328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,16,power_law_1.2,0.060835838317871094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,1,16,power_law_1.01,0.08311936378479004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,16,power_law_1.2,0.054996480941772466
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,16,power_law_1.2,0.08718079566955567
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,16,power_law_1.2,0.09023743629455566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,16,power_law_1.2,0.09088640213012696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,16,power_law_1.2,0.09734656333923339
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,16,power_law_1.2,0.09904000282287598
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,16,power_law_1.2,0.10313471794128418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,16,power_law_1.2,0.10537599563598632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,16,power_law_1.2,0.10978943824768066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,16,power_law_1.2,0.10815487861633302
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,16,power_law_1.2,0.11397760391235352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,16,power_law_1.2,0.112161283493042
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,16,power_law_1.2,0.11730815887451171
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,1,16,power_law_1.01,0.1347481632232666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,16,power_law_1.2,0.1601535987854004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,16,power_law_1.2,0.18001279830932618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,16,power_law_1.2,0.09430527687072754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,16,power_law_1.2,0.24876672744750977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,16,power_law_1.2,0.2797158432006836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,16,power_law_1.2,0.4478668975830078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,16,power_law_1.2,0.5987788772583007
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,1,16,power_law_1.2,0.024235520362854004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,16,power_law_1.2,0.05610112190246582
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,1,16,power_law_1.01,0.23397632598876955
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,1,16,power_law_1.2,0.026414079666137697
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,1,16,power_law_1.2,0.027170560359954833
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,1,16,power_law_1.2,0.027005441188812256
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,1,16,power_law_1.2,0.026009600162506103
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,1,16,power_law_1.2,0.02724992036819458
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,1,16,power_law_1.2,0.026012160778045655
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,1,16,power_law_1.2,0.02605567932128906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,1,16,power_law_1.2,0.027246079444885253
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,1,16,power_law_1.2,0.026607360839843754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,16,power_law_1.2,1.571032257080078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,1,16,power_law_1.2,0.02779263973236084
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,1,16,power_law_1.2,0.02879744052886963
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,16,power_law_1.2,1.021934051513672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,1,16,power_law_1.2,0.02726016044616699
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,1,16,power_law_1.2,0.027718400955200194
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,1,16,power_law_1.2,0.029089279174804687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,1,16,power_law_1.01,0.2829094314575195
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,1,16,power_law_1.2,0.052872958183288576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,16,power_law_1.2,0.8485298919677735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,16,power_law_1.2,2.231994934082031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,1,16,power_law_1.2,0.03145087957382202
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,1,16,power_law_1.2,0.03785471916198731
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,1,16,power_law_1.2,0.04525440216064453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,1,16,power_law_1.2,0.0590284776687622
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,1,16,power_law_1.2,0.03305216073989868
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,1,16,power_law_1.2,0.08390399932861328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,1,16,power_law_1.2,0.11565952301025391
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,1,16,power_law_1.2,0.1682931137084961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,8,balanced,0.023552000522613525
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,8,balanced,0.023490560054779053
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,8,balanced,0.0866982364654541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,8,balanced,0.08883968353271485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,8,balanced,0.09019136428833008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,8,balanced,0.024929280281066897
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,8,balanced,0.09085184097290039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,8,balanced,0.09249792098999024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,8,balanced,0.09294976234436035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,8,balanced,0.10094719886779786
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,8,balanced,0.10574848175048827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,8,balanced,0.09377535820007324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,8,balanced,0.09886591911315919
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,8,balanced,0.09509504318237305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,8,balanced,0.053091840744018556
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,8,balanced,0.10009344100952149
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,8,balanced,0.11058303833007813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,8,balanced,0.12196736335754395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,8,balanced,0.1659916877746582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,8,balanced,0.18197376251220704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,8,balanced,0.25528831481933595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,8,balanced,0.29494016647338867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,8,balanced,0.41284225463867186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,8,balanced,0.041521921157836914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,8,balanced,0.5136857604980468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,8,balanced,0.737017593383789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,8,balanced,0.9673894500732422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,1,16,power_law_1.2,0.28242687225341795
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,1,16,power_law_1.2,0.22045312881469728
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,1,16,power_law_1.2,0.41637248992919923
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,2,8,balanced,0.02027776002883911
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,2,8,balanced,0.0210099196434021
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,2,8,balanced,0.03125760078430176
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,2,8,balanced,0.031820800304412845
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,2,8,balanced,0.03139328002929688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,2,8,balanced,0.03191168069839477
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,2,8,balanced,0.021651198863983152
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,2,8,balanced,0.031596798896789555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,2,8,balanced,0.03223680019378662
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,2,8,balanced,0.031718399524688724
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,2,8,balanced,0.03209088087081909
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,2,8,balanced,0.032330238819122316
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,2,8,balanced,0.032967679500579834
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,2,8,balanced,0.021087999343872073
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,2,8,balanced,0.033269760608673093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,2,8,balanced,0.035659520626068114
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,2,8,balanced,0.03608319997787476
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,2,8,balanced,0.027226879596710208
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,2,8,balanced,0.031146240234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,2,8,balanced,0.0308787202835083
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,2,8,balanced,0.03069567918777466
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,8,power_law_1.01,0.0488102388381958
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,8,power_law_1.01,0.04538623809814453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,2,8,balanced,0.057392640113830565
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,2,8,balanced,0.07619711875915527
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,2,8,balanced,0.07091968059539795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,8,power_law_1.01,0.05386367797851562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,8,power_law_1.01,0.08223103523254395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,8,power_law_1.01,0.0831167984008789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,2,8,balanced,0.05376895904541016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,8,power_law_1.01,0.08839679718017578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,8,power_law_1.01,0.08689791679382323
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,8,power_law_1.01,0.08926976203918457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,8,power_law_1.01,0.09461376190185547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,8,power_law_1.01,0.092871675491333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,8,power_law_1.01,0.09936511993408204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,8,power_law_1.01,0.09543423652648926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,8,power_law_1.01,0.09526271820068359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,8,power_law_1.01,0.1051865577697754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,8,power_law_1.01,0.12067456245422363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,8,power_law_1.01,0.13833344459533692
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,8,power_law_1.01,0.18779136657714843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,8,power_law_1.01,0.22632959365844724
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,8,power_law_1.01,0.3052044868469238
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,8,power_law_1.01,0.37937278747558595
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,2,8,balanced,0.11325823783874511
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,8,power_law_1.01,0.0959603214263916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,8,power_law_1.01,0.5341747283935547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,2,8,power_law_1.01,0.02184319972991943
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,8,power_law_1.01,0.6583296203613281
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,2,8,power_law_1.01,0.026736640930175783
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,2,8,power_law_1.01,0.02893440008163452
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,2,8,power_law_1.01,0.028206079006195067
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,2,8,power_law_1.01,0.029282560348510744
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,2,8,power_law_1.01,0.030411520004272462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,8,power_law_1.01,0.9890022277832031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,8,power_law_1.01,1.0682828521728516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,2,8,power_law_1.01,0.03069567918777466
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,2,8,power_law_1.01,0.031560959815979006
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,2,8,power_law_1.01,0.03052799940109253
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,2,8,power_law_1.01,0.031587839126586914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,2,8,power_law_1.01,0.0315775990486145
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,2,8,power_law_1.01,0.03223423957824707
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,2,8,power_law_1.01,0.03166591882705688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,2,8,power_law_1.01,0.03312639951705933
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,2,8,power_law_1.01,0.032576000690460204
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,2,8,power_law_1.01,0.026263039112091064
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,2,8,power_law_1.01,0.03787008047103882
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,8,power_law_1.01,0.03662847995758057
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,8,power_law_1.01,0.05472896099090576
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,2,8,balanced,0.13157888412475588
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,2,8,power_law_1.01,0.03551232099533081
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,2,8,power_law_1.01,0.04169856071472168
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,2,8,power_law_1.01,0.04995456218719482
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,2,8,power_law_1.01,0.09531135559082031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,8,power_law_1.2,0.036587519645690916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,2,8,power_law_1.01,0.07080832004547119
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,2,8,power_law_1.01,0.05870463848114014
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,2,8,power_law_1.01,0.12436991691589357
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,8,power_law_1.2,0.08010368347167969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,8,power_law_1.2,0.08332159996032715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,8,power_law_1.2,0.0863321590423584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,8,power_law_1.2,0.0887116813659668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,8,power_law_1.2,0.09098624229431153
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,8,power_law_1.2,0.045898241996765135
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,8,power_law_1.2,0.09373951911926269
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,8,power_law_1.2,0.09697024345397949
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,8,power_law_1.2,0.09666560173034669
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,8,power_law_1.2,0.09486975669860839
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,8,power_law_1.2,0.09714816093444825
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,8,power_law_1.2,0.09943936347961427
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,8,power_law_1.2,0.10967424392700195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,8,power_law_1.2,0.1282367992401123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,8,power_law_1.2,0.14966272354125976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,8,power_law_1.2,0.20065919876098634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,8,power_law_1.2,0.25668352127075195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,8,power_law_1.2,0.04900735855102539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,8,power_law_1.2,0.33466110229492185
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,8,power_law_1.2,0.43760768890380863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,8,power_law_1.2,0.04077568054199219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,8,power_law_1.2,0.606115837097168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,8,power_law_1.2,0.7379917144775391
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,2,8,power_law_1.2,0.02166527986526489
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,2,8,power_law_1.2,0.02258687973022461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,8,power_law_1.2,1.5474124145507813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,2,8,power_law_1.2,0.024454400539398194
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,2,8,power_law_1.2,0.028157439231872562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,2,8,power_law_1.2,0.03005568027496338
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,2,8,power_law_1.2,0.02847359895706177
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,2,8,power_law_1.2,0.0301311993598938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,8,power_law_1.2,0.05614848136901855
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,2,8,power_law_1.2,0.030429439544677733
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,2,8,power_law_1.2,0.031063039302825928
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,2,8,power_law_1.2,0.030815999507904052
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,2,8,power_law_1.2,0.03042815923690796
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,2,8,power_law_1.2,0.031485440731048586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,2,8,power_law_1.2,0.031298561096191405
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,2,8,power_law_1.2,0.03214848041534424
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,2,8,power_law_1.2,0.032033278942108154
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,2,8,power_law_1.01,0.16955520629882812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,2,8,power_law_1.2,0.05134719848632813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,8,power_law_1.2,1.1021017456054687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,2,8,power_law_1.01,0.221081600189209
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,2,8,power_law_1.01,0.32039039611816406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,2,8,power_law_1.2,0.03466495990753174
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,2,8,power_law_1.2,0.03636480093002319
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,2,8,power_law_1.2,0.042388482093811033
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,2,8,power_law_1.2,0.04612224102020264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,4,4,balanced,0.023891201019287107
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,4,4,balanced,0.022293760776519775
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,4,4,balanced,0.025949440002441405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,4,4,balanced,0.04038400173187256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,4,4,balanced,0.054261760711669924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,4,4,balanced,0.06638463973999023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,4,4,balanced,0.08846336364746094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,4,4,balanced,0.0899443244934082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,4,4,balanced,0.09136511802673339
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,4,4,balanced,0.094652156829834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,4,4,balanced,0.10124544143676757
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,4,4,balanced,0.10447104454040528
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,2,8,power_law_1.2,0.0712985610961914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,4,4,balanced,0.09140992164611816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,4,4,balanced,0.09224320411682128
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,4,4,balanced,0.09564288139343262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,4,4,balanced,0.1006169605255127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,4,4,balanced,0.11148159980773927
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,4,4,balanced,0.12354175567626953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,4,4,balanced,0.169880313873291
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,2,8,power_law_1.2,0.09328512191772462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,4,4,balanced,0.18732543945312502
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,4,4,balanced,0.31358720779418947
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,4,4,balanced,0.45133056640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,4,4,balanced,0.0961945629119873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,4,4,balanced,0.5654438400268555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,4,4,balanced,0.8218316650390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,4,4,balanced,1.0786086273193358
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,4,4,balanced,0.26673791885375975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,4,4,power_law_1.01,0.03592063903808594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,4,4,power_law_1.01,0.03756671905517578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,4,4,power_law_1.01,0.047399678230285645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,4,4,power_law_1.01,0.0578495979309082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,2,8,power_law_1.2,0.12413951873779297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,2,8,power_law_1.2,0.17331584930419922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,4,4,power_law_1.01,0.08376192092895508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,4,4,power_law_1.01,0.024785919189453127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,4,4,power_law_1.01,0.08914688110351562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,4,4,power_law_1.01,0.09717887878417969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,4,4,power_law_1.01,0.09010432243347169
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,4,4,power_law_1.01,0.09129599571228028
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,4,4,power_law_1.01,0.08569984436035157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,4,4,power_law_1.01,0.09202303886413574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,4,4,power_law_1.01,0.09028351783752442
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,4,4,power_law_1.01,0.10133631706237793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,4,4,power_law_1.01,0.13148927688598633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,4,4,power_law_1.01,0.05033728122711182
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,4,4,power_law_1.01,0.1767616081237793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,4,4,power_law_1.01,0.27765119552612305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,4,4,power_law_1.01,0.09330431938171387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,4,4,power_law_1.01,0.3362496185302734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,4,4,power_law_1.01,0.5004108810424805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,4,4,power_law_1.01,0.6124607849121093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,4,4,power_law_1.01,0.09349504470825196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,4,4,power_law_1.01,0.11333888053894044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,4,4,power_law_1.2,0.024780800342559813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,4,4,power_law_1.2,0.03666048049926758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,4,4,power_law_1.2,0.037740800380706784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,4,4,power_law_1.01,0.905117416381836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,4,4,power_law_1.2,0.0448473596572876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,4,4,power_law_1.2,0.047886080741882324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,4,4,power_law_1.01,0.21107200622558592
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,4,4,power_law_1.2,0.08358783721923828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,4,4,power_law_1.2,0.08502143859863281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,4,4,power_law_1.2,0.08743167877197265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,4,4,power_law_1.2,0.09008255958557129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,4,4,power_law_1.01,1.245000991821289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,4,4,power_law_1.2,0.09463808059692383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,4,4,power_law_1.2,0.10037759780883788
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,4,4,power_law_1.2,0.08870911598205566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,4,4,power_law_1.2,0.09340543746948242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,4,4,power_law_1.2,0.09299967765808106
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,4,4,power_law_1.2,0.09647616386413574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,4,4,power_law_1.2,0.10207615852355958
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,4,4,power_law_1.2,0.114520320892334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,4,4,power_law_1.2,0.14173952102661133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,4,4,power_law_1.2,0.1833113670349121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,4,4,power_law_1.2,0.22450048446655274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,4,4,power_law_1.2,0.2886745643615723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,4,4,power_law_1.2,0.35820545196533204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,4,4,power_law_1.2,0.5283788681030274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,4,4,power_law_1.2,0.735061798095703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,4,4,power_law_1.2,0.9430912017822266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,4,4,power_law_1.2,0.06046720027923584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,4,4,power_law_1.2,1.29552001953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,2,8,power_law_1.2,0.3361407852172852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,8,2,balanced,0.0235481595993042
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,8,2,balanced,0.025525760650634766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,8,2,balanced,0.031357440948486325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,8,2,balanced,0.0854860782623291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,8,2,balanced,0.10804096221923828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,8,2,balanced,0.05544320106506347
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,8,2,balanced,0.11028608322143554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,8,2,balanced,0.1109119987487793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,8,2,balanced,0.14347904205322265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,8,2,balanced,0.14936832427978514
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,8,2,balanced,0.1524287986755371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,8,2,balanced,0.14395903587341308
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,8,2,balanced,0.15415424346923828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,8,2,balanced,0.1628646469116211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,8,2,balanced,0.17471488952636718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,8,2,balanced,0.1458995246887207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,8,2,balanced,0.2708518409729004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,8,2,balanced,0.28528255462646485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,8,2,balanced,0.4031462478637695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,8,2,balanced,0.4462220764160156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,8,2,balanced,0.14754048347473145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,8,2,balanced,0.15140735626220703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,8,2,balanced,0.6682508850097657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,2,8,power_law_1.2,0.427955207824707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,8,2,power_law_1.01,0.025259521007537845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,8,2,power_law_1.01,0.03179008007049561
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,8,2,balanced,0.8224947357177734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,8,2,power_law_1.01,0.09510656356811523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,8,2,balanced,1.1962611389160156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,8,2,power_law_1.01,0.09853568077087402
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,8,2,balanced,1.5685568237304688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,8,2,balanced,0.15019007682800295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,8,2,power_law_1.01,0.07197184085845947
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,8,2,power_law_1.01,0.10423551559448244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,8,2,power_law_1.01,0.1413542366027832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,8,2,power_law_1.01,0.059982080459594724
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,8,2,power_law_1.01,0.046350078582763674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,8,2,power_law_1.01,0.13329407691955567
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,8,2,power_law_1.01,0.16094207763671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,8,2,power_law_1.01,0.13496191978454591
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,8,2,power_law_1.01,0.1813158416748047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,8,2,power_law_1.01,0.21068031311035157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,8,2,power_law_1.01,0.25663616180419924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,8,2,power_law_1.01,0.13869824409484863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,8,2,power_law_1.01,0.32932479858398434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,8,2,power_law_1.01,0.4172108840942383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,8,2,power_law_1.01,0.1374988842010498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,8,2,power_law_1.01,0.5044160079956055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,8,2,power_law_1.01,0.7116595458984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,8,2,power_law_1.01,0.9692057800292968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,8,2,power_law_1.01,0.14526207923889162
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,8,2,power_law_1.2,0.03886591911315918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,8,2,power_law_1.01,0.1467097568511963
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,8,2,power_law_1.01,1.3620428466796874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,8,2,power_law_1.2,0.05575424194335937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,8,2,power_law_1.2,0.06583040237426759
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,8,2,power_law_1.2,0.09439231872558593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,8,2,power_law_1.01,1.7274278259277345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,8,2,power_law_1.2,0.13361920356750487
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,8,2,power_law_1.2,0.13814784049987794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,8,2,power_law_1.2,0.028094720840454102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,8,2,power_law_1.2,0.13911680221557615
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,8,2,power_law_1.2,0.1428121566772461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,8,2,power_law_1.2,0.14697471618652344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,8,2,power_law_1.2,0.15109375953674314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,8,2,power_law_1.2,0.0461952018737793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,8,2,power_law_1.2,0.1555840015411377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,8,2,power_law_1.2,0.16584575653076172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,8,2,power_law_1.2,0.18905088424682617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,8,2,power_law_1.2,0.2153228759765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,8,2,power_law_1.2,0.13426688194274902
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,8,2,power_law_1.2,0.10085247993469237
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,8,2,power_law_1.2,0.0982323169708252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,8,2,power_law_1.01,0.15208703994750977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,8,2,power_law_1.2,0.264846076965332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,8,2,power_law_1.2,0.42413566589355467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,8,2,power_law_1.2,0.3325056076049805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,8,2,power_law_1.2,0.5145702362060547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,8,2,power_law_1.2,0.7284095764160157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,8,2,power_law_1.2,1.0155980682373047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,8,2,power_law_1.2,1.403937225341797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,8,2,power_law_1.2,1.8621836853027343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,16,1,balanced,0.04519167900085449
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,16,1,balanced,0.07149184226989747
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,16,1,balanced,0.03082688093185425
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,16,1,balanced,0.14315008163452148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,16,1,balanced,0.03531968116760254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,16,1,balanced,0.14970303535461427
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,16,1,balanced,0.1441113567352295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,16,1,balanced,0.14297151565551758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,16,1,balanced,0.14430591583251953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,16,1,balanced,0.20215423583984374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,16,1,balanced,0.1124556827545166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,16,1,balanced,0.20212223052978517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,16,1,balanced,0.21284160614013672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,16,1,balanced,0.20523456573486326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,16,1,balanced,0.23305984497070314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,16,1,balanced,0.3465913772583008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,16,1,balanced,0.3647430419921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,16,1,balanced,0.1424729633331299
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,16,1,balanced,0.49511806488037113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,16,1,balanced,0.20371904373168945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,16,1,balanced,0.20707712173461915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,16,1,balanced,0.5361113739013672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,16,1,power_law_1.01,0.030396161079406736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,16,1,balanced,0.7927359771728516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,16,1,power_law_1.01,0.03453183889389038
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,16,1,power_law_1.01,0.05539968013763428
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,16,1,power_law_1.01,0.06976640224456787
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,16,1,power_law_1.01,0.11886015892028809
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,16,1,power_law_1.01,0.12184127807617187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,16,1,power_law_1.01,0.12634048461914063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,16,1,power_law_1.01,0.13165120124816893
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,16,1,balanced,0.22202688217163086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,16,1,power_law_1.01,0.1351846408843994
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,16,1,power_law_1.01,0.14638848304748536
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,16,1,power_law_1.01,0.19284543991088868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,16,1,balanced,1.8785824584960937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,16,1,power_law_1.01,0.19586240768432617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,16,1,power_law_1.01,0.20066112518310547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,16,1,power_law_1.01,0.20274431228637696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,16,1,power_law_1.01,0.2071116828918457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,16,1,power_law_1.01,0.2182636833190918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,16,1,balanced,1.4368173217773437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,16,1,power_law_1.01,0.24010047912597657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,16,1,power_law_1.01,0.26279296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,16,1,power_law_1.01,0.3313254547119141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,16,1,power_law_1.01,0.3762041473388672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,16,1,balanced,0.9675859069824219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,16,1,power_law_1.01,0.4855091094970703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,16,1,power_law_1.01,0.5998067092895508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,16,1,power_law_1.2,0.030740480422973632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,16,1,power_law_1.2,0.0383948802947998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,16,1,power_law_1.01,0.8021600341796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,16,1,power_law_1.2,0.055603837966918944
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,16,1,power_law_1.2,0.06610943794250487
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,16,1,power_law_1.01,0.08420479774475098
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,16,1,power_law_1.01,1.0301222229003906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,16,1,power_law_1.2,0.1172428798675537
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,16,1,power_law_1.2,0.12245439529418944
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,16,1,power_law_1.2,0.13250304222106934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,16,1,power_law_1.2,0.13968192100524904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,16,1,power_law_1.2,0.15031616210937498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,16,1,power_law_1.01,1.4673394775390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,16,1,power_law_1.2,0.19531007766723635
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,16,1,power_law_1.2,0.1982931137084961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,16,1,power_law_1.2,0.20448511123657226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,16,1,power_law_1.2,0.2060576057434082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,16,1,power_law_1.2,0.08151871681213378
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,16,1,power_law_1.01,1.9179827880859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,16,1,power_law_1.2,0.22214143753051757
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,16,1,power_law_1.2,0.24304895401000975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,16,1,power_law_1.2,0.2645561599731445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,16,1,power_law_1.2,0.3346540832519531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,16,1,power_law_1.2,0.12915391921997071
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,16,1,power_law_1.2,0.37798206329345707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,16,1,power_law_1.2,0.4890694427490234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,16,1,power_law_1.2,0.5979206466674805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,16,1,power_law_1.2,0.2119289588928223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,16,1,power_law_1.2,0.8033856201171876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,16,1,power_law_1.2,1.0302931213378907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,16,balanced,0.06467199802398682
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,16,balanced,0.07785215854644775
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,16,balanced,0.12239487648010254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,16,balanced,0.2357375907897949
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,16,1,power_law_1.2,1.4835583496093752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,16,balanced,0.3402048110961914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,16,balanced,0.4072396850585937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,16,balanced,0.4415628814697266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,16,balanced,0.4664896011352539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,16,balanced,0.4809267044067383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,16,balanced,0.5267123031616211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,16,balanced,0.5347635269165039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,16,balanced,0.5646323013305664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,16,balanced,0.5748825454711914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,16,balanced,0.591833610534668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,16,balanced,0.47896190643310543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,16,balanced,0.6384550476074219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,16,balanced,0.7082291412353515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,16,1,power_law_1.2,1.929431610107422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,16,balanced,0.8005785369873047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,16,balanced,1.156526107788086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,16,balanced,1.3113626098632811
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,16,balanced,1.9976588439941405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,16,balanced,0.4892723083496094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,16,balanced,2.4786714172363284
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,16,balanced,0.05261568069458008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,16,balanced,0.12410880088806153
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,16,balanced,0.037946879863739014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,16,balanced,0.1834547233581543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,16,balanced,3.8325323486328124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,16,balanced,0.22533504486083986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,16,balanced,0.23624191284179688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,16,balanced,0.2360601615905762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,16,balanced,0.24654464721679686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,16,balanced,0.2471232032775879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,16,balanced,0.25279104232788085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,16,balanced,0.26544384002685545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,16,balanced,0.29236991882324215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,16,balanced,0.3120947265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,16,balanced,0.2872780799865723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,16,balanced,0.3140620803833008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,16,balanced,4.970252685546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,16,balanced,0.3322995376586914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,16,balanced,0.3916147232055664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,16,balanced,0.45414272308349607
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,16,balanced,0.5431820678710937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,16,balanced,0.660684814453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,16,balanced,0.954287338256836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,16,balanced,1.1528832244873048
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,16,balanced,1.6872781372070313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,16,balanced,7.4608154296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,16,balanced,2.110556182861328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,16,balanced,3.1500006103515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,16,balanced,0.08065279960632324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,16,balanced,9.911527709960938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,16,balanced,4.090494689941407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,1,16,balanced,0.025436160564422605
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,1,16,balanced,0.12545408248901366
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,1,16,balanced,0.12567551612854005
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,1,16,balanced,0.032209920883178714
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,1,16,balanced,0.12597887992858886
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,1,16,balanced,0.1266150379180908
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,1,16,balanced,0.1258726406097412
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,1,16,balanced,0.12691967964172363
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,1,16,balanced,0.12647168159484862
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,1,16,balanced,0.12758784294128417
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,1,16,balanced,0.1269388771057129
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,1,16,balanced,0.048975358009338374
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,1,16,balanced,0.1289087963104248
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,1,16,balanced,0.1324185562133789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,1,16,balanced,0.13450495719909666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,1,16,balanced,0.14042240142822265
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,1,16,balanced,0.1453337574005127
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,1,16,balanced,0.07528831958770751
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,1,16,balanced,0.10750592231750489
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,1,16,balanced,0.12477567672729492
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,1,16,balanced,0.12405759811401366
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,16,power_law_1.01,0.28589439392089844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,16,power_law_1.01,0.38571647644042967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,16,power_law_1.01,0.373015022277832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,16,power_law_1.01,0.3902054214477539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,16,power_law_1.01,0.39548927307128906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,16,power_law_1.01,0.3480729675292969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,16,power_law_1.01,0.40818817138671876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,16,power_law_1.01,0.4261235046386719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,16,power_law_1.01,0.43420673370361323
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,16,power_law_1.01,0.45515007019042975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,16,power_law_1.01,0.4770943832397461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,16,power_law_1.01,0.4940671920776367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,16,power_law_1.01,0.5271129608154297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,16,power_law_1.01,0.5559552001953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,16,power_law_1.01,0.5833087921142578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,16,power_law_1.01,0.6501837158203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,16,power_law_1.01,0.7533990478515624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,16,power_law_1.01,0.8928396606445312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,16,power_law_1.01,0.42240127563476565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,16,power_law_1.01,1.1639014434814452
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,16,power_law_1.01,1.4289753723144532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,16,power_law_1.01,2.037451477050781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,16,power_law_1.01,2.5022962951660155
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,1,16,balanced,0.29540735244750976
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,1,16,balanced,0.1994393539428711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,1,16,balanced,0.21176191329956057
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,16,power_law_1.01,3.7731469726562503
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,16,power_law_1.01,0.20578048706054686
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,1,16,balanced,0.31547136306762696
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,1,16,balanced,0.582295036315918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,16,power_law_1.01,0.2061939239501953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,16,power_law_1.01,0.15783552169799803
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,16,power_law_1.01,0.20905471801757813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,1,16,balanced,0.4709721755981445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,16,power_law_1.01,0.21642751693725587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,16,power_law_1.01,0.2241062355041504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,16,power_law_1.01,0.23451520919799806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,16,power_law_1.01,0.2395865631103516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,16,power_law_1.01,0.24181503295898438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,16,power_law_1.01,0.24578687667846683
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,16,power_law_1.01,0.2614054489135742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,16,power_law_1.01,4.725934143066406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,16,power_law_1.01,0.2996710395812988
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,16,power_law_1.01,0.2905792045593262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,16,power_law_1.01,0.28719104766845704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,16,power_law_1.01,0.32006015777587893
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,16,power_law_1.01,0.37836929321289064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,16,power_law_1.01,0.4312128067016602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,16,power_law_1.01,0.5371750259399415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,16,power_law_1.01,0.6752358245849609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,16,power_law_1.01,0.9450624084472656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,16,power_law_1.01,1.2066738891601563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,16,power_law_1.01,1.6435098266601564
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,1,16,power_law_1.01,0.09332736015319824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,16,power_law_1.01,2.0957452392578126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,16,power_law_1.01,8.038466796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,1,16,power_law_1.01,0.1239692783355713
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,1,16,power_law_1.01,0.12436736106872559
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,1,16,power_law_1.01,0.12191743850708008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,16,power_law_1.01,3.287562255859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,1,16,power_law_1.01,0.12330240249633788
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,1,16,power_law_1.01,0.1222771167755127
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,1,16,power_law_1.01,0.12440959930419922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,16,power_law_1.01,9.796583862304688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,1,16,power_law_1.01,0.11684224128723145
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,1,16,power_law_1.01,0.12142208099365234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,1,16,power_law_1.01,0.11739520072937011
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,1,16,power_law_1.01,0.12328576087951659
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,1,16,power_law_1.01,0.12425855636596679
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,1,16,power_law_1.01,0.11878911972045898
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,1,16,power_law_1.01,0.1263270378112793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,1,16,power_law_1.01,0.1294591999053955
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,1,16,power_law_1.01,0.14054271697998047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,16,power_law_1.01,4.346859436035156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,1,16,power_law_1.01,0.1615782356262207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,16,power_law_1.01,0.2415065574645996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,16,power_law_1.01,0.2255833625793457
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,1,16,power_law_1.01,0.12411775588989257
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,1,16,power_law_1.01,0.1319040012359619
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,1,16,power_law_1.01,0.19640064239501953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,1,16,power_law_1.01,0.23517311096191404
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,1,16,power_law_1.01,0.3599795150756836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,1,16,power_law_1.01,0.3091481590270996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,16,power_law_1.2,0.35345409393310545
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,1,16,power_law_1.01,0.5171456146240234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,16,power_law_1.2,0.3566809463500976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,16,power_law_1.2,0.3887116622924805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,1,16,power_law_1.01,0.654882583618164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,16,power_law_1.2,0.4133004760742187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,16,power_law_1.2,0.4076134490966797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,16,power_law_1.2,0.4409881591796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,16,power_law_1.2,0.45674369812011717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,16,power_law_1.2,0.3680089569091797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,16,power_law_1.2,0.4854361724853516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,16,power_law_1.2,0.5062412643432618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,16,power_law_1.2,0.5511603164672851
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,16,power_law_1.2,0.2865433692932129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,16,power_law_1.2,0.5792665481567383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,16,power_law_1.2,0.6121075057983398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,16,power_law_1.2,0.6538893127441406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,16,power_law_1.2,0.9333964538574218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,16,power_law_1.2,1.2175424194335938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,16,power_law_1.2,0.3629555130004883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,16,power_law_1.2,1.6591384887695313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,16,power_law_1.2,0.43779968261718755
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,16,power_law_1.2,2.113392639160156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,16,power_law_1.2,3.0558590698242187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,16,power_law_1.2,0.1581657600402832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,16,power_law_1.2,0.8213504028320312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,16,power_law_1.2,0.19072256088256836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,16,power_law_1.2,0.2083123207092285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,16,power_law_1.2,0.21351680755615235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,16,power_law_1.2,0.2206643295288086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,16,power_law_1.2,0.24047359466552734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,16,power_law_1.2,0.22401407241821286
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,16,power_law_1.2,6.07944091796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,16,power_law_1.2,0.23112192153930664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,16,power_law_1.2,0.22597120285034178
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,16,power_law_1.2,0.24425727844238282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,16,power_law_1.2,0.2453670310974121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,16,power_law_1.2,0.27191040039062503
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,16,power_law_1.2,0.30950656890869144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,16,power_law_1.2,0.2861964797973633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,16,power_law_1.2,0.29500288009643555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,16,power_law_1.2,0.3346316909790039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,16,power_law_1.2,4.641004943847657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,16,power_law_1.2,0.3858752059936523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,16,power_law_1.2,0.4545177459716797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,16,power_law_1.2,0.5883327865600586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,16,power_law_1.2,0.7303040313720703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,16,power_law_1.2,1.0259404754638672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,16,power_law_1.2,0.20498943328857422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,16,power_law_1.2,1.264691162109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,16,power_law_1.2,1.8068121337890624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,1,16,power_law_1.2,0.09356032371520996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,1,16,power_law_1.2,0.11939840316772461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,16,power_law_1.2,2.370789184570312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,1,16,power_law_1.2,0.12206080436706543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,1,16,power_law_1.2,0.12462080001831055
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,1,16,power_law_1.2,0.1253388786315918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,1,16,power_law_1.2,0.121146879196167
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,1,16,power_law_1.2,0.1190015983581543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,16,power_law_1.2,12.62095458984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,1,16,power_law_1.2,0.11313023567199706
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,1,16,power_law_1.2,0.12256256103515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,1,16,power_law_1.2,0.12092800140380859
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,1,16,power_law_1.2,0.1192742443084717
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,1,16,power_law_1.2,0.12236031532287597
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,1,16,power_law_1.2,0.12220671653747558
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,1,16,power_law_1.2,0.12401920318603517
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,1,16,power_law_1.2,0.12892800331115722
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,1,16,power_law_1.2,0.1460428810119629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,16,power_law_1.2,3.779971923828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,1,16,power_law_1.01,1.0905843353271485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,16,power_law_1.2,9.520361938476563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,16,power_law_1.2,4.781276245117188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,1,16,power_law_1.01,1.1855628967285157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,1,16,power_law_1.2,0.1369484806060791
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,1,16,power_law_1.2,0.18485120773315428
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,1,16,power_law_1.2,0.1614784049987793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,1,16,power_law_1.2,0.2351910400390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,8,balanced,0.05076608180999755
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,8,balanced,0.06655231952667237
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,8,balanced,0.09345536231994629
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,1,16,power_law_1.2,0.28287744522094727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,8,balanced,0.1945484733581543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,8,balanced,0.3758348846435547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,8,balanced,0.3902361679077148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,8,balanced,0.39302398681640627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,8,balanced,0.4510617446899414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,1,16,power_law_1.2,0.3579955291748047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,8,balanced,0.4536832046508789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,1,16,power_law_1.2,0.5403776168823242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,8,balanced,0.4611174392700195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,8,balanced,0.48226303100585943
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,8,balanced,0.4933452987670899
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,8,balanced,0.5187545776367187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,8,balanced,0.5087193679809571
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,8,balanced,0.5277580642700196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,8,balanced,0.561190414428711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,8,balanced,0.6360268783569336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,8,balanced,0.7019084930419922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,8,balanced,0.9860928344726563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,8,balanced,1.1151283264160157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,1,16,power_law_1.2,0.7100889587402344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,8,balanced,1.6588633728027342
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,8,balanced,1.8656665039062499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,8,balanced,0.3115660858154297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,8,balanced,2.9700262451171873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,8,balanced,0.049039359092712405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,8,balanced,3.914501037597656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,8,balanced,0.10935551643371581
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,8,balanced,0.17355392456054688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,8,balanced,0.21993600845336916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,8,balanced,0.2296691131591797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,8,balanced,0.23367168426513674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,8,balanced,0.25490816116333004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,8,balanced,0.03218303918838501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,8,balanced,0.25650943756103517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,8,balanced,5.99889404296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,8,balanced,0.2592076873779297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,8,balanced,0.27683200836181643
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,8,balanced,0.28920831680297854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,8,balanced,0.07087232112884521
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,8,balanced,0.28179967880249024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,8,balanced,0.2979020881652832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,8,balanced,0.3223654556274414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,8,balanced,0.36362625122070313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,8,balanced,0.40270336151123043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,8,balanced,0.4991385650634766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,8,balanced,0.30906751632690427
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,8,balanced,0.5675289535522461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,8,balanced,0.848678436279297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,8,balanced,7.942909545898438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,8,balanced,1.0141401672363282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,8,balanced,1.4636927795410155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,8,balanced,1.7829452514648438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,8,balanced,2.6928280639648436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,8,balanced,3.4745050048828126
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,1,16,power_law_1.2,0.8870886230468751
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,1,16,power_law_1.2,1.6178521728515627
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,1,16,power_law_1.2,1.9570022583007813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,2,8,balanced,0.029308159351348877
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,2,8,balanced,0.1346623992919922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,2,8,balanced,0.033729279041290285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,2,8,balanced,0.13426176071166993
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,2,8,balanced,0.13542400360107423
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,2,8,balanced,0.1351731204986572
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,2,8,balanced,0.1353548812866211
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,2,8,balanced,0.046915841102600095
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,2,8,balanced,0.1369638442993164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,2,8,balanced,0.13778176307678222
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,2,8,balanced,0.13798656463623046
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,2,8,balanced,0.14008319854736329
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,2,8,balanced,0.14392704010009766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,2,8,balanced,0.14836735725402833
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,2,8,balanced,0.15546624183654784
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,2,8,balanced,0.16167423248291016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,2,8,balanced,0.07100543975830079
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,2,8,balanced,0.10575103759765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,2,8,balanced,0.13274623870849608
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,2,8,balanced,0.13387776374816895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,8,power_law_1.01,0.1631999969482422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,8,power_law_1.01,0.2682496070861816
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,2,8,balanced,0.1348031997680664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,8,power_law_1.01,0.2957632064819336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,8,power_law_1.01,0.3217011260986328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,8,power_law_1.01,0.3541670227050781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,8,power_law_1.01,0.3732620620727539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,8,power_law_1.01,0.23104639053344728
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,8,power_law_1.01,0.3567283248901367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,8,power_law_1.01,0.42451713562011717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,8,power_law_1.01,0.41716350555419923
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,8,power_law_1.01,0.4328768157958985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,8,power_law_1.01,0.4640192031860352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,8,power_law_1.01,0.46596607208251956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,8,power_law_1.01,0.5050867080688477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,8,power_law_1.01,0.48989952087402344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,8,power_law_1.01,0.517262077331543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,8,power_law_1.01,0.5584320068359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,2,8,balanced,0.2296998405456543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,2,8,balanced,0.3456921768188477
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,2,8,balanced,0.24639488220214845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,8,power_law_1.01,0.6743846130371094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,8,power_law_1.01,0.7845452880859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,2,8,balanced,0.3774118423461914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,8,power_law_1.01,1.2240332794189452
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,8,power_law_1.01,1.698796844482422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,8,power_law_1.01,2.123962860107422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,2,8,balanced,0.5726924896240234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,8,power_law_1.01,0.15796223640441895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,8,power_law_1.01,3.0701849365234373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,8,power_law_1.01,0.09496319770812989
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,8,power_law_1.01,0.17381120681762696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,8,power_law_1.01,0.18913791656494142
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,8,power_law_1.01,3.992262268066406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,8,power_law_1.01,0.2268070411682129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,8,power_law_1.01,0.9807103729248047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,8,power_law_1.01,5.798533325195312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,8,power_law_1.01,0.24061695098876953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,8,power_law_1.01,0.145350399017334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,8,power_law_1.01,0.2574143981933594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,8,power_law_1.01,0.2677260780334473
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,8,power_law_1.01,0.2985561561584473
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,8,power_law_1.01,0.216760311126709
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,8,power_law_1.01,0.27060096740722656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,8,power_law_1.01,0.2822604751586914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,8,power_law_1.01,0.3136767959594727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,8,power_law_1.01,0.3690099334716797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,2,8,balanced,0.71831298828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,8,power_law_1.01,0.41904510498046876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,8,power_law_1.01,8.125595092773438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,8,power_law_1.01,0.23556352615356446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,8,power_law_1.01,0.6368755340576172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,8,power_law_1.01,0.23533056259155272
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,8,power_law_1.01,0.8642086029052735
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,2,8,power_law_1.01,0.06652416229248047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,8,power_law_1.01,1.052327651977539
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,2,8,power_law_1.01,0.09232128143310547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,2,8,power_law_1.01,0.10164735794067384
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,8,power_law_1.01,1.5029286193847657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,2,8,power_law_1.01,0.10394880294799805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,2,8,power_law_1.01,0.12148608207702635
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,2,8,power_law_1.01,0.13407999992370606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,8,power_law_1.01,1.9374079895019531
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,2,8,power_law_1.01,0.11421695709228516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,2,8,power_law_1.01,0.11641983985900879
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,2,8,power_law_1.01,0.11870847702026369
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,2,8,power_law_1.01,0.12016127586364747
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,2,8,power_law_1.01,0.12559103965759277
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,8,power_law_1.01,0.209617919921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,2,8,power_law_1.01,0.12779904365539552
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,2,8,power_law_1.01,0.1301593589782715
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,2,8,power_law_1.01,0.13346176147460936
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,2,8,power_law_1.01,0.13622143745422363
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,2,8,power_law_1.01,0.13686400413513183
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,8,power_law_1.01,2.9055859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,2,8,power_law_1.01,0.1657036781311035
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,2,8,power_law_1.01,0.18543487548828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,8,power_law_1.01,3.7606118774414066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,8,power_law_1.01,0.5164902496337891
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,2,8,power_law_1.01,0.14787967681884767
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,2,8,power_law_1.01,0.2703462409973144
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,2,8,power_law_1.01,0.2230963134765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,2,8,power_law_1.01,0.6313062286376953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,2,8,power_law_1.01,0.4206707382202149
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,8,power_law_1.2,0.2050777626037598
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,2,8,power_law_1.01,0.34802688598632814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,8,power_law_1.2,0.2959526443481445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,8,power_law_1.2,0.34397953033447265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,8,power_law_1.2,0.3716172790527344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,2,8,power_law_1.01,0.7770713806152344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,8,power_law_1.2,0.3540300750732422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,8,power_law_1.2,0.4047436904907227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,8,power_law_1.2,0.4365427017211914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,8,power_law_1.2,0.4257984161376953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,2,8,power_law_1.01,1.3182170104980468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,8,power_law_1.2,0.45615615844726565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,8,power_law_1.2,0.4762905502319336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,8,power_law_1.2,0.508031997680664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,8,power_law_1.2,0.5418048095703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,8,power_law_1.2,0.161079044342041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,8,power_law_1.2,0.5083993530273437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,8,power_law_1.2,0.5815910339355469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,8,power_law_1.2,0.7026675415039063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,8,power_law_1.2,0.8147763061523439
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,8,power_law_1.2,1.0193791961669922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,8,power_law_1.2,1.261990432739258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,8,power_law_1.2,1.7167948913574218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,8,power_law_1.2,2.2853543090820314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,8,power_law_1.2,0.09451135635375976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,8,power_law_1.2,0.3595302581787109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,8,power_law_1.2,0.11906559944152832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,8,power_law_1.2,3.28021240234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,8,power_law_1.2,0.14101375579833983
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,8,power_law_1.2,0.17465343475341796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,8,power_law_1.2,0.20304384231567382
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,8,power_law_1.2,0.2169843292236328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,8,power_law_1.2,0.20959871292114257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,8,power_law_1.2,0.2464614486694336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,8,power_law_1.2,4.342657165527344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,8,power_law_1.2,0.21632768630981447
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,8,power_law_1.2,0.24850816726684571
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,8,power_law_1.2,0.25566591262817384
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,8,power_law_1.2,0.23792896270751954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,8,power_law_1.2,0.27476863861083983
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,8,power_law_1.2,0.2770470428466797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,8,power_law_1.2,0.30241024017333984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,8,power_law_1.2,0.292544002532959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,8,power_law_1.2,0.316549129486084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,8,power_law_1.2,0.3605376052856445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,8,power_law_1.2,0.42297470092773437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,8,power_law_1.2,0.5399705505371093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,8,power_law_1.2,0.6500723266601562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,8,power_law_1.2,6.299175415039063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,8,power_law_1.2,0.880337905883789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,8,power_law_1.2,1.0922675323486328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,2,8,power_law_1.2,0.06716032028198242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,8,power_law_1.2,1.5408230590820313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,2,8,power_law_1.2,0.0797056007385254
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,2,8,power_law_1.2,0.09298687934875488
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,2,8,power_law_1.2,0.10411904335021974
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,8,power_law_1.2,2.010007019042969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,8,power_law_1.2,0.23749759674072263
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,2,8,power_law_1.2,0.12352383613586426
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,2,8,power_law_1.2,0.13153535842895508
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,2,8,power_law_1.2,0.12576767921447754
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,2,8,power_law_1.2,0.12088064193725585
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,2,8,power_law_1.2,0.12027647972106933
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,2,8,power_law_1.2,0.12155136108398437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,2,8,power_law_1.2,0.1272268772125244
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,2,8,power_law_1.2,0.1287718391418457
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,2,8,power_law_1.2,0.1309286403656006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,8,power_law_1.2,2.9319000244140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,2,8,power_law_1.2,0.1400870418548584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,2,8,power_law_1.2,0.14560000419616698
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,8,power_law_1.2,8.620610351562501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,8,power_law_1.2,3.9801831054687495
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,2,8,power_law_1.01,1.5708454895019532
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,2,8,power_law_1.2,0.16229375839233398
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,2,8,power_law_1.2,0.15267583847045899
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,2,8,power_law_1.2,0.16979583740234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,2,8,power_law_1.2,0.20829824447631834
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,2,8,power_law_1.2,0.2636313629150391
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,2,8,power_law_1.2,0.322371826171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,4,4,balanced,0.06487296104431153
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,4,4,balanced,0.09198975563049316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,4,4,balanced,0.04870016098022461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,4,4,balanced,0.17209983825683595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,4,4,balanced,0.286242561340332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,2,8,power_law_1.2,0.44149887084960937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,4,4,balanced,0.3944217681884766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,4,4,balanced,0.39745025634765624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,4,4,balanced,0.4096640014648438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,4,4,balanced,0.3804595184326172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,4,4,balanced,0.4139507293701172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,4,4,balanced,0.4203379058837891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,4,4,balanced,0.463183364868164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,4,4,balanced,0.4695616149902344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,2,8,power_law_1.2,0.5450675201416015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,4,4,balanced,0.496869125366211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,4,4,balanced,0.4784435272216797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,4,4,balanced,0.4924838256835938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,4,4,balanced,0.5247321701049804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,4,4,balanced,0.5897753524780274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,4,4,balanced,0.6509286499023437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,4,4,balanced,0.9111833953857422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,4,4,balanced,1.021338882446289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,4,4,balanced,1.4410086059570313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,4,4,balanced,1.6871423339843752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,4,4,balanced,2.517295379638672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,4,4,balanced,0.03260799884796143
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,4,4,balanced,0.04589312076568604
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,4,4,balanced,0.07358592033386231
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,4,4,balanced,3.193045654296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,4,4,balanced,0.11725055694580078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,4,4,balanced,0.18436735153198242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,4,4,balanced,0.23627904891967771
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,4,4,balanced,0.24849792480468752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,4,4,balanced,0.25271039962768554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,4,4,balanced,0.26232831954956054
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,4,4,balanced,0.2654745674133301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,4,4,balanced,0.2715328025817871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,4,4,balanced,4.791522521972657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,4,4,balanced,0.31689216613769533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,4,4,balanced,0.3223513412475586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,4,4,balanced,0.35180416107177737
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,4,4,balanced,0.32285438537597655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,4,4,balanced,0.32790782928466794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,4,4,balanced,0.3644262313842773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,4,4,balanced,0.3999219131469727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,4,4,balanced,0.4317068862915039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,4,4,balanced,6.3527899169921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,4,4,balanced,0.5000102233886718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,4,4,balanced,0.5716812896728516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,4,4,balanced,0.852089614868164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,4,4,balanced,0.9848716735839844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,4,4,balanced,1.4258111572265624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,4,4,balanced,1.7170034790039062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,4,4,balanced,2.5723559570312498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,4,4,balanced,3.295024719238281
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,2,8,power_law_1.2,0.8535155487060546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,2,8,power_law_1.2,1.143915557861328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,2,8,power_law_1.2,2.6070721435546877
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,2,8,power_law_1.2,2.032316131591797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,4,4,balanced,0.041103358268737795
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,4,4,balanced,0.15738368034362793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,4,4,balanced,0.15794816017150878
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,4,4,balanced,0.15882240295410158
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,4,4,balanced,0.15935744285583495
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,4,4,balanced,0.16291711807250978
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,4,4,balanced,0.04411520004272461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,4,4,balanced,0.16262144088745117
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,4,4,balanced,0.16371456146240235
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,4,4,balanced,0.16669952392578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,4,4,balanced,0.05731584072113037
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,4,4,balanced,0.16780160903930663
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,4,4,balanced,0.17484159469604493
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,4,4,balanced,0.17977216720581055
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,4,4,balanced,0.1933670425415039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,4,4,balanced,0.2027993583679199
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,4,4,balanced,0.07941376209259034
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,4,4,balanced,0.12053631782531737
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,4,4,balanced,0.15675647735595705
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,4,4,balanced,0.1574143981933594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,4,4,balanced,0.15784064292907715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,4,4,power_law_1.01,0.0852454376220703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,4,4,power_law_1.01,0.14749055862426758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,4,4,power_law_1.01,0.15804032325744627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,4,4,power_law_1.01,0.2229158401489258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,4,4,power_law_1.01,0.23680255889892576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,4,4,power_law_1.01,0.3420620727539062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,4,4,power_law_1.01,0.32580478668212887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,4,4,power_law_1.01,0.3511423873901367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,4,4,power_law_1.01,0.3666700744628906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,4,4,power_law_1.01,0.3672768020629883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,4,4,power_law_1.01,0.398606071472168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,4,4,power_law_1.01,0.4419123077392578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,4,4,power_law_1.01,0.4457356643676758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,4,4,power_law_1.01,0.4779276657104492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,4,4,power_law_1.01,0.46168960571289064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,4,4,power_law_1.01,0.4827008056640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,4,4,power_law_1.01,0.5269836807250977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,4,4,power_law_1.01,0.6204915237426758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,4,4,power_law_1.01,0.710882568359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,4,4,balanced,0.31198591232299805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,4,4,balanced,0.484005126953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,4,4,balanced,0.3400332641601563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,4,4,power_law_1.01,0.8910540771484374
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,4,4,balanced,0.5362982559204101
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,4,4,power_law_1.01,1.0805734252929686
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,4,4,balanced,0.9072345733642578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,4,4,power_law_1.01,1.465233917236328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,4,4,power_law_1.01,0.06453887939453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,4,4,power_law_1.01,0.0993830394744873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,4,4,power_law_1.01,1.8345318603515626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,4,4,power_law_1.01,0.10759552001953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,4,4,balanced,1.0564313507080079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,4,4,power_law_1.01,0.1459712028503418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,4,4,power_law_1.01,0.1590988826751709
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,4,4,power_law_1.01,0.23216512680053708
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,4,4,power_law_1.01,2.6197479248046873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,4,4,power_law_1.01,0.23915775299072264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,4,4,power_law_1.01,0.23655296325683595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,4,4,power_law_1.01,0.24451583862304688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,4,4,power_law_1.01,0.25182336807250977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,4,4,power_law_1.01,0.29354751586914063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,4,4,power_law_1.01,0.29419008255004886
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,4,4,power_law_1.01,0.3229350280761719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,4,4,power_law_1.01,0.29529087066650395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,4,4,power_law_1.01,3.439833679199219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,4,4,power_law_1.01,0.21841407775878907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,4,4,power_law_1.01,0.31316991806030275
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,4,4,power_law_1.01,0.34248703002929687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,4,4,power_law_1.01,0.39779327392578123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,4,4,power_law_1.01,0.4348608016967773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,4,4,power_law_1.01,0.5347366333007812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,4,4,power_law_1.01,0.6249113464355469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,4,4,power_law_1.01,0.8310015869140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,4,4,power_law_1.01,1.0242713928222655
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,4,4,power_law_1.01,0.06087552070617676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,4,4,power_law_1.01,4.970718688964844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,4,4,power_law_1.01,1.4160205078124999
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,4,4,power_law_1.01,0.07572735786437988
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,4,4,power_law_1.01,0.09695743560791016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,4,4,power_law_1.01,0.10369791984558105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,4,4,power_law_1.01,1.8167616271972655
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,4,4,power_law_1.01,0.13224063873291017
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,4,4,power_law_1.01,0.14317184448242187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,4,4,power_law_1.01,0.15017215728759764
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,4,4,power_law_1.01,0.08307711601257324
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,4,4,power_law_1.01,0.14991488456726074
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,4,4,power_law_1.01,0.1485977554321289
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,4,4,power_law_1.01,0.15221759796142578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,4,4,power_law_1.01,2.62616455078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,4,4,power_law_1.01,6.63966064453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,4,4,power_law_1.01,0.1565721607208252
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,4,4,power_law_1.01,0.16162431716918946
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,4,4,power_law_1.01,0.1673561668395996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,4,4,power_law_1.01,0.17330944061279296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,4,4,power_law_1.01,3.5011480712890624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,4,4,power_law_1.01,0.15826687812805176
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,4,4,power_law_1.01,0.20825983047485352
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,4,4,power_law_1.01,0.18750335693359374
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,4,4,power_law_1.01,0.24204032897949218
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,4,4,power_law_1.01,0.28721792221069337
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,4,4,power_law_1.01,0.3627750396728516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,4,4,power_law_1.2,0.08700032234191894
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,4,4,power_law_1.01,0.48242046356201174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,4,4,power_law_1.2,0.10865792274475097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,4,4,power_law_1.2,0.1530892753601074
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,4,4,power_law_1.01,0.6282598495483398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,4,4,power_law_1.2,0.24074880599975587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,4,4,power_law_1.2,0.331776008605957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,4,4,power_law_1.2,0.35123840332031253
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,4,4,power_law_1.2,0.3504000091552734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,4,4,power_law_1.2,0.36811008453369143
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,4,4,power_law_1.2,0.4048307037353515
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,4,4,power_law_1.01,0.8888166046142578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,4,4,power_law_1.2,0.44255233764648433
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,4,4,power_law_1.2,0.4568153762817383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,4,4,power_law_1.2,0.47812862396240235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,4,4,power_law_1.2,0.47123840332031247
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,4,4,power_law_1.2,0.4903705596923828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,4,4,power_law_1.2,0.3583014297485351
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,4,4,power_law_1.2,0.5479270553588866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,4,4,power_law_1.2,0.6302105712890624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,4,4,power_law_1.2,0.21980031967163086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,4,4,power_law_1.2,0.7289766693115235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,4,4,power_law_1.2,0.9146227264404297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,4,4,power_law_1.2,1.1211775970458984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,4,4,power_law_1.2,1.5158618164062498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,4,4,power_law_1.2,1.8953485107421877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,4,4,power_law_1.2,0.06273151874542236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,4,4,power_law_1.2,0.08696703910827637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,4,4,power_law_1.2,2.773104553222656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,4,4,power_law_1.2,0.1076416015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,4,4,power_law_1.2,0.13845888137817383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,4,4,power_law_1.2,0.16640768051147461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,4,4,power_law_1.2,3.565148010253906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,4,4,power_law_1.2,0.21487232208251955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,4,4,power_law_1.2,0.2260095977783203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,4,4,power_law_1.2,0.21747583389282227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,4,4,power_law_1.2,0.24033792495727538
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,4,4,power_law_1.2,0.24754432678222654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,4,4,power_law_1.2,0.2586227226257324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,4,4,power_law_1.2,0.29649791717529295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,4,4,power_law_1.2,0.30069248199462895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,4,4,power_law_1.2,0.3328550338745117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,4,4,power_law_1.2,0.3068953514099121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,4,4,power_law_1.2,0.3199398422241211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,4,4,power_law_1.2,0.36019073486328124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,4,4,power_law_1.2,5.2337011718749995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,4,4,power_law_1.2,0.4009523010253906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,4,4,power_law_1.2,0.44871551513671876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,4,4,power_law_1.2,0.5436774444580078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,4,4,power_law_1.2,0.6542182159423828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,4,4,power_law_1.2,0.8481932830810546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,4,4,power_law_1.2,1.0541324615478516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,4,4,power_law_1.2,1.4442098999023438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,4,4,power_law_1.2,0.060165119171142575
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,4,4,power_law_1.2,1.8698892211914064
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,4,4,power_law_1.2,0.07032959938049317
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,4,4,power_law_1.2,0.08007807731628418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,4,4,power_law_1.2,6.766366577148437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,4,4,power_law_1.2,0.09254400253295898
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,4,4,power_law_1.2,0.09541888236999511
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,4,4,power_law_1.2,0.13960320472717286
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,4,4,power_law_1.2,0.14213503837585448
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,4,4,power_law_1.2,0.14474495887756347
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,4,4,power_law_1.2,0.14628992080688477
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,4,4,power_law_1.2,0.1545907211303711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,4,4,power_law_1.2,0.1548518371582031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,4,4,power_law_1.2,0.15563136100769043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,4,4,power_law_1.2,0.16079999923706054
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,4,4,power_law_1.2,0.16850175857543945
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,4,4,power_law_1.2,0.17796607971191408
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,4,4,power_law_1.2,0.18625663757324218
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,4,4,power_law_1.2,0.23212799072265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,4,4,power_law_1.2,2.6977856445312502
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,4,4,power_law_1.2,3.6093518066406247
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,4,4,power_law_1.01,1.143837432861328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,4,4,power_law_1.01,1.7946290588378908
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,4,4,power_law_1.01,2.291875915527344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,4,4,power_law_1.2,0.19970304489135743
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,4,4,power_law_1.2,0.27633535385131835
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,4,4,power_law_1.2,0.35208831787109374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,8,2,balanced,0.04603775978088379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,8,2,balanced,0.06465663909912109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,8,2,balanced,0.09190912246704101
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,4,4,power_law_1.2,0.4586188888549804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,8,2,balanced,0.29077375411987305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,8,2,balanced,0.40375553131103514
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,8,2,balanced,0.41211391448974605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,8,2,balanced,0.4140044784545898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,8,2,balanced,0.42827392578124995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,8,2,balanced,0.17233152389526368
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,4,4,power_law_1.2,0.6508902740478516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,8,2,balanced,0.43590145111083983
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,8,2,balanced,0.4492774581909179
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,8,2,balanced,0.4581248092651367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,8,2,balanced,0.4303488159179688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,8,2,balanced,0.47760257720947263
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,4,4,power_law_1.2,0.8248614501953124
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,4,4,power_law_1.2,1.1591193389892578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,8,2,balanced,0.48241409301757815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,8,2,balanced,0.49258369445800787
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,8,2,balanced,0.6415206146240234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,8,2,balanced,0.5225126266479492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,8,2,balanced,0.9313676452636719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,8,2,balanced,1.031088638305664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,8,2,balanced,0.5855680084228515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,8,2,balanced,1.4366566467285158
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,8,2,balanced,1.6886361694335938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,8,2,balanced,0.035450880527496335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,8,2,balanced,0.04884352207183838
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,8,2,balanced,0.07666816234588622
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,8,2,balanced,0.13269503593444826
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,8,2,balanced,2.5007603454589846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,8,2,balanced,0.22215295791625977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,8,2,balanced,0.29813888549804685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,8,2,balanced,0.3068070411682129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,8,2,balanced,0.3096473693847656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,8,2,balanced,3.137018737792969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,8,2,balanced,0.32057857513427734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,8,2,balanced,0.32326656341552734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,8,2,balanced,0.32827903747558596
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,8,2,balanced,0.34065536499023436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,8,2,balanced,0.34562049865722655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,8,2,balanced,0.3633113479614258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,8,2,balanced,0.3765081787109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,8,2,balanced,0.3917452621459961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,8,2,balanced,0.39433471679687504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,8,2,balanced,0.4401203155517578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,8,2,balanced,0.4647987365722656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,8,2,balanced,0.5320320129394531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,8,2,balanced,0.5936934280395507
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,8,2,balanced,4.630999145507813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,8,2,balanced,5.925562744140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,8,2,balanced,1.048947219848633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,8,2,balanced,1.51112060546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,8,2,balanced,0.9230783843994141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,8,2,power_law_1.01,0.09003520011901855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,8,2,balanced,1.7659954833984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,8,2,power_law_1.01,0.11120384216308594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,8,2,power_law_1.01,0.2340800094604492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,8,2,balanced,2.710069885253906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,8,2,power_law_1.01,0.06160384178161621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,8,2,power_law_1.01,0.3313740921020508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,8,2,power_law_1.01,0.3622208023071289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,8,2,power_law_1.01,0.37100414276123045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,8,2,power_law_1.01,0.38516353607177733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,8,2,power_law_1.01,0.38663169860839847
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,8,2,balanced,3.405177612304688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,8,2,power_law_1.01,0.4216524887084961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,8,2,power_law_1.01,0.45139198303222655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,8,2,power_law_1.01,0.47686912536621096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,8,2,power_law_1.01,0.5344243240356445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,8,2,power_law_1.01,0.47281536102294924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,8,2,power_law_1.01,0.4897907257080078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,8,2,power_law_1.01,0.5349913787841797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,8,2,power_law_1.01,0.6238451385498047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,8,2,power_law_1.01,0.18210304260253907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,8,2,power_law_1.01,0.706456298828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,8,2,power_law_1.01,0.881553955078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,8,2,power_law_1.01,1.0845721435546873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,8,2,power_law_1.01,1.4150527954101562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,4,4,power_law_1.2,2.694609985351563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,8,2,power_law_1.01,1.7970726013183593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,8,2,power_law_1.01,0.04591487884521485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,8,2,power_law_1.01,0.06858367919921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,8,2,power_law_1.01,0.09594752311706542
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,8,2,power_law_1.01,2.536678466796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,8,2,power_law_1.01,0.1848691177368164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,8,2,power_law_1.01,0.13738880157470704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,8,2,power_law_1.01,0.24422143936157226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,8,2,power_law_1.01,0.2654374313354492
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,4,4,power_law_1.2,3.4828057861328126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,8,2,power_law_1.01,0.2761715126037598
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,8,2,power_law_1.01,3.2649227905273435
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,8,2,power_law_1.01,0.290830078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,8,2,power_law_1.01,0.3139737510681152
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,8,2,power_law_1.01,0.2972991943359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,8,2,power_law_1.01,0.32000896453857425
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,8,2,power_law_1.01,0.3389478302001953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,8,2,power_law_1.01,0.34889343261718747
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,8,2,power_law_1.01,0.36361087799072267
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,8,2,power_law_1.01,0.28593408584594726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,8,2,power_law_1.01,0.3801881790161133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,8,2,power_law_1.01,0.4390220642089844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,8,2,power_law_1.01,4.750174865722657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,8,2,power_law_1.01,0.47612159729003906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,8,2,power_law_1.01,0.574576644897461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,8,2,power_law_1.01,0.6693504333496094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,8,2,power_law_1.01,0.8694771575927736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,8,2,power_law_1.01,1.0903449249267578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,8,2,power_law_1.2,0.0640012788772583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,8,2,power_law_1.01,1.4718988037109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,8,2,power_law_1.2,0.07820799827575683
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,8,2,power_law_1.01,6.2721228027343745
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,8,2,power_law_1.01,1.8698074340820312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,8,2,power_law_1.2,0.2224742317199707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,8,2,power_law_1.2,0.16989311218261718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,8,2,power_law_1.2,0.3261439895629883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,8,2,power_law_1.2,0.36010494232177737
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,8,2,power_law_1.2,0.11213312149047852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,8,2,power_law_1.2,0.3843353652954101
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,8,2,power_law_1.2,0.3700787353515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,8,2,power_law_1.2,0.3908710479736328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,8,2,power_law_1.01,2.694915771484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,8,2,power_law_1.2,0.4252799987792969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,8,2,power_law_1.2,0.4599347305297852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,8,2,power_law_1.2,0.49521793365478517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,8,2,power_law_1.2,0.47710208892822265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,8,2,power_law_1.2,0.5774335861206055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,8,2,power_law_1.2,0.49218048095703126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,8,2,power_law_1.01,3.5008895874023436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,8,2,power_law_1.2,0.548042221069336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,8,2,power_law_1.2,0.7255347442626954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,8,2,power_law_1.2,1.1024447631835936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,8,2,power_law_1.2,0.8958719635009766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,8,2,power_law_1.2,0.6246694564819336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,8,2,power_law_1.2,0.04608895778656006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,8,2,power_law_1.2,1.441739501953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,8,2,power_law_1.2,0.06292863845825195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,8,2,power_law_1.2,0.09543807983398438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,8,2,power_law_1.2,0.13134079933166504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,8,2,power_law_1.2,0.17655935287475585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,8,2,power_law_1.2,2.60305419921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,8,2,power_law_1.2,0.24827903747558594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,8,2,power_law_1.2,0.2629875183105469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,8,2,power_law_1.2,0.274202880859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,8,2,power_law_1.2,0.2875379180908203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,8,2,power_law_1.2,0.2939967918395996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,8,2,power_law_1.2,3.3203570556640627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,8,2,power_law_1.2,0.3020505523681641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,8,2,power_law_1.2,0.3168115234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,8,2,power_law_1.2,0.32425216674804686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,8,2,power_law_1.2,0.35838207244873044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,8,2,power_law_1.2,1.8140325927734373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,8,2,power_law_1.2,0.36216705322265624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,8,2,power_law_1.2,0.37225086212158204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,8,2,power_law_1.2,0.3839052963256836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,8,2,power_law_1.2,0.44073345184326174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,8,2,power_law_1.2,0.49846782684326174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,8,2,power_law_1.2,0.5851289749145507
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,8,2,power_law_1.2,0.6862924957275391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,8,2,power_law_1.2,0.8875276947021484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,8,2,power_law_1.2,1.1089997100830078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,8,2,power_law_1.2,4.813392639160156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,8,2,power_law_1.2,1.4875430297851562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,8,2,power_law_1.2,1.9108799743652345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,8,2,power_law_1.2,6.24956298828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,16,1,balanced,0.08110400199890136
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,16,1,balanced,0.1206022357940674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,16,1,balanced,0.1869503974914551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,8,2,power_law_1.2,2.7603570556640626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,16,1,balanced,0.3452921676635742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,16,1,balanced,0.8924838256835936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,16,1,balanced,0.9025491333007812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,16,1,balanced,0.6217203140258789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,16,1,balanced,0.9296781158447265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,16,1,balanced,0.9448544311523438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,16,1,balanced,0.9088716888427735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,16,1,balanced,0.919767074584961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,8,2,power_law_1.2,3.5634561157226563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,16,1,balanced,0.9743270111083984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,16,1,balanced,0.9608927917480468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,16,1,balanced,0.9225215911865234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,16,1,balanced,1.2963372802734374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,16,1,balanced,0.9541017913818359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,16,1,balanced,1.677821502685547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,16,1,balanced,1.179509735107422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,16,1,balanced,1.2298400115966797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,16,1,balanced,2.0141017150878904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,16,1,balanced,2.127838134765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,16,1,balanced,3.0109841918945315
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,16,1,balanced,3.247054748535156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,16,1,balanced,4.987130126953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,16,1,balanced,6.110631713867187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,16,1,balanced,8.890189208984376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,16,1,balanced,11.04826904296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,16,1,balanced,0.20657215118408204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,16,1,balanced,0.35135040283203123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,16,1,balanced,0.6297907257080079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,16,1,balanced,1.558871612548828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,16,1,balanced,3.1500628662109373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,16,1,balanced,3.2361395263671873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,16,1,balanced,3.2367193603515623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,16,1,balanced,3.7967514038085937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,16,1,balanced,3.253623657226562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,16,1,balanced,3.943489379882813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,16,1,balanced,3.246980590820313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,16,1,balanced,3.216120300292969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,16,1,balanced,3.156816711425781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,16,1,balanced,3.992745666503906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,16,1,balanced,4.110691223144531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,16,1,balanced,3.786909484863281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,16,1,balanced,4.3129574584960935
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,16,1,balanced,5.228798217773438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,16,1,balanced,5.595457153320313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,16,1,balanced,5.650936279296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,16,1,balanced,4.301923217773437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,16,1,power_law_1.01,0.0787398386001587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,16,1,power_law_1.01,0.18615936279296874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,16,1,power_law_1.01,0.3017990493774414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,16,1,power_law_1.01,0.4410118484497071
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,16,1,balanced,10.784599609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,16,1,power_law_1.01,0.6904844665527344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,16,1,balanced,11.219599609374999
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,16,1,power_law_1.01,0.7383334350585937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,16,1,power_law_1.01,0.7636402893066406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,16,1,power_law_1.01,0.7907673645019531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,16,1,power_law_1.01,0.8067948913574219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,16,1,power_law_1.01,0.8715782165527344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,16,1,power_law_1.01,0.9387340545654297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,16,1,power_law_1.01,1.0005222320556642
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,16,1,power_law_1.01,1.1487814331054689
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,16,1,power_law_1.01,0.11483712196350097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,16,1,power_law_1.01,1.2508992004394532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,16,1,balanced,12.905025634765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,16,1,power_law_1.01,1.6516333007812498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,16,1,power_law_1.01,1.3909254455566407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,16,1,power_law_1.01,1.4734124755859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,16,1,power_law_1.01,1.8082374572753905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,16,1,power_law_1.01,2.142321319580078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,16,1,power_law_1.01,2.7990930175781252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,16,1,power_law_1.01,3.448164367675781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,16,1,power_law_1.01,1.3175482177734374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,16,1,power_law_1.01,4.756332092285156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,16,1,balanced,17.503924560546874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,16,1,power_law_1.01,6.069949951171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,16,1,power_law_1.01,8.720761108398438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,16,1,balanced,26.564926757812497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,16,1,power_law_1.01,11.351591796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,16,1,power_law_1.01,3.222109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,16,1,power_law_1.01,3.820432739257812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,16,1,power_law_1.01,3.493006591796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,16,1,balanced,33.4581005859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,16,1,power_law_1.01,3.636636047363281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,16,1,power_law_1.01,4.1031796264648435
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,16,1,power_law_1.01,3.57451904296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,16,1,power_law_1.01,0.2543571281433105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,16,1,power_law_1.01,0.36467201232910157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,16,1,power_law_1.01,0.8865004730224608
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,16,1,power_law_1.01,3.777492980957031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,16,1,power_law_1.01,3.7816052246093745
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,16,1,power_law_1.01,4.361947631835937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,16,1,power_law_1.01,1.4004742431640627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,16,1,power_law_1.01,3.9789331054687502
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,16,1,power_law_1.01,4.7512097167968745
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,16,1,power_law_1.01,2.0800869750976565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,16,1,power_law_1.01,5.2723846435546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,16,1,power_law_1.01,5.7881689453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,16,1,power_law_1.01,6.9305224609375005
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,16,1,power_law_1.01,7.522791748046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,16,1,power_law_1.2,0.08093567848205566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,16,1,power_law_1.01,7.631187744140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,16,1,power_law_1.2,0.11953472137451171
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,16,1,power_law_1.2,0.2783852767944336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,16,1,power_law_1.2,0.39935745239257814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,16,1,power_law_1.2,0.18853887557983398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,16,1,power_law_1.2,0.6795046234130859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,16,1,power_law_1.2,0.7349874877929687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,16,1,power_law_1.2,0.7571616363525391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,16,1,power_law_1.2,0.7911686706542969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,16,1,power_law_1.2,0.827343978881836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,16,1,power_law_1.2,0.8950208282470703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,16,1,power_law_1.01,11.751473388671874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,16,1,power_law_1.2,0.9709113311767578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,16,1,power_law_1.2,1.0385298919677735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,16,1,power_law_1.2,1.3030720520019532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,16,1,power_law_1.2,1.4314253234863281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,16,1,power_law_1.2,1.1861843109130858
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,16,1,power_law_1.2,1.6928186035156252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,16,1,power_law_1.2,1.3619552612304688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,16,1,power_law_1.01,15.238856201171876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,16,1,power_law_1.2,1.5298457336425781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,16,1,power_law_1.01,3.4610855102539064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,16,1,power_law_1.2,1.8466957092285157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,16,1,power_law_1.2,2.168811492919922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,16,1,power_law_1.2,2.8252276611328124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,16,1,power_law_1.01,18.461378173828127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,16,1,power_law_1.2,3.473867492675781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,16,1,power_law_1.2,0.21705984115600585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,16,1,power_law_1.2,0.45704063415527346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,16,1,power_law_1.2,0.6366425704956055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,16,1,power_law_1.2,4.795687561035156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,16,1,power_law_1.2,1.098177261352539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,16,1,power_law_1.2,1.9248345947265626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,16,1,power_law_1.2,6.098540039062501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,16,1,power_law_1.2,2.631844482421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,16,1,power_law_1.2,3.5231481933593747
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,16,1,power_law_1.2,3.273899536132812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,16,1,power_law_1.2,2.8188705444335938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,16,1,power_law_1.01,26.50244140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,16,1,power_law_1.2,3.151560974121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,16,1,power_law_1.2,3.969967956542969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,16,1,power_law_1.2,3.6415960693359373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,16,1,power_law_1.2,3.1080813598632813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,16,1,power_law_1.2,3.6968249511718754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,16,1,power_law_1.2,4.012842102050781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,16,1,power_law_1.2,8.745691528320313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,16,1,power_law_1.2,4.7201708984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,16,1,power_law_1.2,4.496819763183593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,16,1,power_law_1.2,5.7295361328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,16,1,power_law_1.2,4.967639770507812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,16,1,power_law_1.01,35.1057958984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,16,1,power_law_1.2,7.0091357421875005
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,16,1,power_law_1.2,8.001426391601562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,4,8,balanced,0.05066880226135254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,4,8,balanced,0.04865664005279541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,4,8,balanced,0.048911361694335936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,4,8,balanced,0.11182975769042969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,4,8,balanced,0.1118553638458252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,4,8,balanced,0.11197823524475096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,4,8,balanced,0.11241344451904296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,4,8,balanced,0.11149439811706544
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,4,8,balanced,0.11262335777282714
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,4,8,balanced,0.09961728096008302
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,4,8,balanced,0.10356096267700195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,4,8,balanced,0.10351103782653807
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,4,8,balanced,0.10937472343444825
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,4,8,balanced,0.1128767967224121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,16,1,power_law_1.2,9.220101318359374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,4,8,balanced,0.11699071884155274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,4,8,balanced,0.11656703948974609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,4,8,balanced,0.12032256126403809
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,16,1,power_law_1.2,10.1885009765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,4,8,balanced,0.1302015972137451
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,4,8,balanced,0.13591551780700684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,4,8,balanced,0.16109312057495118
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,4,8,balanced,0.15730303764343262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,4,8,balanced,0.21030656814575197
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,4,8,balanced,0.2530726432800293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,4,8,balanced,0.3500966262817383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,4,8,balanced,0.4509337615966797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,4,8,balanced,0.0354201602935791
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,4,8,balanced,0.6500313568115235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,4,8,balanced,0.035408639907836915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,4,8,balanced,0.03554687976837158
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,4,8,balanced,0.04567552089691162
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,4,8,balanced,0.8494579315185546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,4,8,balanced,0.046071038246154786
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,4,8,balanced,0.04649727821350098
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,4,8,balanced,0.04727935791015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,4,8,balanced,0.04607999801635742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,4,8,balanced,0.04723968029022217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,4,8,balanced,0.04651008129119873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,4,8,balanced,0.04727551937103271
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,4,8,balanced,0.04653183937072754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,4,8,balanced,0.048095998764038084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,4,8,balanced,0.04733823776245118
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,4,8,balanced,0.04768383979797363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,4,8,balanced,0.04941696166992188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,4,8,balanced,0.051811838150024415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,4,8,balanced,0.05454976081848144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,4,8,balanced,0.06777344226837159
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,4,8,balanced,0.06326784133911133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,4,8,balanced,0.08644224166870117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,4,8,balanced,0.04648320198059082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,4,8,balanced,0.09955967903137207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,4,8,balanced,0.1383846378326416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,4,8,balanced,0.16651391983032227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,4,8,balanced,0.2344921684265137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,4,8,balanced,0.3131929588317871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,16,1,power_law_1.2,11.384334716796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,16,1,power_law_1.2,15.184393310546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,16,1,power_law_1.2,14.759107666015627
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,4,8,balanced,0.017994240522384644
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,4,8,balanced,0.017821439504623414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,4,8,balanced,0.017807359695434573
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,4,8,balanced,0.018238719701766968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,4,8,balanced,0.018210560083389282
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,4,8,balanced,0.018167040348052978
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,4,8,balanced,0.018252799510955812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,4,8,balanced,0.01823232054710388
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,4,8,balanced,0.018557440042495727
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,4,8,balanced,0.01815935969352722
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,4,8,balanced,0.01821823954582214
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,4,8,balanced,0.018586879968643187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,4,8,balanced,0.017756160497665405
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,4,8,balanced,0.018254079818725587
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,4,8,balanced,0.018982399702072144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,16,1,power_law_1.2,27.60335205078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,4,8,balanced,0.01860991954803467
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,4,8,balanced,0.018167040348052978
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,4,8,balanced,0.02021888017654419
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,4,8,balanced,0.020357120037078857
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,4,8,balanced,0.025575680732727053
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,4,8,balanced,0.027747840881347657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,16,1,power_law_1.2,34.49302734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,4,8,power_law_1.01,0.05839231967926025
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,4,8,balanced,0.04208000183105469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,4,8,power_law_1.01,0.1070131206512451
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,4,8,power_law_1.01,0.10099455833435059
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,4,8,power_law_1.01,0.10782208442687988
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,4,8,balanced,0.03612927913665771
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,4,8,power_law_1.01,0.1129702377319336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,4,8,power_law_1.01,0.11675264358520507
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,4,8,power_law_1.01,0.09824895858764648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,4,8,power_law_1.01,0.11707776069641114
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,4,8,power_law_1.01,0.1197977638244629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,4,8,power_law_1.01,0.12466815948486328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,4,8,power_law_1.01,0.1286630439758301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,4,8,power_law_1.01,0.13409919738769532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,4,8,power_law_1.01,0.14445695877075196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,4,8,power_law_1.01,0.1495321559906006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,4,8,power_law_1.01,0.11427200317382813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,4,8,power_law_1.01,0.16100479125976563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,4,8,power_law_1.01,0.09668224334716798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,4,8,power_law_1.01,0.19723264694213866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,4,8,power_law_1.01,0.24344575881958005
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,4,8,power_law_1.01,0.29724544525146485
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,4,8,balanced,0.05568895816802979
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,4,8,power_law_1.01,0.5965427017211914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,4,8,power_law_1.01,0.3851583862304687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,4,8,power_law_1.01,0.7307839965820312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,4,8,power_law_1.01,0.03625087976455689
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,4,8,power_law_1.01,0.04494080066680908
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,4,8,power_law_1.01,0.046083841323852535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,4,8,power_law_1.01,0.04609151840209961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,4,8,power_law_1.01,0.04650496006011963
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,4,8,power_law_1.01,0.04649727821350098
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,4,8,power_law_1.01,0.046937599182128906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,4,8,power_law_1.01,0.04763519763946533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,4,8,power_law_1.01,0.047489280700683596
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,4,8,power_law_1.01,0.0478988790512085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,4,8,power_law_1.01,0.048679680824279786
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,4,8,power_law_1.01,0.049625601768493656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,4,8,power_law_1.01,0.05062784194946289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,4,8,power_law_1.01,0.0528115177154541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,4,8,power_law_1.01,0.057966079711914066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,4,8,power_law_1.01,0.05944064140319825
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,4,8,power_law_1.01,0.06327807903289795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,4,8,power_law_1.01,0.0844595241546631
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,4,8,power_law_1.01,1.1640345764160156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,4,8,power_law_1.01,0.09553279876708984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,4,8,power_law_1.01,1.5421157836914063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,4,8,power_law_1.01,0.1183232021331787
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,4,8,power_law_1.01,0.15150336265563963
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,4,8,balanced,0.07209727764129639
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,4,8,power_law_1.01,0.21672960281372072
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,4,8,balanced,0.10027903556823729
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,4,8,balanced,0.125153284072876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,4,8,power_law_1.01,0.2612556838989258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,4,8,power_law_1.01,0.017859840393066408
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,4,8,power_law_1.01,0.018231040239334105
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,4,8,power_law_1.01,0.017822719812393188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,4,8,power_law_1.01,0.40164352416992183
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,4,8,power_law_1.01,0.018256640434265135
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,4,8,power_law_1.01,0.018210560083389282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,4,8,power_law_1.01,2.446000671386719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,4,8,power_law_1.01,0.018283519744873047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,4,8,power_law_1.01,0.019216639995574953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,4,8,power_law_1.01,0.01864575982093811
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,4,8,power_law_1.01,0.019022079706192015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,4,8,power_law_1.01,0.843526382446289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,4,8,power_law_1.01,0.593433609008789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,4,8,power_law_1.01,0.018202879428863526
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,4,8,power_law_1.01,1.0498854064941407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,4,8,power_law_1.01,0.01783679962158203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,4,8,power_law_1.01,0.019468799829483033
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,4,8,power_law_1.01,0.019429119825363158
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,4,8,power_law_1.01,0.033475840091705324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,4,8,power_law_1.01,3.3005862426757817
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,4,8,power_law_1.01,0.020285439491271973
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,4,8,power_law_1.01,0.0245632004737854
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,4,8,power_law_1.01,0.025582079887390134
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,4,8,power_law_1.01,0.026800639629364013
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,4,8,power_law_1.01,0.05444352149963379
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,4,8,power_law_1.01,0.041982722282409665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,4,8,power_law_1.2,0.04885119915008545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,4,8,power_law_1.2,0.10790783882141113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,4,8,power_law_1.2,0.08964096069335938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,4,8,power_law_1.2,0.09074432373046874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,4,8,power_law_1.2,0.10869759559631346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,4,8,power_law_1.2,0.10995967864990235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,4,8,power_law_1.2,0.11255552291870116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,4,8,power_law_1.2,0.11482239723205567
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,4,8,power_law_1.2,0.11657216072082519
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,4,8,power_law_1.01,0.06534143924713134
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,4,8,power_law_1.2,0.11768192291259764
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,4,8,power_law_1.2,0.12057087898254396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,4,8,power_law_1.2,0.12345855712890626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,4,8,power_law_1.2,0.12919808387756349
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,4,8,power_law_1.2,0.13434240341186524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,4,8,power_law_1.2,0.14156288146972656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,4,8,power_law_1.2,0.1530790424346924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,4,8,power_law_1.2,0.16663679122924804
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,4,8,power_law_1.01,0.09312000274658203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,4,8,power_law_1.2,0.1988492774963379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,4,8,power_law_1.2,0.3120652770996094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,4,8,power_law_1.2,0.24934400558471678
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,4,8,power_law_1.2,0.4402380752563476
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,4,8,power_law_1.01,0.11325823783874511
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,4,8,power_law_1.2,0.7605542755126953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,4,8,power_law_1.2,0.5991321563720703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,4,8,power_law_1.01,0.19735679626464844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,4,8,power_law_1.2,0.03512576103210449
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,4,8,power_law_1.2,0.044396800994873045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,4,8,power_law_1.2,0.04600704193115234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,4,8,power_law_1.2,0.04578944206237793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,4,8,power_law_1.2,0.04650496006011963
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,4,8,power_law_1.2,0.046535677909851074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,4,8,power_law_1.2,0.04687104225158691
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,4,8,power_law_1.2,1.2735948944091797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,4,8,power_law_1.2,0.047007999420166015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,4,8,power_law_1.2,0.04811391830444336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,4,8,power_law_1.2,0.04792319774627686
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,4,8,power_law_1.01,0.16362112045288085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,4,8,power_law_1.2,0.04892799854278564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,4,8,power_law_1.2,0.04979712009429932
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,4,8,power_law_1.2,0.05053696155548095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,4,8,power_law_1.2,0.052771840095520016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,4,8,power_law_1.2,0.05847551822662354
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,4,8,power_law_1.2,0.0645747184753418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,4,8,power_law_1.2,0.0688422393798828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,4,8,power_law_1.2,0.09742207527160644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,4,8,power_law_1.2,0.08481280326843262
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,4,8,power_law_1.01,0.31581567764282226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,4,8,power_law_1.2,0.1264256000518799
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,4,8,power_law_1.2,0.16115840911865234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,4,8,power_law_1.2,0.2157913589477539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,4,8,power_law_1.2,0.4408742523193359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,4,8,power_law_1.2,0.2989068794250488
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,4,8,power_law_1.2,0.017800960540771484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,4,8,power_law_1.2,0.0178547203540802
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,4,8,power_law_1.2,0.018246400356292724
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,4,8,power_law_1.2,0.017815040349960325
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,4,8,power_law_1.2,0.017849600315093993
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,4,8,power_law_1.2,0.018228479623794556
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,4,8,power_law_1.2,0.018250240087509157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,4,8,power_law_1.2,0.01819391965866089
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,4,8,power_law_1.2,0.01820799946784973
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,4,8,power_law_1.2,0.018617600202560425
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,4,8,power_law_1.2,0.01855232000350952
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,4,8,power_law_1.2,0.01984768033027649
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,4,8,power_law_1.2,0.019900159835815428
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,4,8,power_law_1.2,0.02025984048843384
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,4,8,power_law_1.2,1.6675672912597654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,4,8,power_law_1.2,0.8281613159179688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,4,8,power_law_1.01,0.3589708709716797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,4,8,power_law_1.2,2.398096618652344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,4,8,power_law_1.2,0.592366065979004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,4,8,power_law_1.2,3.4172839355468754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,4,8,power_law_1.2,1.0894143676757813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,4,8,power_law_1.2,0.025331199169158936
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,4,8,power_law_1.2,0.024665598869323728
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,4,8,power_law_1.2,0.027263998985290527
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,4,8,power_law_1.2,0.031010560989379883
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,4,8,power_law_1.2,0.04233344078063965
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,4,8,power_law_1.2,0.0525875186920166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,8,4,balanced,0.030108160972595215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,8,4,balanced,0.030265600681304933
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,8,4,balanced,0.04925183773040771
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,8,4,balanced,0.08465663909912109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,8,4,balanced,0.08470399856567383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,8,4,balanced,0.08535679817199707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,8,4,balanced,0.0857100772857666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,8,4,balanced,0.08822527885437012
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,8,4,balanced,0.09093759536743165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,8,4,balanced,0.09119615554809571
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,8,4,balanced,0.09171072006225586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,8,4,balanced,0.08996735572814941
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,8,4,balanced,0.0919052791595459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,8,4,balanced,0.09593471527099609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,8,4,balanced,0.09635199546813965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,8,4,balanced,0.09743743896484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,8,4,balanced,0.1042841625213623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,8,4,balanced,0.12120575904846191
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,8,4,balanced,0.1103116798400879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,8,4,balanced,0.1280627155303955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,8,4,balanced,0.1548313617706299
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,8,4,balanced,0.21147903442382812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,8,4,balanced,0.2574118423461914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,8,4,balanced,0.36935680389404296
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,4,8,power_law_1.2,0.07147903919219971
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,8,4,balanced,0.4797158432006836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,8,4,balanced,0.7021286773681641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,8,4,balanced,0.030920960903167725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,8,4,balanced,0.9273945617675782
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,4,8,power_law_1.2,0.09331968307495117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,8,4,balanced,0.03044991970062256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,8,4,balanced,0.03298815965652466
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,8,4,balanced,0.039777278900146484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,8,4,balanced,0.040332798957824705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,8,4,balanced,0.03994368076324463
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,8,4,balanced,0.04079103946685791
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,8,4,balanced,0.040724477767944335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,8,4,balanced,0.040746240615844725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,8,4,balanced,0.04051328182220459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,8,4,balanced,0.041587200164794925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,8,4,balanced,0.041218562126159666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,8,4,balanced,0.04233727931976318
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,8,4,balanced,0.04249599933624267
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,8,4,balanced,0.04322815895080566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,8,4,balanced,0.044026880264282225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,8,4,balanced,0.04700160026550293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,8,4,balanced,0.05126783847808838
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,8,4,balanced,0.05139328002929687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,8,4,balanced,0.061656317710876464
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,8,4,balanced,0.07207295894622803
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,8,4,balanced,0.10051967620849608
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,8,4,balanced,0.1168499183654785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,8,4,balanced,0.1606937599182129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,8,4,balanced,0.21620096206665038
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,8,4,balanced,0.3031206321716309
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,8,4,balanced,0.40318462371826175
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,4,8,power_law_1.2,0.22805248260498048
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,4,8,power_law_1.2,0.12095104217529298
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,4,8,power_law_1.2,0.17633920669555664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,8,4,balanced,0.017345279455184937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,8,4,balanced,0.017743359804153442
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,8,4,balanced,0.01742591977119446
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,8,4,balanced,0.017847039699554444
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,8,4,balanced,0.017812479734420777
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,8,4,balanced,0.01740288019180298
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,8,4,balanced,0.0181388795375824
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,8,4,balanced,0.0186572802066803
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,8,4,balanced,0.01823232054710388
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,8,4,balanced,0.01855487942695618
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,8,4,balanced,0.01866495966911316
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,8,4,balanced,0.01861631989479065
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,8,4,balanced,0.01854464054107666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,8,4,balanced,0.01881600022315979
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,8,4,balanced,0.017789440155029295
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,4,8,power_law_1.2,0.42132480621337887
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,4,8,power_law_1.2,0.3079539108276367
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,8,4,balanced,0.017791999578475954
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,8,4,balanced,0.01778048038482666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,8,4,balanced,0.02805248022079468
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,8,4,balanced,0.02141439914703369
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,8,4,balanced,0.020660479068756104
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,8,4,power_law_1.01,0.035814399719238284
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,8,4,power_law_1.01,0.047848958969116215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,8,4,power_law_1.01,0.08079360008239747
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,8,4,power_law_1.01,0.08937215805053711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,8,4,power_law_1.01,0.08951168060302735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,8,4,power_law_1.01,0.08945023536682128
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,8,4,power_law_1.01,0.09163392066955567
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,8,4,power_law_1.01,0.0920627212524414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,8,4,power_law_1.01,0.09499520301818848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,8,4,power_law_1.01,0.0954867172241211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,8,4,power_law_1.01,0.09832703590393066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,8,4,power_law_1.01,0.09974528312683105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,8,4,power_law_1.01,0.10257408142089844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,8,4,power_law_1.01,0.10674688339233399
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,8,4,power_law_1.01,0.11423999786376952
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,8,4,balanced,0.029724159240722657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,8,4,power_law_1.01,0.1253004837036133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,8,4,power_law_1.01,0.12181376457214355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,8,4,power_law_1.01,0.1427353572845459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,8,4,power_law_1.01,0.23943679809570312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,8,4,power_law_1.01,0.18402687072753907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,8,4,power_law_1.01,0.3250137710571289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,8,4,power_law_1.01,0.4043801498413086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,8,4,balanced,0.040687360763549806
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,8,4,balanced,0.04983168125152588
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,8,4,power_law_1.01,0.5035993576049804
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,8,4,balanced,0.06556032180786134
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,8,4,power_law_1.01,0.030446081161499022
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,8,4,power_law_1.01,0.031896319389343265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,8,4,power_law_1.01,0.039263999462127684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,8,4,power_law_1.01,0.03996927976608276
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,8,4,power_law_1.01,0.04009088039398193
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,8,4,power_law_1.01,0.0407539176940918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,8,4,power_law_1.01,0.04075007915496826
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,8,4,power_law_1.01,0.041137919425964356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,8,4,power_law_1.01,0.041301760673522946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,8,4,power_law_1.01,0.042000641822814946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,8,4,power_law_1.01,0.7164415740966797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,8,4,power_law_1.01,0.04285696029663086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,8,4,power_law_1.01,0.04367104053497314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,8,4,power_law_1.01,0.04441472053527832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,8,4,power_law_1.01,0.04648064136505127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,8,4,power_law_1.01,0.04821375846862793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,8,4,power_law_1.01,0.05060480117797852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,8,4,power_law_1.01,0.053738241195678715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,8,4,power_law_1.01,0.06513792037963867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,8,4,power_law_1.01,0.07669119834899903
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,8,4,balanced,0.08078335762023926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,8,4,power_law_1.01,0.09461248397827149
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,8,4,power_law_1.01,0.12445055961608886
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,8,4,power_law_1.01,0.9347980499267579
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,8,4,power_law_1.01,0.20699136734008788
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,8,4,power_law_1.01,0.16888959884643556
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,8,4,balanced,0.13926783561706543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,8,4,balanced,0.1120025634765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,8,4,power_law_1.01,0.017495039701461792
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,8,4,power_law_1.01,0.01739392042160034
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,8,4,power_law_1.01,0.017404160499572753
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,8,4,power_law_1.01,0.017802239656448365
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,8,4,power_law_1.01,0.017827839851379396
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,8,4,power_law_1.01,0.017918720245361328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,8,4,power_law_1.01,0.017876479625701904
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,8,4,power_law_1.01,0.018309119939804077
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,8,4,power_law_1.01,0.01814272046089172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,8,4,power_law_1.01,0.018238719701766968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,8,4,power_law_1.01,0.018488320112228392
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,8,4,power_law_1.01,0.019397120475769043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,8,4,power_law_1.01,0.02034048080444336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,8,4,power_law_1.01,0.37230720520019533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,8,4,power_law_1.01,0.28795007705688475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,8,4,power_law_1.01,0.5356211090087891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,8,4,power_law_1.01,1.8764633178710937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,8,4,power_law_1.01,1.3944268798828126
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,8,4,power_law_1.01,0.019549440145492553
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,8,4,power_law_1.01,0.021047039031982424
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,8,4,power_law_1.01,0.02190079927444458
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,8,4,power_law_1.01,0.025017600059509277
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,8,4,power_law_1.01,0.032001280784606935
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,8,4,power_law_1.01,0.03584768056869507
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,8,4,power_law_1.01,0.04727551937103271
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,8,4,power_law_1.01,0.9023590087890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,8,4,power_law_1.2,0.04215295791625977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,8,4,power_law_1.2,0.04501247882843017
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,8,4,power_law_1.2,0.06891392230987549
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,8,4,power_law_1.2,0.07687168121337891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,8,4,power_law_1.2,0.08111871719360352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,8,4,power_law_1.2,0.0843712043762207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,8,4,power_law_1.2,0.0875814437866211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,8,4,power_law_1.2,0.0895680046081543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,8,4,power_law_1.2,0.0905907154083252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,8,4,power_law_1.2,0.09239168167114258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,8,4,power_law_1.2,0.09484416007995605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,8,4,power_law_1.2,0.09783679962158202
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,8,4,power_law_1.2,0.09899264335632324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,8,4,power_law_1.2,0.10560256004333497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,8,4,power_law_1.2,0.11482624053955078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,8,4,power_law_1.2,0.12245120048522948
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,8,4,power_law_1.2,0.12574336051940918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,8,4,power_law_1.2,0.14390912055969238
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,8,4,power_law_1.2,0.18397823333740235
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,8,4,power_law_1.01,0.058150401115417485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,8,4,power_law_1.2,0.2364441680908203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,8,4,power_law_1.01,0.09389311790466308
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,8,4,power_law_1.2,0.28543872833251954
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,8,4,power_law_1.01,0.07374847888946533
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,8,4,power_law_1.01,0.12011136054992674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,8,4,power_law_1.2,0.3997350311279297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,8,4,power_law_1.2,0.49292545318603515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,8,4,power_law_1.2,0.03134464025497437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,8,4,power_law_1.2,0.03248255968093872
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,8,4,power_law_1.2,0.03954303979873657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,8,4,power_law_1.2,0.03996416091918945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,8,4,power_law_1.2,0.04038527965545654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,8,4,power_law_1.2,0.04078976154327393
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,8,4,power_law_1.2,0.04136703968048096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,8,4,power_law_1.2,0.04158592224121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,8,4,power_law_1.2,0.0419814395904541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,8,4,power_law_1.2,0.04200319766998291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,8,4,power_law_1.2,0.04354559898376465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,8,4,power_law_1.2,0.04404096126556396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,8,4,power_law_1.2,0.045342721939086915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,8,4,power_law_1.2,0.04706943988800048
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,8,4,power_law_1.2,0.04989183902740478
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,8,4,power_law_1.2,0.05086080074310303
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,8,4,power_law_1.2,0.05230463981628418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,8,4,power_law_1.2,0.06442880153656005
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,8,4,power_law_1.2,0.07728511810302734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,8,4,power_law_1.2,0.09808128356933594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,8,4,power_law_1.01,0.15796863555908203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,8,4,power_law_1.2,0.7142387390136719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,8,4,power_law_1.2,0.12001664161682128
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,8,4,power_law_1.2,0.16312320709228517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,8,4,power_law_1.2,0.2048588752746582
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,8,4,power_law_1.01,0.2623372840881347
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,8,4,power_law_1.2,0.3149350357055664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,8,4,power_law_1.2,0.8966873931884767
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,8,4,power_law_1.2,0.01771775960922241
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,8,4,power_law_1.2,0.017489919662475588
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,8,4,power_law_1.2,0.017797119617462158
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,8,4,power_law_1.2,0.017653759717941284
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,8,4,power_law_1.2,0.017807359695434573
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,8,4,power_law_1.2,0.017791999578475954
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,8,4,power_law_1.2,0.018186240196228026
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,8,4,power_law_1.2,0.018191360235214234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,8,4,power_law_1.2,0.018220800161361694
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,8,4,power_law_1.2,0.018126080036163332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,8,4,power_law_1.2,0.01943935990333557
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,8,4,power_law_1.01,0.2264998435974121
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,8,4,power_law_1.2,0.020185599327087404
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,8,4,power_law_1.2,0.02142848014831543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,8,4,power_law_1.2,0.022911999225616455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,8,4,power_law_1.2,0.5970291137695313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,8,4,power_law_1.2,1.3616358947753906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,8,4,power_law_1.2,0.42095489501953126
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,8,4,power_law_1.2,0.04490111827850342
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,8,4,power_law_1.2,1.8487257385253906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,8,4,power_law_1.2,0.01738368034362793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,8,4,power_law_1.2,0.10075903892517091
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,8,4,power_law_1.2,0.019576319456100465
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,8,4,power_law_1.2,0.025946879386901857
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,8,4,power_law_1.2,0.03281791925430298
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,8,4,power_law_1.2,0.0363916802406311
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,16,2,balanced,0.026754560470581057
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,16,2,balanced,0.028853759765624997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,16,2,balanced,0.04306431770324707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,16,2,balanced,0.06134016036987304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,16,2,balanced,0.061552639007568355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,16,2,balanced,0.07612800121307373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,8,4,power_law_1.2,0.8443942260742187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,16,2,balanced,0.07599487781524658
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,16,2,balanced,0.07664000034332276
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,16,2,balanced,0.07578623771667481
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,16,2,balanced,0.08180095672607421
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,16,2,balanced,0.0813811206817627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,16,2,balanced,0.08415616035461426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,16,2,balanced,0.0890777587890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,16,2,balanced,0.09344511985778808
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,16,2,balanced,0.09979264259338379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,16,2,balanced,0.10240768432617187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,16,2,balanced,0.09562623977661133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,16,2,balanced,0.1065727996826172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,16,2,balanced,0.11751040458679199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,16,2,balanced,0.1397670364379883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,16,2,balanced,0.16596864700317382
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,16,2,balanced,0.23772544860839845
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,8,4,power_law_1.2,0.055490560531616204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,16,2,balanced,0.2984896087646484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,16,2,balanced,0.4320127868652344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,16,2,balanced,0.5667737579345703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,16,2,balanced,0.028078079223632812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,16,2,balanced,0.8301964569091798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,16,2,balanced,0.02971519947052002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,16,2,balanced,0.03151103973388672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,16,2,balanced,0.03708159923553467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,16,2,balanced,0.037414400577545165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,16,2,balanced,0.037470719814300536
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,16,2,balanced,0.037722880840301516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,16,2,balanced,1.097323532104492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,16,2,balanced,0.038302719593048096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,16,2,balanced,0.0387059211730957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,16,2,balanced,0.03929343938827515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,16,2,balanced,0.04036096096038818
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,16,2,balanced,0.040770559310913085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,16,2,balanced,0.04278656005859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,16,2,balanced,0.048093438148498535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,16,2,balanced,0.047713279724121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,16,2,balanced,0.04599936008453369
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,8,4,power_law_1.2,0.07554944038391113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,16,2,balanced,0.05585023880004882
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,16,2,balanced,0.06299776077270508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,16,2,balanced,0.07726975917816162
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,16,2,balanced,0.09018112182617187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,16,2,balanced,0.03797631978988648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,16,2,balanced,0.12964863777160646
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,16,2,balanced,0.1622118377685547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,16,2,balanced,0.23325439453125002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,16,2,balanced,0.3046284866333008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,16,2,balanced,0.44092544555664065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,16,2,balanced,0.580371208190918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,8,4,power_law_1.2,0.16275199890136718
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,8,4,power_law_1.2,0.12960639953613282
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,8,4,power_law_1.2,0.23089664459228515
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,16,2,balanced,0.017845760583877563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,16,2,balanced,0.018763519525527954
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,16,2,balanced,0.01862015962600708
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,16,2,balanced,0.018220800161361694
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,16,2,balanced,0.019037439823150634
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,16,2,balanced,0.018240000009536746
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,16,2,balanced,0.01938688039779663
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,16,2,balanced,0.019691519737243653
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,16,2,balanced,0.019879679679870605
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,16,2,balanced,0.02023168087005615
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,16,2,balanced,0.018160640001296996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,16,2,balanced,0.0201907205581665
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,16,2,balanced,0.019022079706192015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,16,2,balanced,0.019037439823150634
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,8,4,power_law_1.2,0.26973567962646483
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,16,2,balanced,0.01858944058418274
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,16,2,balanced,0.018540799617767334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,16,2,balanced,0.01857792019844055
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,16,2,balanced,0.023824639320373535
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,16,2,balanced,0.032538878917694095
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,16,2,balanced,0.02419327974319458
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,16,2,power_law_1.01,0.02513024091720581
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,16,2,power_law_1.01,0.036487679481506344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,16,2,power_law_1.01,0.03790848016738892
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,16,2,power_law_1.01,0.07068927764892578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,16,2,balanced,0.03833728075027466
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,16,2,power_law_1.01,0.06885888099670409
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,16,2,power_law_1.01,0.07284480094909668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,16,2,power_law_1.01,0.07616511821746827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,16,2,power_law_1.01,0.07373311996459961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,16,2,power_law_1.01,0.0787340784072876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,16,2,power_law_1.01,0.07900288105010986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,16,2,power_law_1.01,0.08299903869628907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,16,2,power_law_1.01,0.08559103965759277
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,16,2,power_law_1.01,0.08912256240844726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,16,2,power_law_1.01,0.0929958438873291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,16,2,power_law_1.01,0.10230912208557127
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,16,2,balanced,0.053600001335144046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,16,2,power_law_1.01,0.10485376358032225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,16,2,power_law_1.01,0.10214271545410156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,16,2,power_law_1.01,0.11295104026794434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,16,2,power_law_1.01,0.1757516860961914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,16,2,power_law_1.01,0.13828351974487305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,16,2,balanced,0.06160255908966065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,16,2,power_law_1.01,0.21508735656738281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,16,2,power_law_1.01,0.3498316955566406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,16,2,power_law_1.01,0.6557414245605468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,16,2,power_law_1.01,0.275545597076416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,16,2,power_law_1.01,0.02811392068862915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,16,2,power_law_1.01,0.030094079971313477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,16,2,power_law_1.01,0.03091455936431885
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,16,2,power_law_1.01,0.03707007884979248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,16,2,power_law_1.01,0.03746687889099121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,16,2,power_law_1.01,0.037474560737609866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,16,2,power_law_1.01,0.038311679363250736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,16,2,power_law_1.01,0.03827199935913086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,16,2,power_law_1.01,0.038675200939178464
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,16,2,power_law_1.01,0.03909888029098511
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,16,2,power_law_1.01,0.0397708797454834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,16,2,power_law_1.01,0.040761599540710444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,16,2,power_law_1.01,0.04157695770263672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,16,2,power_law_1.01,0.04361472129821777
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,16,2,power_law_1.01,0.046456317901611324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,16,2,power_law_1.01,0.050007038116455085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,16,2,power_law_1.01,0.048519678115844726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,16,2,power_law_1.01,0.05851903915405273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,16,2,power_law_1.01,0.0647103977203369
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,16,2,balanced,0.08392319679260254
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,16,2,balanced,0.10439167976379396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,16,2,power_law_1.01,0.0798310422897339
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,16,2,power_law_1.01,0.1011673641204834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,16,2,power_law_1.01,0.5580863952636719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,16,2,power_law_1.01,0.32756481170654295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,16,2,power_law_1.01,0.13887999534606935
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,16,2,balanced,0.14457088470458984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,16,2,power_law_1.01,0.017789440155029295
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,16,2,power_law_1.01,0.017830400466918944
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,16,2,power_law_1.01,0.01820032000541687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,16,2,balanced,0.18865280151367186
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,16,2,power_law_1.01,0.018201600313186645
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,16,2,power_law_1.01,0.01875200033187866
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,16,2,power_law_1.01,0.01868288040161133
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,16,2,power_law_1.01,0.01869312047958374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,16,2,power_law_1.01,0.16945024490356447
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,16,2,power_law_1.01,0.019069440364837646
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,16,2,power_law_1.01,0.018963199853897095
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,16,2,power_law_1.01,0.01903488039970398
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,16,2,power_law_1.01,0.01946239948272705
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,16,2,power_law_1.01,0.02106879949569702
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,16,2,power_law_1.01,0.24403200149536133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,16,2,power_law_1.01,0.9163967895507813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,16,2,power_law_1.01,1.2961164855957032
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,16,2,power_law_1.01,0.032320001125335694
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,16,2,power_law_1.01,0.48485759735107425
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,16,2,power_law_1.01,0.022649600505828857
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,16,2,power_law_1.01,0.02193023920059204
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,16,2,power_law_1.01,0.021507840156555175
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,16,2,power_law_1.01,0.023128321170806883
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,16,2,power_law_1.01,0.0248524808883667
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,16,2,power_law_1.01,0.03675519943237305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,16,2,power_law_1.01,0.04229119777679443
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,16,2,power_law_1.01,0.6834713745117187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,16,2,power_law_1.2,0.028446719646453855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,16,2,power_law_1.2,0.03943936109542846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,16,2,power_law_1.2,0.041232638359069824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,16,2,power_law_1.2,0.07988736152648926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,16,2,power_law_1.2,0.08050047874450683
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,16,2,power_law_1.2,0.08046591758728028
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,16,2,power_law_1.2,0.08169983863830567
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,16,2,power_law_1.2,0.08178560256958008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,16,2,power_law_1.2,0.08441344261169434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,16,2,power_law_1.2,0.08399999618530274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,16,2,power_law_1.2,0.08598015785217286
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,16,2,power_law_1.2,0.0861235237121582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,16,2,power_law_1.2,0.08870271682739259
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,16,2,power_law_1.2,0.09351807594299316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,16,2,power_law_1.2,0.09943936347961427
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,16,2,power_law_1.2,0.10461183547973633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,16,2,power_law_1.2,0.10428159713745118
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,16,2,power_law_1.2,0.11708031654357912
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,16,2,power_law_1.2,0.13859328269958496
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,16,2,power_law_1.01,0.06137983798980713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,16,2,power_law_1.2,0.17183231353759765
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,16,2,power_law_1.01,0.07230336189270019
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,16,2,power_law_1.01,0.08552576065063476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,16,2,power_law_1.2,0.2816524887084961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,16,2,power_law_1.2,0.21347583770751952
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,16,2,power_law_1.01,0.12501888275146483
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,16,2,power_law_1.2,0.46488574981689457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,16,2,power_law_1.2,0.028053760528564453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,16,2,power_law_1.2,0.36416641235351566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,16,2,power_law_1.2,0.02991744041442871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,16,2,power_law_1.2,0.030109438896179196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,16,2,power_law_1.2,0.03662336111068726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,16,2,power_law_1.2,0.037084159851074217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,16,2,power_law_1.2,0.03747711896896362
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,16,2,power_law_1.2,0.03795200109481812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,16,2,power_law_1.2,0.0382809591293335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,16,2,power_law_1.2,0.038278400897979736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,16,2,power_law_1.2,0.03910399913787842
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,16,2,power_law_1.2,0.03992448091506958
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,16,2,power_law_1.2,0.04074751853942871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,16,2,power_law_1.2,0.04329984188079834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,16,2,power_law_1.2,0.043550720214843755
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,16,2,power_law_1.2,0.7244467163085938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,16,2,power_law_1.2,0.048209919929504394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,16,2,power_law_1.2,0.04868607997894287
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,16,2,power_law_1.2,0.05294847965240479
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,16,2,power_law_1.2,0.06452479839324951
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,16,2,power_law_1.2,0.0586188793182373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,16,2,power_law_1.2,0.08503295898437499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,16,2,power_law_1.2,0.10422400474548339
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,16,2,power_law_1.2,0.1375603199005127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,16,2,power_law_1.2,0.2586355209350586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,16,2,power_law_1.01,0.24753536224365233
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,16,2,power_law_1.2,0.1825651168823242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,16,2,power_law_1.2,0.017830400466918944
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,16,2,power_law_1.2,0.017909760475158694
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,16,2,power_law_1.01,0.30747648239135744
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,16,2,power_law_1.2,0.018197760581970215
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,16,2,power_law_1.2,0.01856256008148193
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,16,2,power_law_1.2,0.0182041597366333
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,16,2,power_law_1.2,0.018598400354385376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,16,2,power_law_1.2,0.018978559970855714
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,16,2,power_law_1.2,0.019073280096054076
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,16,2,power_law_1.2,0.018959360122680662
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,16,2,power_law_1.2,0.0189900803565979
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,16,2,power_law_1.2,0.019434239864349365
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,16,2,power_law_1.2,0.02148736000061035
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,16,2,power_law_1.2,0.021416959762573243
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,16,2,power_law_1.2,0.02270335912704468
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,16,2,power_law_1.2,0.023054080009460447
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,16,2,power_law_1.01,0.14943488121032716
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,16,2,power_law_1.2,0.9220301055908203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,16,2,power_law_1.2,0.49907070159912104
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,16,2,power_law_1.2,0.332940788269043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,16,2,power_law_1.2,0.021456639766693115
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,16,2,power_law_1.2,0.025102078914642334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,16,2,power_law_1.2,0.030922238826751706
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,16,2,power_law_1.2,0.03908224105834961
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,16,2,power_law_1.2,0.04679423809051514
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,16,2,power_law_1.2,0.05231359958648681
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,16,2,power_law_1.2,0.07154816150665283
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,32,1,balanced,0.030390400886535645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,32,1,balanced,0.03292351961135864
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,32,1,balanced,0.04847040176391602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,32,1,balanced,0.048877439498901366
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,32,1,balanced,0.07547520160675049
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,32,1,balanced,0.077674880027771
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,16,2,power_law_1.2,0.6947071838378907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,32,1,balanced,0.07898687839508056
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,32,1,balanced,0.07672768115997315
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,32,1,balanced,0.0805958366394043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,32,1,balanced,0.08165120124816895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,32,1,balanced,0.08232000350952148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,32,1,balanced,0.08430912017822265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,32,1,balanced,0.08573311805725098
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,32,1,balanced,0.08809151649475097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,32,1,balanced,0.09561920166015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,32,1,balanced,0.09858367919921876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,32,1,balanced,0.09499648094177246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,32,1,balanced,0.10686783790588379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,32,1,balanced,0.11530048370361327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,32,1,balanced,0.13436991691589356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,32,1,balanced,0.17010879516601562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,32,1,balanced,0.23409215927124025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,32,1,balanced,0.2839385604858399
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,32,1,balanced,0.40944446563720704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,32,1,balanced,0.5352678298950195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,32,1,balanced,0.7778963470458985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,32,1,balanced,1.0249440002441408
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,16,2,power_law_1.2,1.5360525512695313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,32,1,balanced,0.03418623924255371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,32,1,balanced,0.03530944108963013
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,32,1,balanced,0.03790719985961914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,16,2,power_law_1.2,0.0914419174194336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,32,1,balanced,0.036556799411773686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,32,1,balanced,0.043471360206604005
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,32,1,balanced,0.04553599834442139
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,32,1,balanced,0.04597951889038086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,32,1,balanced,0.046809601783752444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,32,1,balanced,0.04761727809906006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,32,1,balanced,0.04789567947387695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,32,1,balanced,0.050091519355773925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,32,1,balanced,0.05303743839263916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,16,2,power_law_1.2,0.12200960159301757
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,32,1,balanced,0.04398528099060058
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,32,1,balanced,0.04543935775756836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,32,1,balanced,0.0528550386428833
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,32,1,balanced,0.062427520751953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,32,1,balanced,0.06689599990844727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,32,1,balanced,0.08392831802368164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,32,1,balanced,0.09994239807128906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,32,1,balanced,0.14415167808532714
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,32,1,balanced,0.17050304412841796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,32,1,balanced,0.2509766387939453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,32,1,balanced,0.04388415813446045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,32,1,balanced,0.32609920501708983
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,32,1,balanced,0.4773427200317383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,32,1,balanced,0.615093765258789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,32,1,balanced,0.05663040161132813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,16,2,power_law_1.2,0.2481023979187012
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,16,2,power_law_1.2,0.17494144439697265
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,32,1,balanced,0.018125439882278445
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,32,1,balanced,0.01915071964263916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,32,1,balanced,0.020949120521545413
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,32,1,balanced,0.021157119274139404
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,32,1,balanced,0.020983679294586183
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,32,1,balanced,0.021071999073028563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,32,1,balanced,0.020991361141204833
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,16,2,power_law_1.2,0.2737651252746582
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,32,1,balanced,0.021203200817108154
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,32,1,balanced,0.021245439052581788
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,32,1,balanced,0.022644479274749756
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,32,1,balanced,0.02260672092437744
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,32,1,balanced,0.020576000213623047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,32,1,balanced,0.02054784059524536
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,32,1,balanced,0.020383999347686768
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,32,1,balanced,0.02099519968032837
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,32,1,balanced,0.020666880607604982
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,32,1,balanced,0.022311038970947265
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,32,1,balanced,0.030496640205383303
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,32,1,balanced,0.043055357933044436
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,32,1,balanced,0.028337919712066652
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,32,1,power_law_1.01,0.03247872114181519
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,32,1,power_law_1.01,0.03922431945800781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,32,1,power_law_1.01,0.046129279136657715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,32,1,power_law_1.01,0.04670335769653321
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,32,1,balanced,0.049456639289855955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,32,1,power_law_1.01,0.07932096004486085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,32,1,power_law_1.01,0.07425536155700684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,32,1,power_law_1.01,0.08194560050964356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,32,1,power_law_1.01,0.07563392162322999
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,32,1,power_law_1.01,0.08250752449035645
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,32,1,balanced,0.06852672100067139
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,32,1,power_law_1.01,0.08317760467529298
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,32,1,power_law_1.01,0.08764032363891602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,32,1,power_law_1.01,0.08776896476745606
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,32,1,balanced,0.08258432388305664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,32,1,power_law_1.01,0.08990143775939942
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,32,1,power_law_1.01,0.09427264213562012
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,32,1,power_law_1.01,0.09714943885803222
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,32,1,power_law_1.01,0.10003711700439452
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,32,1,power_law_1.01,0.10172991752624512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,32,1,power_law_1.01,0.11267200469970703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,32,1,power_law_1.01,0.12384960174560547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,32,1,power_law_1.01,0.14861184120178222
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,32,1,power_law_1.01,0.1805286407470703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,32,1,power_law_1.01,0.2475872039794922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,32,1,power_law_1.01,0.30708160400390627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,32,1,power_law_1.01,0.43013374328613285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,32,1,power_law_1.01,0.03205120086669922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,32,1,power_law_1.01,0.5523110580444336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,32,1,balanced,0.11214464187622071
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,32,1,power_law_1.01,0.03365184068679809
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,32,1,power_law_1.01,0.03537983894348144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,32,1,power_law_1.01,0.04351295948028565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,32,1,power_law_1.01,0.04401023864746094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,32,1,power_law_1.01,0.04478655815124512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,32,1,power_law_1.01,0.045598721504211424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,32,1,power_law_1.01,0.04614272117614746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,32,1,power_law_1.01,0.046456317901611324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,32,1,power_law_1.01,0.047674880027771
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,32,1,power_law_1.01,0.04884992122650146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,32,1,power_law_1.01,0.05416575908660889
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,32,1,power_law_1.01,0.05760191917419434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,32,1,power_law_1.01,0.05916287899017334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,32,1,power_law_1.01,0.03656575918197632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,32,1,power_law_1.01,0.05707647800445557
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,32,1,balanced,0.20507648468017575
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,32,1,balanced,0.14374208450317383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,32,1,power_law_1.01,0.07318784236907959
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,32,1,balanced,0.26859840393066403
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,32,1,power_law_1.01,0.042362241744995116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,32,1,power_law_1.01,0.09010623931884766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,32,1,power_law_1.01,0.14513728141784668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,32,1,power_law_1.01,0.06438591957092285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,32,1,power_law_1.01,0.7969548797607422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,32,1,power_law_1.01,0.01811840057373047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,32,1,power_law_1.01,0.17761856079101562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,32,1,power_law_1.01,0.018940800428390504
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,32,1,power_law_1.01,0.019557759761810303
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,32,1,power_law_1.01,0.02030080080032349
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,32,1,power_law_1.01,0.020599040985107422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,32,1,power_law_1.01,0.021002240180969238
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,32,1,power_law_1.01,0.02068160057067871
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,32,1,power_law_1.01,0.02099967956542969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,32,1,power_law_1.01,0.020619521141052248
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,32,1,power_law_1.01,0.020994560718536375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,32,1,power_law_1.01,0.1074233627319336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,32,1,power_law_1.01,0.020979199409484863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,32,1,power_law_1.01,0.25612224578857423
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,32,1,power_law_1.01,0.33521854400634765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,32,1,power_law_1.01,1.044110107421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,32,1,power_law_1.01,0.4870240020751953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,32,1,power_law_1.01,0.02144511938095093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,32,1,power_law_1.01,0.022227840423583983
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,32,1,power_law_1.01,0.02317312002182007
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,32,1,power_law_1.01,0.024634881019592287
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,32,1,power_law_1.01,0.6304499053955078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,32,1,power_law_1.01,0.025121281147003172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,32,1,power_law_1.01,0.028361599445343017
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,32,1,power_law_1.01,0.03348223924636841
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,32,1,power_law_1.01,0.04032832145690918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,32,1,power_law_1.01,0.05193024158477784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,32,1,power_law_1.2,0.02999808073043823
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,32,1,power_law_1.2,0.041637759208679195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,32,1,power_law_1.2,0.04243904113769531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,32,1,power_law_1.2,0.04432447910308838
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,32,1,power_law_1.2,0.06457600116729736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,32,1,power_law_1.2,0.07438720226287841
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,32,1,power_law_1.2,0.07425983905792236
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,32,1,power_law_1.01,0.061114878654479975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,32,1,power_law_1.2,0.07405824184417724
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,32,1,power_law_1.2,0.07744256019592285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,32,1,power_law_1.2,0.08023360252380371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,32,1,power_law_1.2,0.08457728385925292
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,32,1,power_law_1.2,0.08790975570678711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,32,1,power_law_1.2,0.08799296379089355
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,32,1,power_law_1.01,0.08302207946777343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,32,1,power_law_1.2,0.09378432273864747
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,32,1,power_law_1.2,0.09630911827087403
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,32,1,power_law_1.2,0.10107775688171387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,32,1,power_law_1.2,0.10209088325500489
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,32,1,power_law_1.2,0.11228863716125488
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,32,1,power_law_1.01,0.10356160163879395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,32,1,power_law_1.2,0.12408703804016112
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,32,1,power_law_1.2,0.14939071655273437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,32,1,power_law_1.2,0.18094144821166994
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,32,1,power_law_1.01,0.1418553638458252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,32,1,power_law_1.2,0.2473766326904297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,32,1,power_law_1.01,0.18403327941894532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,32,1,power_law_1.2,0.3068595123291016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,32,1,power_law_1.2,0.03613568067550659
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,32,1,power_law_1.2,0.03614720106124878
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,32,1,power_law_1.2,0.037789440155029295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,32,1,power_law_1.2,0.04392000198364258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,32,1,power_law_1.2,0.044327039718627934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,32,1,power_law_1.2,0.04515520095825195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,32,1,power_law_1.2,0.045564160346984864
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,32,1,power_law_1.2,0.04601151943206787
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,32,1,power_law_1.2,0.046794881820678716
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,32,1,power_law_1.2,0.03341248035430908
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,32,1,power_law_1.2,0.04733248233795166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,32,1,power_law_1.2,0.04813759803771973
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,32,1,power_law_1.2,0.050079998970031736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,32,1,power_law_1.2,0.054157438278198244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,32,1,power_law_1.2,0.05915008068084717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,32,1,power_law_1.2,0.06047552108764649
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,32,1,power_law_1.2,0.058256640434265136
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,32,1,power_law_1.2,0.06337664127349854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,32,1,power_law_1.2,0.07508416175842285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,32,1,power_law_1.2,0.43012161254882814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,32,1,power_law_1.2,0.08920639991760254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,32,1,power_law_1.2,0.10659775733947754
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,32,1,power_law_1.01,0.26678016662597653
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,32,1,power_law_1.2,0.14533632278442382
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,32,1,power_law_1.2,0.5543526458740234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,32,1,power_law_1.2,0.17840063095092773
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,32,1,power_law_1.2,0.018149759769439697
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,32,1,power_law_1.2,0.018926719427108767
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,32,1,power_law_1.2,0.018944640159606934
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,32,1,power_law_1.2,0.020547199249267577
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,32,1,power_law_1.2,0.020545918941497803
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,32,1,power_law_1.2,0.02057728052139282
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,32,1,power_law_1.2,0.020981760025024415
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,32,1,power_law_1.2,0.020570240020751952
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,32,1,power_law_1.2,0.020986878871917726
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,32,1,power_law_1.2,0.02058880090713501
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,32,1,power_law_1.2,0.021235840320587156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,32,1,power_law_1.2,0.25603776931762695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,32,1,power_law_1.2,0.022157440185546877
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,32,1,power_law_1.2,0.023444480895996093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,32,1,power_law_1.2,0.02467967987060547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,32,1,power_law_1.2,0.026334080696105953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,32,1,power_law_1.01,0.35015167236328126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,32,1,power_law_1.2,0.7972921752929688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,32,1,power_law_1.2,0.3353184127807617
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,32,1,power_law_1.2,0.02176959991455078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,32,1,power_law_1.2,0.028492159843444824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,32,1,power_law_1.2,1.0433023834228516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,32,1,power_law_1.2,0.034225919246673585
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,32,1,power_law_1.2,0.040644478797912595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,32,1,power_law_1.2,0.48696640014648435
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,32,1,power_law_1.2,0.08574272155761718
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,32,1,power_law_1.2,0.052641282081604
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,4,8,balanced,0.08285951614379883
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,32,1,power_law_1.2,0.10676095962524415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,4,8,balanced,0.08310272216796874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,4,8,balanced,0.08334976196289062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,4,8,balanced,0.15089152336120606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,4,8,balanced,0.15106944084167478
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,4,8,balanced,0.15153152465820313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,4,8,balanced,0.15201536178588868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,4,8,balanced,0.15228416442871093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,4,8,balanced,0.15252479553222656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,4,8,balanced,0.15294848442077638
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,4,8,balanced,0.15251968383789063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,4,8,balanced,0.15466879844665526
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,4,8,balanced,0.15545344352722168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,4,8,balanced,0.1578598403930664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,4,8,balanced,0.1581068801879883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,4,8,balanced,0.15803903579711914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,4,8,balanced,0.1645631980895996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,4,8,balanced,0.16982784271240234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,4,8,balanced,0.18171648025512693
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,4,8,balanced,0.24801279067993165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,4,8,balanced,0.23202560424804686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,4,8,balanced,0.32121982574462893
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,4,8,balanced,0.3945075225830078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,4,8,balanced,0.5632691192626953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,4,8,balanced,0.7223296356201172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,4,8,balanced,1.0775603485107421
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,4,8,balanced,0.04362112045288086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,4,8,balanced,0.04292479991912842
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,4,8,balanced,0.04280704021453857
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,4,8,balanced,1.4471270751953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,4,8,balanced,0.055105280876159665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,4,8,balanced,0.05508480072021484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,4,8,balanced,0.05386367797851562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,4,8,balanced,0.054551038742065426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,4,8,balanced,0.053884158134460444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,4,8,balanced,0.05505919933319092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,4,8,balanced,0.0538969612121582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,4,8,balanced,0.05430655956268311
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,32,1,power_law_1.2,0.63325439453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,4,8,balanced,0.05518847942352295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,4,8,balanced,0.05711103916168213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,4,8,balanced,0.05490687847137451
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,4,8,balanced,0.05676671981811523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,4,8,balanced,0.05809919834136963
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,4,8,balanced,0.05836800098419189
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,4,8,balanced,0.06541183948516846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,4,8,balanced,0.07204351902008058
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,4,8,balanced,0.08525312423706055
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,32,1,power_law_1.2,0.06279359817504883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,4,8,balanced,0.08373632431030273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,4,8,balanced,0.11681407928466796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,4,8,balanced,0.13756928443908692
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,4,8,balanced,0.1850444793701172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,4,8,balanced,0.232488956451416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,4,8,balanced,0.3403443145751953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,4,8,balanced,0.4415488052368164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,32,1,power_law_1.2,0.14721152305603027
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,32,1,power_law_1.2,0.19099456787109376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,4,8,balanced,0.02068095922470093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,4,8,balanced,0.020696320533752442
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,4,8,balanced,0.02022655963897705
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,4,8,balanced,0.020736000537872314
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,4,8,balanced,0.020101120471954344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,4,8,balanced,0.020262401103973388
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,4,8,balanced,0.020666880607604982
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,4,8,balanced,0.020592639446258547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,4,8,balanced,0.020280320644378663
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,4,8,balanced,0.021076478958129884
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,4,8,balanced,0.020317440032958982
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,4,8,balanced,0.020675840377807616
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,4,8,balanced,0.021091840267181396
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,4,8,balanced,0.021470720767974856
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,4,8,balanced,0.02059648036956787
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,32,1,power_law_1.2,0.2773734474182129
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,4,8,balanced,0.02027008056640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,4,8,balanced,0.020213758945465087
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,4,8,balanced,0.02313215970993042
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,4,8,balanced,0.023582720756530763
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,4,8,balanced,0.031752960681915285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,4,8,balanced,0.03878144025802612
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,4,8,power_law_1.01,0.08317824363708495
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,4,8,power_law_1.01,0.14941951751708984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,4,8,power_law_1.01,0.150067195892334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,32,1,power_law_1.2,0.36572158813476563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,4,8,power_law_1.01,0.15074687957763672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,4,8,power_law_1.01,0.15217663764953612
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,4,8,power_law_1.01,0.1532172775268555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,4,8,power_law_1.01,0.15560447692871093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,4,8,power_law_1.01,0.1503116798400879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,4,8,power_law_1.01,0.1569651222229004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,4,8,power_law_1.01,0.1581811237335205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,4,8,power_law_1.01,0.16101375579833985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,4,8,power_law_1.01,0.1657164764404297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,4,8,power_law_1.01,0.16963840484619142
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,4,8,power_law_1.01,0.17570560455322265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,4,8,power_law_1.01,0.19717760086059571
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,4,8,power_law_1.01,0.20634880065917968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,4,8,power_law_1.01,0.22384767532348632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,4,8,power_law_1.01,0.30034303665161133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,4,8,power_law_1.01,0.37878528594970706
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,4,8,power_law_1.01,0.5113459014892578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,4,8,power_law_1.01,0.6121676635742188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,4,8,power_law_1.01,0.873785629272461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,4,8,balanced,0.04898816108703613
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,4,8,power_law_1.01,1.7587954711914062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,4,8,power_law_1.01,1.2827813720703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,4,8,balanced,0.061582078933715825
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,4,8,power_law_1.01,0.042421760559082033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,4,8,power_law_1.01,0.05222400188446045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,4,8,power_law_1.01,0.053917441368103024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,4,8,power_law_1.01,0.05337984085083007
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,4,8,power_law_1.01,0.05426688194274902
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,4,8,power_law_1.01,0.05343616008758545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,4,8,power_law_1.01,0.05388800144195557
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,4,8,power_law_1.01,0.05471615791320801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,4,8,power_law_1.01,0.05713024139404297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,4,8,power_law_1.01,0.05791999816894531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,4,8,power_law_1.01,0.059644160270690916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,4,8,power_law_1.01,0.06162047863006592
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,4,8,power_law_1.01,0.06427648067474365
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,4,8,power_law_1.01,0.06957695960998536
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,4,8,power_law_1.01,0.07485568046569824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,4,8,power_law_1.01,0.07911295890808105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,4,8,power_law_1.01,0.08144767761230469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,4,8,balanced,0.08410752296447754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,4,8,power_law_1.01,0.10218496322631836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,4,8,power_law_1.01,3.9574029541015627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,4,8,power_law_1.01,0.1269260787963867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,4,8,power_law_1.01,0.1712076759338379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,4,8,power_law_1.01,0.20060672760009765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,4,8,power_law_1.01,0.30846208572387696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,4,8,power_law_1.01,0.5116608047485351
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,4,8,balanced,0.10458111763000488
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,4,8,balanced,0.14436351776123046
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,4,8,power_law_1.01,0.020700159072875975
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,4,8,balanced,0.18599552154541016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,4,8,power_law_1.01,0.020293118953704833
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,4,8,power_law_1.01,0.020200960636138916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,4,8,power_law_1.01,0.02065920114517212
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,4,8,power_law_1.01,0.020280320644378663
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,4,8,power_law_1.01,0.02067456007003784
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,4,8,power_law_1.01,0.020661759376525878
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,4,8,power_law_1.01,0.020711679458618164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,4,8,power_law_1.01,0.020522239208221434
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,4,8,power_law_1.01,0.02108288049697876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,4,8,power_law_1.01,0.3615500640869141
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,4,8,power_law_1.01,0.02068351984024048
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,4,8,power_law_1.01,0.02192255973815918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,4,8,power_law_1.01,2.656344299316406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,4,8,power_law_1.01,1.1363340759277345
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,4,8,power_law_1.01,0.03493247985839844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,4,8,power_law_1.01,5.8075714111328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,4,8,power_law_1.01,0.7784333038330078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,4,8,power_law_1.01,0.07720191955566406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,4,8,power_law_1.01,1.5612608337402345
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,4,8,power_law_1.01,0.022782719135284422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,4,8,power_law_1.01,0.0297216010093689
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,4,8,power_law_1.01,0.023070719242095947
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,4,8,power_law_1.01,0.027809278964996336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,4,8,power_law_1.01,0.18124671936035158
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,4,8,power_law_1.01,0.04862080097198486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,4,8,power_law_1.2,0.08333951950073241
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,4,8,power_law_1.2,0.1497689628601074
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,4,8,power_law_1.01,0.058607358932495114
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,4,8,power_law_1.2,0.15073023796081544
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,4,8,power_law_1.2,0.15118080139160156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,4,8,power_law_1.2,0.15097984313964843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,4,8,power_law_1.2,0.152424955368042
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,4,8,power_law_1.2,0.15408384323120117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,4,8,power_law_1.2,0.15633152008056642
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,4,8,power_law_1.2,0.1569651222229004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,4,8,power_law_1.2,0.15827327728271484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,4,8,power_law_1.2,0.1615679931640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,4,8,power_law_1.2,0.16499967575073243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,4,8,power_law_1.2,0.168538875579834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,4,8,power_law_1.2,0.1767616081237793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,4,8,power_law_1.2,0.19959936141967774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,4,8,power_law_1.2,0.1986944007873535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,4,8,power_law_1.2,0.2354751968383789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,4,8,power_law_1.2,0.28961023330688473
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,4,8,power_law_1.2,0.38000640869140623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,4,8,power_law_1.2,0.4725465774536133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,4,8,power_law_1.2,0.8629299163818359
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,4,8,power_law_1.01,0.09430272102355956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,4,8,power_law_1.2,0.6154751968383789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,4,8,power_law_1.01,0.13238911628723143
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,4,8,power_law_1.2,1.3590412902832032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,4,8,power_law_1.2,2.418682861328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,4,8,power_law_1.2,0.04372352123260498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,4,8,power_law_1.2,0.054237442016601564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,4,8,power_law_1.2,0.055404801368713376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,4,8,power_law_1.2,0.0550387191772461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,4,8,power_law_1.2,0.05452032089233398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,4,8,power_law_1.2,0.0539251184463501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,4,8,power_law_1.2,0.05556992053985595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,4,8,power_law_1.2,0.054202880859375004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,4,8,power_law_1.2,0.054611201286315914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,4,8,power_law_1.2,0.05539711952209473
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,4,8,power_law_1.2,0.055776000022888184
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,4,8,power_law_1.2,0.06197247982025147
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,4,8,power_law_1.2,0.06287744045257568
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,4,8,power_law_1.2,0.06679296016693115
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,4,8,power_law_1.2,0.07251200199127197
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,4,8,power_law_1.2,0.0784281587600708
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,4,8,power_law_1.2,0.09084032058715821
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,4,8,power_law_1.2,2.0169471740722655
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,4,8,power_law_1.01,0.3197542381286621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,4,8,power_law_1.2,0.10994688034057618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,4,8,power_law_1.2,0.13157888412475588
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,4,8,power_law_1.2,0.17999616622924805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,4,8,power_law_1.2,0.209036808013916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,4,8,power_law_1.01,0.2647590446472168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,4,8,power_law_1.2,0.3045286369323731
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,4,8,power_law_1.01,0.48009601593017576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,4,8,power_law_1.2,0.37737728118896485
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,4,8,power_law_1.2,0.019833600521087645
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,4,8,power_law_1.2,0.02021631956100464
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,4,8,power_law_1.2,0.01983232021331787
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,4,8,power_law_1.2,0.02026623964309692
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,4,8,power_law_1.2,0.01989375948905945
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,4,8,power_law_1.2,0.020273919105529788
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,4,8,power_law_1.2,0.02059391975402832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,4,8,power_law_1.2,0.5776448059082031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,4,8,power_law_1.2,0.020268800258636473
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,4,8,power_law_1.2,0.020576000213623047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,4,8,power_law_1.2,0.020700159072875975
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,4,8,power_law_1.2,0.021011199951171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,4,8,power_law_1.2,0.02189440011978149
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,4,8,power_law_1.2,0.023946239948272705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,4,8,power_law_1.2,1.1421542358398438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,4,8,power_law_1.2,0.8257958221435547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,4,8,power_law_1.2,4.363008117675781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,4,8,power_law_1.2,1.5549900817871094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,4,8,power_law_1.2,5.482713623046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,4,8,power_law_1.2,0.02193536043167114
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,4,8,power_law_1.2,0.03192064046859741
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,4,8,power_law_1.01,0.7016038513183593
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,4,8,power_law_1.2,0.1606553649902344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,4,8,power_law_1.2,0.028076798915863038
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,4,8,power_law_1.2,0.03630208015441895
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,4,8,power_law_1.2,0.04462080001831055
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,4,8,power_law_1.2,0.07195519924163818
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,4,8,power_law_1.2,0.055015678405761714
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,8,4,balanced,0.03789695978164673
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,8,4,balanced,0.03958784103393555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,8,4,balanced,0.07184768199920655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,8,4,balanced,0.12363136291503907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,8,4,balanced,0.12473343849182128
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,8,4,balanced,0.1242956829071045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,8,4,balanced,0.12495231628417969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,8,4,balanced,0.1254310417175293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,8,4,balanced,0.12701312065124512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,8,4,balanced,0.12731776237487794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,8,4,balanced,0.1275635242462158
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,8,4,balanced,0.13047679901123047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,8,4,balanced,0.132488956451416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,8,4,balanced,0.13584383964538574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,8,4,balanced,0.1381708812713623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,8,4,balanced,0.14009344100952148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,8,4,balanced,0.14435327529907227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,8,4,balanced,0.16642303466796876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,8,4,balanced,0.15491071701049805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,8,4,balanced,0.2042304039001465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,8,4,balanced,0.2435916709899902
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,8,4,balanced,0.3260454559326172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,8,4,balanced,0.4013875198364258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,8,4,balanced,0.5885440063476562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,8,4,balanced,0.7610726165771484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,8,4,balanced,1.1261222076416015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,8,4,balanced,0.03665791988372803
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,8,4,balanced,0.03625600099563599
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,8,4,balanced,0.038677759170532226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,8,4,balanced,0.046489601135253904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,8,4,balanced,1.4981146240234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,8,4,balanced,0.04690176010131836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,8,4,balanced,0.048162560462951656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,8,4,balanced,0.04772863864898681
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,8,4,balanced,0.04726784229278565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,8,4,balanced,0.04790527820587158
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,8,4,balanced,0.04891520023345947
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,8,4,balanced,0.047722239494323734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,8,4,balanced,0.050114560127258304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,8,4,balanced,0.05019904136657714
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,8,4,balanced,0.050632958412170415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,8,4,balanced,0.053864960670471196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,8,4,balanced,0.055656957626342776
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,8,4,balanced,0.05956992149353028
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,8,4,balanced,0.07362175941467285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,8,4,balanced,0.06880640029907227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,8,4,balanced,0.08494720458984376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,8,4,balanced,0.09820544242858886
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,8,4,balanced,0.13609600067138672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,8,4,balanced,0.16040447235107422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,8,4,balanced,0.22916479110717775
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,8,4,balanced,0.3051519966125488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,8,4,balanced,0.43557247161865237
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,8,4,balanced,0.5631103897094727
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,4,8,power_law_1.2,0.10223232269287111
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,4,8,power_law_1.2,0.13378560066223144
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,4,8,power_law_1.2,0.2497177505493164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,8,4,balanced,0.019425280094146728
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,8,4,balanced,0.019819519519805907
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,8,4,balanced,0.019843839406967163
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,8,4,balanced,0.019454720020294188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,8,4,balanced,0.019874559640884398
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,8,4,balanced,0.020168960094451904
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,8,4,balanced,0.020241920948028562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,8,4,balanced,0.02020224094390869
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,8,4,balanced,0.020293118953704833
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,8,4,balanced,0.02062335968017578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,8,4,balanced,0.020696320533752442
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,8,4,balanced,0.020684800148010253
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,8,4,balanced,0.021018879413604735
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,8,4,balanced,0.021477119922637938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,4,8,power_law_1.2,0.3614553451538086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,8,4,balanced,0.01944831967353821
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,8,4,balanced,0.02005120038986206
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,4,8,power_law_1.2,0.5142668914794922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,4,8,power_law_1.2,0.6473382568359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,8,4,balanced,0.019811840057373048
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,8,4,balanced,0.025968639850616454
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,8,4,balanced,0.02639359951019287
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,8,4,balanced,0.03537280082702637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,8,4,power_law_1.01,0.056570878028869634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,8,4,power_law_1.01,0.06751359939575195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,8,4,power_law_1.01,0.12187392234802245
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,8,4,power_law_1.01,0.12428928375244142
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,8,4,power_law_1.01,0.12434047698974608
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,8,4,balanced,0.04220799922943115
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,8,4,power_law_1.01,0.12608511924743654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,8,4,power_law_1.01,0.1271615982055664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,8,4,power_law_1.01,0.12841856002807617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,8,4,power_law_1.01,0.13049471855163575
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,8,4,power_law_1.01,0.1308351993560791
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,8,4,power_law_1.01,0.13502464294433594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,8,4,power_law_1.01,0.13853440284729004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,8,4,power_law_1.01,0.14090496063232422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,8,4,power_law_1.01,0.1465664005279541
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,8,4,balanced,0.056040959358215334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,8,4,power_law_1.01,0.1598374366760254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,8,4,power_law_1.01,0.16606592178344726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,8,4,power_law_1.01,0.17428991317749024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,8,4,power_law_1.01,0.2176140785217285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,8,4,power_law_1.01,0.2700799942016602
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,8,4,balanced,0.06860288143157958
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,8,4,power_law_1.01,0.42020225524902344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,8,4,power_law_1.01,0.32703231811523437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,8,4,balanced,0.09232895851135255
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,8,4,power_law_1.01,1.450100555419922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,8,4,balanced,0.1154368019104004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,8,4,power_law_1.01,0.03585536003112793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,8,4,power_law_1.01,0.03786240100860595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,8,4,power_law_1.01,0.6208166503906251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,8,4,power_law_1.01,0.881195526123047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,8,4,power_law_1.01,0.04648575782775879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,8,4,power_law_1.01,0.04674816131591797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,8,4,power_law_1.01,0.04749311923980713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,8,4,power_law_1.01,0.04727168083190918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,8,4,power_law_1.01,0.047710719108581545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,8,4,power_law_1.01,0.048126721382141115
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,8,4,power_law_1.01,0.04894591808319092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,8,4,power_law_1.01,0.050481920242309575
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,8,4,power_law_1.01,0.05310080051422119
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,8,4,power_law_1.01,0.05466112136840821
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,8,4,power_law_1.01,1.198092803955078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,8,4,power_law_1.01,0.0634009599685669
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,8,4,power_law_1.01,0.06722303867340088
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,8,4,power_law_1.01,0.07171199798583985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,8,4,power_law_1.01,0.04812160015106201
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,8,4,power_law_1.01,0.07532032012939453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,8,4,power_law_1.01,0.10276608467102051
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,8,4,power_law_1.01,0.08687487602233887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,8,4,power_law_1.01,0.16723712921142578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,8,4,power_law_1.01,0.1345088005065918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,8,4,power_law_1.01,0.4501555252075195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,8,4,power_law_1.01,0.2409971237182617
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,8,4,balanced,0.16141696929931643
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,8,4,power_law_1.01,0.019472639560699462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,8,4,power_law_1.01,0.3310335922241211
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,8,4,power_law_1.01,0.01971583962440491
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,8,4,power_law_1.01,0.5679833602905273
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,8,4,power_law_1.01,0.019792640209198
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,8,4,power_law_1.01,0.01985152006149292
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,8,4,power_law_1.01,0.01987328052520752
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,8,4,power_law_1.01,0.020254719257354736
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,8,4,power_law_1.01,0.020200960636138916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,8,4,power_law_1.01,0.020258560180664062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,8,4,power_law_1.01,0.020285439491271973
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,8,4,power_law_1.01,0.02068351984024048
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,8,4,power_law_1.01,0.021515519618988038
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,8,4,balanced,0.20681472778320314
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,8,4,power_law_1.01,0.02396159887313843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,8,4,power_law_1.01,2.955146179199219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,8,4,power_law_1.01,0.8723442840576172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,8,4,power_law_1.01,2.338995208740234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,8,4,power_law_1.01,1.2017740631103515
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,8,4,power_law_1.01,0.019866880178451538
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,8,4,power_law_1.01,0.021960959434509278
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,8,4,power_law_1.01,0.02807039976119995
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,8,4,power_law_1.01,0.02886528015136719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,8,4,power_law_1.01,0.10660736083984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,8,4,power_law_1.01,0.1423961639404297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,8,4,power_law_1.01,0.03340672016143799
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,8,4,power_law_1.01,0.04157055854797363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,8,4,power_law_1.2,0.06239359855651856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,8,4,power_law_1.2,0.0676800012588501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,8,4,power_law_1.2,0.12197631835937499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,8,4,power_law_1.2,0.12420479774475097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,8,4,power_law_1.2,0.12428671836853027
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,8,4,power_law_1.2,0.12544639587402345
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,8,4,power_law_1.01,0.04650239944458008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,8,4,power_law_1.2,0.12694399833679199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,8,4,power_law_1.2,0.12959360122680663
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,8,4,power_law_1.2,0.1299507236480713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,8,4,power_law_1.2,0.13051136016845705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,8,4,power_law_1.2,0.13516032218933105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,8,4,power_law_1.2,0.1382271957397461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,8,4,power_law_1.2,0.13943039894104003
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,8,4,power_law_1.2,0.14852352142333985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,8,4,power_law_1.2,0.15838463783264162
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,8,4,power_law_1.2,0.17086591720581054
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,8,4,power_law_1.2,0.17465087890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,8,4,power_law_1.2,0.22288768768310546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,8,4,power_law_1.2,0.263317756652832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,8,4,power_law_1.2,0.33753601074218753
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,8,4,power_law_1.2,0.6574809265136718
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,8,4,power_law_1.01,0.0610752010345459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,8,4,power_law_1.2,0.4706278228759766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,8,4,power_law_1.01,0.07297152042388916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,8,4,power_law_1.2,0.7561293029785157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,8,4,power_law_1.2,1.1216793823242188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,8,4,power_law_1.2,0.03502847909927368
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,8,4,power_law_1.2,0.038270719051361084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,8,4,power_law_1.2,0.046082558631896975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,8,4,power_law_1.2,0.04657279968261719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,8,4,power_law_1.2,0.04673791885375976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,8,4,power_law_1.2,0.047298560142517085
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,8,4,power_law_1.01,0.19410303115844726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,8,4,power_law_1.2,0.04742656230926513
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,8,4,power_law_1.2,0.04773119926452637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,8,4,power_law_1.2,0.04887296199798584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,8,4,power_law_1.2,0.04942975997924805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,8,4,power_law_1.2,0.05230591773986817
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,8,4,power_law_1.2,0.056543998718261726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,8,4,power_law_1.2,0.05741312026977539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,8,4,power_law_1.2,0.063439359664917
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,8,4,power_law_1.2,0.06818431854248047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,8,4,power_law_1.2,0.07307007789611816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,8,4,power_law_1.2,0.07572991847991943
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,8,4,power_law_1.2,1.4647398376464844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,8,4,power_law_1.2,0.08697088241577149
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,8,4,power_law_1.2,0.10491519927978517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,8,4,power_law_1.2,0.13451135635375977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,8,4,power_law_1.2,0.16162944793701173
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,8,4,power_law_1.01,0.31021823883056643
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,8,4,power_law_1.2,0.41314945220947263
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,8,4,power_law_1.2,0.25675392150878906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,8,4,power_law_1.01,0.24362112045288087
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,8,4,power_law_1.2,0.019403519630432128
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,8,4,power_law_1.2,0.019639040231704714
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,8,4,power_law_1.2,0.3249919891357422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,8,4,power_law_1.2,0.019397120475769043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,8,4,power_law_1.2,0.019855359792709353
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,8,4,power_law_1.2,0.01986176013946533
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,8,4,power_law_1.2,0.02019711971282959
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,8,4,power_law_1.2,0.020273919105529788
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,8,4,power_law_1.2,0.020212481021881105
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,8,4,power_law_1.2,0.020625920295715333
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,8,4,power_law_1.2,0.02062848091125488
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,8,4,power_law_1.2,0.02186624050140381
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,8,4,power_law_1.2,0.02209280014038086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,8,4,power_law_1.2,0.02350719928741455
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,8,4,power_law_1.2,0.02634495973587036
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,8,4,power_law_1.2,0.028053760528564453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,8,4,power_law_1.2,0.6242623901367187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,8,4,power_law_1.2,2.231786193847656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,8,4,power_law_1.2,0.04650368213653565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,8,4,power_law_1.2,0.8422310638427735
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,8,4,power_law_1.2,0.07481855869293212
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,8,4,power_law_1.2,3.1708441162109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,8,4,power_law_1.2,0.019749120473861695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,8,4,power_law_1.01,0.43223297119140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,8,4,power_law_1.2,0.035191040039062496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,8,4,power_law_1.2,1.1752537536621095
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,8,4,power_law_1.2,0.04246399879455566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,16,2,balanced,0.03337088108062744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,16,2,balanced,0.055118079185485835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,16,2,balanced,0.06669439792633056
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,16,2,balanced,0.11369983673095703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,16,2,balanced,0.1141260814666748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,16,2,balanced,0.11451647758483888
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,16,2,balanced,0.11654911994934083
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,16,2,balanced,0.11717503547668458
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,16,2,balanced,0.11895296096801758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,16,2,balanced,0.11966976165771484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,16,2,balanced,0.12082431793212892
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,16,2,balanced,0.12337280273437501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,16,2,balanced,0.12518272399902344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,8,4,power_law_1.2,0.06534399986267089
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,16,2,balanced,0.1286911964416504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,16,2,balanced,0.13999744415283205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,16,2,balanced,0.14428288459777833
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,16,2,balanced,0.1352729606628418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,16,2,balanced,0.15295231819152832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,16,2,balanced,0.16871423721313478
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,16,2,balanced,0.2008742332458496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,16,2,balanced,0.2674163246154785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,16,2,balanced,0.36603904724121095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,16,2,balanced,0.4543078231811523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,16,2,balanced,0.666398696899414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,16,2,balanced,0.8738508605957032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,16,2,balanced,1.2972454833984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,16,2,balanced,0.03296256065368652
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,16,2,balanced,0.03458559989929199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,16,2,balanced,0.037285120487213136
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,16,2,balanced,0.04325888156890869
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,16,2,balanced,1.7099136352539062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,16,2,balanced,0.04402304172515869
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,16,2,balanced,0.04444416046142578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,16,2,balanced,0.0459174394607544
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,16,2,balanced,0.047791361808776855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,16,2,balanced,0.04606592178344727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,16,2,balanced,0.047667198181152344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,16,2,balanced,0.04897408008575439
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,16,2,balanced,0.05100160121917725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,16,2,balanced,0.04449408054351807
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,16,2,balanced,0.05505280017852783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,16,2,balanced,0.05934720039367676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,16,2,balanced,0.06943103790283203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,8,4,power_law_1.2,0.10980607986450194
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,16,2,balanced,0.0636032009124756
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,16,2,balanced,0.07627647876739502
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,16,2,balanced,0.08451968193054199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,16,2,balanced,0.10527359962463378
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,16,2,balanced,0.12315263748168945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,16,2,balanced,0.18250495910644532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,16,2,balanced,0.2299648094177246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,16,2,balanced,0.32762622833251953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,16,2,balanced,0.4325721740722656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,16,2,balanced,0.6349772644042969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,16,2,balanced,0.8137420654296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,8,4,power_law_1.2,0.12290944099426268
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,8,4,power_law_1.2,0.2184409523010254
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,8,4,power_law_1.2,0.2392192077636719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,16,2,balanced,0.019031039476394652
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,16,2,balanced,0.021964800357818604
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,16,2,balanced,0.019636479616165162
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,16,2,balanced,0.022366719245910643
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,16,2,balanced,0.022215681076049803
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,16,2,balanced,0.022302720546722412
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,8,4,power_law_1.2,0.355175666809082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,16,2,balanced,0.02230015993118286
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,16,2,balanced,0.02239743947982788
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,16,2,balanced,0.022330880165100098
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,16,2,balanced,0.022750720977783204
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,16,2,balanced,0.0227455997467041
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,16,2,balanced,0.024367361068725585
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,16,2,balanced,0.021610240936279296
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,16,2,balanced,0.02183680057525635
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,16,2,balanced,0.02153599977493286
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,8,4,power_law_1.2,0.4871308898925781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,16,2,balanced,0.02231168031692505
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,16,2,balanced,0.02232320070266724
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,16,2,balanced,0.02861311912536621
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,16,2,balanced,0.030252799987792966
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,16,2,balanced,0.04368256092071533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,16,2,power_law_1.01,0.05228928089141845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,16,2,power_law_1.01,0.05831039905548095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,16,2,power_law_1.01,0.06245503902435303
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,16,2,power_law_1.01,0.11374464035034179
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,16,2,power_law_1.01,0.11395199775695801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,16,2,power_law_1.01,0.11528191566467286
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,16,2,power_law_1.01,0.11666560173034668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,16,2,power_law_1.01,0.11862527847290039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,16,2,power_law_1.01,0.11921536445617675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,16,2,power_law_1.01,0.11975040435791015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,16,2,power_law_1.01,0.12151424407958984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,16,2,power_law_1.01,0.12380031585693359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,16,2,power_law_1.01,0.12696831703186034
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,16,2,power_law_1.01,0.13560704231262205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,16,2,power_law_1.01,0.1434496021270752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,16,2,power_law_1.01,0.14910592079162596
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,16,2,balanced,0.05164927959442138
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,16,2,power_law_1.01,0.1522214412689209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,16,2,power_law_1.01,0.16778623580932617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,16,2,power_law_1.01,0.20999423980712892
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,16,2,power_law_1.01,0.2707814407348633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,16,2,power_law_1.01,0.3375193786621094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,16,2,balanced,0.06673920154571533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,16,2,power_law_1.01,0.5641433715820312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,16,2,power_law_1.01,0.48980865478515623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,16,2,power_law_1.01,1.0790361785888671
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,16,2,balanced,0.0815436840057373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,16,2,power_law_1.01,0.03273727893829346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,16,2,power_law_1.01,0.035484158992767335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,16,2,power_law_1.01,0.039221758842468264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,16,2,power_law_1.01,0.04333951950073242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,16,2,power_law_1.01,0.04365056037902832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,16,2,balanced,0.11209216117858886
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,16,2,power_law_1.01,0.044375038146972655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,16,2,power_law_1.01,0.04486271858215332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,16,2,power_law_1.01,0.04525568008422852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,16,2,power_law_1.01,0.04604159832000733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,16,2,power_law_1.01,0.04673408031463623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,16,2,power_law_1.01,0.840728302001953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,16,2,power_law_1.01,0.04867839813232422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,16,2,power_law_1.01,0.053038082122802734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,16,2,power_law_1.01,0.05189760208129883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,16,2,power_law_1.01,0.060912637710571295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,16,2,power_law_1.01,0.06779520034790039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,16,2,power_law_1.01,0.07341184139251708
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,16,2,power_law_1.01,0.07132287979125976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,16,2,power_law_1.01,0.08956928253173828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,16,2,power_law_1.01,0.07730815887451173
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,16,2,balanced,0.13837823867797852
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,16,2,balanced,0.20040576934814452
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,16,2,power_law_1.01,0.11545472145080567
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,16,2,balanced,0.26034816741943356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,16,2,power_law_1.01,0.1485324764251709
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,16,2,power_law_1.01,0.19960575103759765
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,16,2,power_law_1.01,0.019866880178451538
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,16,2,power_law_1.01,0.01955072045326233
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,16,2,power_law_1.01,0.02069119930267334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,16,2,power_law_1.01,0.021512959003448486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,16,2,power_law_1.01,1.6365196228027343
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,16,2,power_law_1.01,0.021831679344177245
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,16,2,power_law_1.01,0.0217740797996521
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,16,2,power_law_1.01,0.021916160583496096
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,16,2,power_law_1.01,0.022305281162261964
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,16,2,power_law_1.01,0.022266879081726074
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,16,2,power_law_1.01,0.021923840045928955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,16,2,power_law_1.01,0.4546163177490234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,16,2,power_law_1.01,0.022361600399017335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,16,2,power_law_1.01,0.23575807571411134
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,16,2,power_law_1.01,0.3520729446411133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,16,2,power_law_1.01,2.2124954223632813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,16,2,power_law_1.01,0.7097395324707032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,16,2,power_law_1.01,0.9411199951171876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,16,2,power_law_1.01,0.049236478805541994
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,16,2,power_law_1.01,0.02293503999710083
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,16,2,power_law_1.01,0.025128960609436035
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,16,2,power_law_1.01,0.026357760429382326
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,16,2,power_law_1.01,0.023054080009460447
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,16,2,power_law_1.01,0.027989759445190433
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,16,2,power_law_1.01,0.03093247890472412
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,16,2,power_law_1.01,0.041602559089660644
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,16,2,power_law_1.01,0.0627622413635254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,16,2,power_law_1.2,0.050704641342163084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,16,2,power_law_1.2,0.05937024116516113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,16,2,power_law_1.2,0.06146048069000244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,16,2,power_law_1.2,0.11321472167968749
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,16,2,power_law_1.2,0.11339520454406737
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,16,2,power_law_1.2,0.11450240135192871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,16,2,power_law_1.2,0.11667967796325684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,16,2,power_law_1.2,0.11807231903076172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,16,2,power_law_1.2,0.11973247528076172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,16,2,power_law_1.2,0.11991552352905273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,16,2,power_law_1.2,0.12175744056701661
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,16,2,power_law_1.2,0.12386303901672364
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,16,2,power_law_1.2,0.12671232223510742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,16,2,power_law_1.2,0.1361062431335449
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,16,2,power_law_1.2,0.14225407600402834
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,16,2,power_law_1.01,0.12356608390808106
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,16,2,power_law_1.2,0.1498406410217285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,16,2,power_law_1.2,0.15438207626342773
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,16,2,power_law_1.01,0.09706368446350097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,16,2,power_law_1.2,0.1715852737426758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,16,2,power_law_1.2,0.20774015426635742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,16,2,power_law_1.01,0.07598591804504394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,16,2,power_law_1.2,0.3366463851928711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,16,2,power_law_1.2,0.27999359130859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,16,2,power_law_1.01,0.18345855712890624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,16,2,power_law_1.2,0.5048473739624024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,16,2,power_law_1.2,0.5587251281738281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,16,2,power_law_1.2,0.0329804801940918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,16,2,power_law_1.2,0.036078081130981446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,16,2,power_law_1.2,0.03783423900604248
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,16,2,power_law_1.01,0.23532543182373047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,16,2,power_law_1.2,0.04394752025604248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,16,2,power_law_1.2,0.044359679222106936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,16,2,power_law_1.2,0.045099520683288576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,16,2,power_law_1.2,0.0455244779586792
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,16,2,power_law_1.2,0.04605696201324463
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,16,2,power_law_1.2,0.046982398033142095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,16,2,power_law_1.2,0.04765952110290527
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,16,2,power_law_1.2,0.0447705602645874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,16,2,power_law_1.2,0.05016704082489014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,16,2,power_law_1.2,0.0535155200958252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,16,2,power_law_1.2,0.061080322265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,16,2,power_law_1.2,0.06947328090667725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,16,2,power_law_1.2,0.07481088161468505
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,16,2,power_law_1.2,1.1829798126220703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,16,2,power_law_1.2,0.07115007877349853
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,16,2,power_law_1.2,0.08987263679504395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,16,2,power_law_1.2,0.07745664119720459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,16,2,power_law_1.2,0.11777152061462401
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,16,2,power_law_1.2,0.8779302215576171
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,16,2,power_law_1.2,0.14584320068359374
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,16,2,power_law_1.01,0.35060096740722657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,16,2,power_law_1.2,0.20731647491455077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,16,2,power_law_1.2,0.2600831985473633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,16,2,power_law_1.2,0.019813120365142822
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,16,2,power_law_1.2,0.019943679571151736
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,16,2,power_law_1.2,0.02024960041046143
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,16,2,power_law_1.2,0.021880319118499754
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,16,2,power_law_1.2,0.021500160694122313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,16,2,power_law_1.2,0.021913599967956544
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,16,2,power_law_1.2,0.02221695899963379
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,16,2,power_law_1.2,0.021886720657348632
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,16,2,power_law_1.2,0.022328319549560546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,16,2,power_law_1.2,0.022307839393615723
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,16,2,power_law_1.2,0.022717440128326417
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,16,2,power_law_1.2,0.023121919631958008
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,16,2,power_law_1.2,0.023535358905792236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,16,2,power_law_1.2,0.3838399887084961
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,16,2,power_law_1.2,0.024556798934936525
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,16,2,power_law_1.2,0.02753024101257324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,16,2,power_law_1.2,2.0505830383300783
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,16,2,power_law_1.01,0.3906598281860352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,16,2,power_law_1.2,1.714239959716797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,16,2,power_law_1.2,0.4877798461914063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,16,2,power_law_1.2,0.669401626586914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,16,2,power_law_1.2,0.027233281135559083
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,16,2,power_law_1.2,0.0339084792137146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,16,2,power_law_1.2,0.8916748809814454
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,16,2,power_law_1.2,0.04079360008239746
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,16,2,power_law_1.2,0.048275198936462406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,16,2,power_law_1.2,0.06212351799011231
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,32,1,balanced,0.03492095947265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,32,1,balanced,0.059474558830261225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,32,1,balanced,0.07038335800170899
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,32,1,balanced,0.07124224185943603
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,32,1,balanced,0.11413311958312988
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,32,1,balanced,0.11505087852478027
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,32,1,balanced,0.1165817642211914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,32,1,balanced,0.11728768348693848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,32,1,balanced,0.11830719947814941
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,32,1,balanced,0.11869888305664063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,16,2,power_law_1.2,0.06729983806610107
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,32,1,balanced,0.11967295646667479
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,32,1,balanced,0.12183679580688475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,32,1,balanced,0.1230412769317627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,32,1,balanced,0.127257604598999
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,32,1,balanced,0.1354854393005371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,32,1,balanced,0.13890624046325684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,32,1,balanced,0.1378694438934326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,32,1,balanced,0.15158592224121095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,32,1,balanced,0.1652364730834961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,32,1,balanced,0.19818815231323242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,32,1,balanced,0.25584320068359373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,32,1,balanced,0.3542451095581055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,32,1,balanced,0.44461822509765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,32,1,balanced,0.6412569427490235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,32,1,balanced,0.8463136291503905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,32,1,balanced,1.2271635437011719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,32,1,balanced,0.03971456050872803
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,32,1,balanced,0.041865601539611816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,32,1,balanced,0.045885438919067385
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,32,1,balanced,0.04687039852142334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,32,1,balanced,0.05465536117553711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,32,1,balanced,1.6593023681640624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,32,1,balanced,0.05570432186126709
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,32,1,balanced,0.0566374397277832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,32,1,balanced,0.056304001808166505
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,32,1,balanced,0.05624703884124756
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,32,1,balanced,0.05871744155883789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,32,1,balanced,0.05754879951477051
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,32,1,balanced,0.05944255828857422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,32,1,balanced,0.060798721313476564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,32,1,balanced,0.06573952198028564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,32,1,balanced,0.07134975910186767
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,32,1,balanced,0.07212800025939942
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,32,1,balanced,0.0851846408843994
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,32,1,balanced,0.09160256385803223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,32,1,balanced,0.11155455589294434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,32,1,balanced,0.06653120040893554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,32,1,balanced,0.13283583641052246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,32,1,balanced,0.1948761558532715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,32,1,balanced,0.2390444755554199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,32,1,balanced,0.345715217590332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,32,1,balanced,0.44879104614257814
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,16,2,power_law_1.2,0.10037887573242188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,32,1,balanced,0.6584429168701172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,32,1,balanced,0.843567352294922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,16,2,power_law_1.2,0.12483327865600587
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,16,2,power_law_1.2,0.1889958381652832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,16,2,power_law_1.2,0.17435647964477538
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,32,1,balanced,0.01931007981300354
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,32,1,balanced,0.02340224027633667
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,32,1,balanced,0.023043839931488036
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,32,1,balanced,0.023438720703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,32,1,balanced,0.023437440395355225
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,32,1,balanced,0.02140223979949951
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,32,1,balanced,0.023647999763488768
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,32,1,balanced,0.0234879994392395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,32,1,balanced,0.023639678955078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,32,1,balanced,0.025491199493408202
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,32,1,balanced,0.02550528049468994
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,32,1,balanced,0.022649600505828857
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,32,1,balanced,0.027158401012420653
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,32,1,balanced,0.022642560005187988
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,32,1,balanced,0.0257587194442749
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,16,2,power_law_1.2,0.383006706237793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,32,1,balanced,0.022922239303588866
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,32,1,balanced,0.023242878913879394
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,16,2,power_law_1.2,0.3311411285400391
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,32,1,balanced,0.03860543966293335
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,32,1,balanced,0.04290175914764404
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,32,1,balanced,0.057332482337951664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,32,1,power_law_1.01,0.048877439498901366
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,32,1,power_law_1.01,0.06021440029144287
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,32,1,power_law_1.01,0.06258560180664062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,32,1,power_law_1.01,0.07023295879364014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,32,1,power_law_1.01,0.11419327735900879
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,32,1,balanced,0.0661894416809082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,32,1,power_law_1.01,0.11528063774108888
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,32,1,power_law_1.01,0.11644288063049317
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,32,1,power_law_1.01,0.11717311859130859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,32,1,power_law_1.01,0.11844415664672851
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,32,1,power_law_1.01,0.11931136131286621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,32,1,power_law_1.01,0.12472319602966309
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,32,1,balanced,0.08812928199768066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,32,1,power_law_1.01,0.12586496353149415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,32,1,power_law_1.01,0.12931391716003418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,32,1,power_law_1.01,0.13266495704650877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,32,1,power_law_1.01,0.13595840454101563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,32,1,power_law_1.01,0.14209407806396485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,32,1,power_law_1.01,0.1453113555908203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,32,1,power_law_1.01,0.1605708885192871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,32,1,power_law_1.01,0.17788543701171874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,32,1,power_law_1.01,0.22707904815673827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,32,1,power_law_1.01,0.2780243110656738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,32,1,power_law_1.01,0.3810419082641602
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,32,1,balanced,0.10912704467773438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,32,1,power_law_1.01,0.47574462890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,32,1,balanced,0.1531180763244629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,32,1,power_law_1.01,0.03810816049575806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,32,1,power_law_1.01,0.04147007942199707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,32,1,power_law_1.01,0.043171839714050295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,32,1,power_law_1.01,0.04550271987915039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,32,1,power_law_1.01,0.054365439414978026
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,32,1,power_law_1.01,0.05422976016998291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,32,1,power_law_1.01,0.0556876802444458
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,32,1,power_law_1.01,0.6705868530273438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,32,1,power_law_1.01,0.0565990400314331
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,32,1,power_law_1.01,0.05744319915771484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,32,1,power_law_1.01,0.05833536148071289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,32,1,power_law_1.01,0.06024831771850586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,32,1,balanced,0.19241535186767578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,32,1,power_law_1.01,0.06020736217498779
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,32,1,power_law_1.01,0.06234432220458984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,32,1,power_law_1.01,0.0643174409866333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,32,1,power_law_1.01,0.06948224067687989
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,32,1,power_law_1.01,0.07199679851531983
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,32,1,power_law_1.01,0.07419263839721679
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,32,1,power_law_1.01,0.08529664039611816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,32,1,power_law_1.01,0.09722751617431641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,32,1,power_law_1.01,0.1161292839050293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,32,1,power_law_1.01,0.14191807746887206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,32,1,power_law_1.01,0.8629644775390626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,32,1,power_law_1.01,0.20027839660644534
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,32,1,power_law_1.01,0.24580223083496094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,32,1,balanced,0.2826694488525391
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,32,1,power_law_1.01,0.019754879474639893
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,32,1,power_law_1.01,0.02057408094406128
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,32,1,balanced,0.3673766326904297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,32,1,power_law_1.01,0.02279103994369507
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,32,1,power_law_1.01,0.35202430725097655
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,32,1,power_law_1.01,0.023473920822143553
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,32,1,power_law_1.01,0.02304447889328003
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,32,1,power_law_1.01,0.023072640895843505
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,32,1,power_law_1.01,0.023452160358428956
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,32,1,power_law_1.01,0.023320319652557372
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,32,1,power_law_1.01,0.023445119857788084
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,32,1,power_law_1.01,0.02098112106323242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,32,1,power_law_1.01,0.02401535987854004
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,32,1,power_law_1.01,0.0266483211517334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,32,1,power_law_1.01,0.02808703899383545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,32,1,power_law_1.01,1.2567091369628907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,32,1,power_law_1.01,0.4571763229370117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,32,1,power_law_1.01,0.6663123321533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,32,1,power_law_1.01,1.6485568237304686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,32,1,power_law_1.01,0.8633554840087891
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,32,1,power_law_1.01,0.08095295906066895
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,32,1,power_law_1.01,0.025495679378509523
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,32,1,power_law_1.01,0.030897920131683347
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,32,1,power_law_1.01,0.03210304021835327
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,32,1,power_law_1.01,0.03584831953048706
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,32,1,power_law_1.01,0.044604158401489256
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,32,1,power_law_1.01,0.056241922378540046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,32,1,power_law_1.2,0.05244351863861084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,32,1,power_law_1.2,0.06025087833404541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,32,1,power_law_1.2,0.061418881416320806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,32,1,power_law_1.2,0.07018239974975586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,32,1,power_law_1.2,0.11360128402709961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,32,1,power_law_1.2,0.1151091194152832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,32,1,power_law_1.01,0.06876992225646973
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,32,1,power_law_1.2,0.11636416435241699
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,32,1,power_law_1.2,0.1173740768432617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,32,1,power_law_1.2,0.11819519996643066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,32,1,power_law_1.2,0.11938816070556642
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,32,1,power_law_1.2,0.124453763961792
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,32,1,power_law_1.2,0.1267296028137207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,32,1,power_law_1.2,0.12865535736083983
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,32,1,power_law_1.2,0.13350208282470705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,32,1,power_law_1.2,0.13672191619873048
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,32,1,power_law_1.2,0.14250432014465333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,32,1,power_law_1.2,0.14547776222229003
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,32,1,power_law_1.2,0.16078399658203124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,32,1,power_law_1.2,0.17795391082763673
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,32,1,power_law_1.2,0.23176639556884765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,32,1,power_law_1.2,0.28329727172851565
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,32,1,power_law_1.01,0.1095635223388672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,32,1,power_law_1.2,0.38146625518798827
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,32,1,power_law_1.01,0.13812992095947266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,32,1,power_law_1.01,0.19680383682250976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,32,1,power_law_1.2,0.4778707122802734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,32,1,power_law_1.2,0.037159678936004636
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,32,1,power_law_1.2,0.04114496231079102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,32,1,power_law_1.2,0.041879677772521974
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,32,1,power_law_1.2,0.04593023777008057
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,32,1,power_law_1.2,0.053372159004211425
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,32,1,power_law_1.2,0.054595198631286625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,32,1,power_law_1.2,0.0558022403717041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,32,1,power_law_1.2,0.05586048126220703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,32,1,power_law_1.2,0.056822400093078616
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,32,1,power_law_1.2,0.05652544021606445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,32,1,power_law_1.2,0.058405761718750004
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,32,1,power_law_1.01,0.2565088081359863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,32,1,power_law_1.2,0.05873983860015869
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,32,1,power_law_1.2,0.06100160121917725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,32,1,power_law_1.2,0.06514368057250977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,32,1,power_law_1.2,0.06855487823486328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,32,1,power_law_1.2,0.07310719966888428
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,32,1,power_law_1.2,0.07541567802429198
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,32,1,power_law_1.2,0.08541440010070801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,32,1,power_law_1.2,0.09700608253479004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,32,1,power_law_1.2,0.6711820983886719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,32,1,power_law_1.2,0.11650752067565917
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,32,1,power_law_1.01,0.37271167755126955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,32,1,power_law_1.2,0.14087295532226562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,32,1,power_law_1.2,0.20069503784179688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,32,1,power_law_1.2,0.8653638458251953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,32,1,power_law_1.01,0.4900147247314453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,32,1,power_law_1.2,0.019348479509353637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,32,1,power_law_1.2,0.2453363227844238
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,32,1,power_law_1.2,0.02054271936416626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,32,1,power_law_1.2,0.022573440074920653
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,32,1,power_law_1.2,0.022768640518188478
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,32,1,power_law_1.2,0.022884480953216553
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,32,1,power_law_1.2,0.023034880161285402
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,32,1,power_law_1.2,0.02303999900817871
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,32,1,power_law_1.2,0.023406078815460203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,32,1,power_law_1.2,0.023450241088867188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,32,1,power_law_1.2,0.02384704113006592
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,32,1,power_law_1.2,0.3527436828613281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,32,1,power_law_1.2,0.4568582534790039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,32,1,power_law_1.2,0.02863424062728882
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,32,1,power_law_1.2,1.2570387268066407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,32,1,power_law_1.2,0.0304915189743042
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,32,1,power_law_1.2,0.02132352113723755
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,32,1,power_law_1.2,0.044961280822753906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,32,1,power_law_1.2,0.025173759460449217
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,32,1,power_law_1.2,0.02632960081100464
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,32,1,power_law_1.2,0.6663231658935547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,32,1,power_law_1.2,1.644757080078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,32,1,power_law_1.2,0.11308863639831543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,32,1,power_law_1.2,0.03276864051818847
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,32,1,power_law_1.2,0.03669183969497681
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,32,1,power_law_1.2,0.8705286407470704
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,32,1,power_law_1.2,0.05734528064727783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,32,balanced,0.06900735855102538
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,32,balanced,0.06907392024993897
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,32,balanced,0.06876031875610351
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,32,balanced,0.07856383800506592
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,32,balanced,0.09599360466003418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,32,balanced,0.14419455528259278
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,32,1,power_law_1.2,0.07081471920013428
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,32,balanced,0.1868185615539551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,32,balanced,0.20027135848999023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,32,balanced,0.19893503189086914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,32,balanced,0.20550399780273437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,32,balanced,0.2056575965881348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,32,balanced,0.18980480194091798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,32,balanced,0.20883840560913086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,32,balanced,0.20394496917724608
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,32,balanced,0.2116339111328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,32,balanced,0.22670080184936525
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,32,balanced,0.2194483184814453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,32,balanced,0.23326080322265624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,32,balanced,0.24726400375366211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,32,balanced,0.2716339111328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,32,balanced,0.2710758399963379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,32,balanced,0.4296793746948242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,32,balanced,0.4326976013183594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,32,balanced,0.5817740631103516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,32,balanced,0.7124237060546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,32,balanced,1.0325926208496095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,32,balanced,1.3186329650878907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,32,balanced,0.038604800701141355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,32,balanced,0.03871871948242188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,32,balanced,0.03852159976959228
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,32,balanced,0.04492159843444824
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,32,1,power_law_1.2,0.08410880088806152
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,32,balanced,0.05980800151824951
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,32,balanced,0.08596735954284668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,32,balanced,0.09318016052246095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,32,balanced,0.09392127990722657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,32,balanced,0.09541119575500488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,32,balanced,0.09495424270629883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,32,balanced,0.0974937629699707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,32,balanced,0.09831935882568359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,32,balanced,0.09913472175598145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,32,balanced,0.10190848350524902
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,32,balanced,0.10569472312927246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,32,balanced,0.10960895538330077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,32,balanced,0.10233599662780761
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,32,balanced,0.1104588794708252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,32,balanced,0.1186508846282959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,32,balanced,0.129169921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,32,balanced,0.14230655670166015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,32,balanced,0.17199487686157228
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,32,balanced,0.19402496337890623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,32,balanced,0.2763776016235352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,32,balanced,0.32382080078125003
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,32,balanced,0.45188735961914067
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,32,balanced,0.5894976043701172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,32,1,power_law_1.2,0.14319295883178712
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,32,1,power_law_1.2,0.20558271408081055
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,32,1,power_law_1.2,0.26667200088500975
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,1,32,balanced,0.02143615961074829
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,1,32,balanced,0.054364161491394045
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,1,32,balanced,0.02157952070236206
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,1,32,balanced,0.054049282073974615
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,1,32,balanced,0.054584321975708006
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,1,32,balanced,0.05447679996490479
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,1,32,balanced,0.054968318939208984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,1,32,balanced,0.05457151889801025
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,1,32,balanced,0.0548415994644165
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,1,32,balanced,0.05474944114685058
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,1,32,balanced,0.055448322296142584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,1,32,balanced,0.021864960193634035
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,1,32,balanced,0.055153918266296384
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,1,32,balanced,0.05547391891479492
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,1,32,balanced,0.05549824237823486
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,1,32,balanced,0.05558656215667725
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,1,32,balanced,0.05775743961334229
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,1,32,balanced,0.05820543766021728
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,1,32,balanced,0.05974400043487549
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,1,32,balanced,0.06164991855621338
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,1,32,balanced,0.02355072021484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,32,1,power_law_1.2,0.3864332962036133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,32,power_law_1.01,0.1439948844909668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,32,power_law_1.01,0.14378879547119142
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,32,power_law_1.01,0.14386176109313964
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,32,power_law_1.01,0.14555264472961427
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,32,power_law_1.01,0.17691520690917967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,32,power_law_1.01,0.1721651268005371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,32,power_law_1.01,0.1861631965637207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,32,power_law_1.01,0.19796352386474608
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,1,32,balanced,0.033788158893585205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,32,power_law_1.01,0.1974847984313965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,32,power_law_1.01,0.19350784301757812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,32,power_law_1.01,0.20265472412109373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,32,power_law_1.01,0.19355648040771484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,32,power_law_1.01,0.20736255645751953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,32,power_law_1.01,0.2092697525024414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,32,power_law_1.01,0.21976959228515627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,32,power_law_1.01,0.21601535797119142
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,32,power_law_1.01,0.2396518325805664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,32,power_law_1.01,0.3036172866821289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,32,power_law_1.01,0.33711231231689454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,32,power_law_1.01,0.4359539031982422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,32,power_law_1.01,0.5090662384033203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,32,power_law_1.01,0.7464435577392579
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,32,power_law_1.01,0.962471694946289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,32,power_law_1.01,1.3293580627441406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,32,power_law_1.01,1.680011444091797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,32,power_law_1.01,2.7291879272460937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,32,power_law_1.01,0.0851353645324707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,32,power_law_1.01,0.08430463790893554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,32,power_law_1.01,0.08448896408081055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,32,power_law_1.01,0.08609151840209961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,32,power_law_1.01,3.3533453369140624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,32,power_law_1.01,0.09104000091552734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,32,power_law_1.01,0.08431615829467773
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,1,32,balanced,0.05434879779815673
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,32,power_law_1.01,0.08824576377868652
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,32,power_law_1.01,0.09359744071960449
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,32,power_law_1.01,0.09179391860961914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,32,power_law_1.01,0.09437439918518067
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,32,power_law_1.01,0.09728639602661134
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,32,power_law_1.01,0.09968128204345703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,32,power_law_1.01,0.10667776107788085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,32,power_law_1.01,0.10470784187316895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,32,power_law_1.01,0.10262528419494629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,32,power_law_1.01,0.11012479782104492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,32,power_law_1.01,0.09708928108215333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,32,power_law_1.01,0.1358348846435547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,32,power_law_1.01,0.1444057559967041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,32,power_law_1.01,0.18940671920776367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,32,power_law_1.01,0.22076927185058595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,32,power_law_1.01,0.29158912658691405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,32,power_law_1.01,0.3707571029663086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,32,power_law_1.01,0.5376422500610352
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,32,1,power_law_1.2,0.5126496124267578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,32,power_law_1.01,0.6559257507324219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,1,32,balanced,0.08765952110290527
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,1,32,power_law_1.01,0.05454720020294189
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,1,32,power_law_1.01,0.05476352214813233
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,32,power_law_1.01,1.0426995086669923
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,1,32,balanced,0.08422656059265136
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,1,32,power_law_1.01,0.054583039283752445
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,1,32,power_law_1.01,0.05425280094146728
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,1,32,power_law_1.01,0.048474879264831544
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,1,32,power_law_1.01,0.051225600242614744
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,1,32,power_law_1.01,0.0521395206451416
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,1,32,power_law_1.01,0.05228799819946289
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,1,32,power_law_1.01,0.050959358215332026
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,1,32,power_law_1.01,0.05118080139160156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,1,32,power_law_1.01,0.051333122253417976
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,1,32,power_law_1.01,0.05147264003753662
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,1,32,power_law_1.01,0.0516761589050293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,1,32,power_law_1.01,0.05262335777282715
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,1,32,power_law_1.01,0.05270527839660645
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,1,32,power_law_1.01,0.05407872200012207
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,1,32,balanced,0.11318400382995605
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,1,32,power_law_1.01,0.06325119972229004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,32,power_law_1.01,1.2376191711425781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,1,32,balanced,0.13603967666625977
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,1,32,power_law_1.01,0.0542848014831543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,1,32,power_law_1.01,0.05782527923583984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,1,32,power_law_1.01,0.08198143959045409
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,1,32,power_law_1.01,0.07180416107177734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,1,32,power_law_1.01,0.10168576240539551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,32,power_law_1.2,0.1437491226196289
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,1,32,power_law_1.01,0.1258521556854248
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,1,32,power_law_1.01,0.17142784118652343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,32,power_law_1.2,0.14400768280029297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,32,power_law_1.2,0.1434496021270752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,32,power_law_1.2,0.18225791931152344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,32,power_law_1.2,0.15835647583007811
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,32,power_law_1.2,0.18945152282714844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,32,power_law_1.2,0.2014886474609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,32,power_law_1.2,0.19969280242919923
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,32,power_law_1.2,0.19909759521484377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,32,power_law_1.2,0.2046681594848633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,32,power_law_1.2,0.14334976196289062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,32,power_law_1.2,0.20256000518798828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,32,power_law_1.2,0.2069055938720703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,32,power_law_1.2,0.21364864349365234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,32,power_law_1.2,0.22931840896606442
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,32,power_law_1.2,0.2415551948547363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,32,power_law_1.2,0.23351295471191405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,32,power_law_1.2,0.3156236839294434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,32,power_law_1.2,0.375728645324707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,32,power_law_1.2,0.4584908676147461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,32,power_law_1.2,0.6503449249267578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,32,power_law_1.2,0.9512857818603516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,32,power_law_1.2,1.2073958587646483
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,32,power_law_1.2,0.08735103607177734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,32,power_law_1.2,1.7041357421874999
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,32,power_law_1.2,0.08464127540588379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,32,power_law_1.2,0.08531840324401856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,32,power_law_1.2,0.08484992027282715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,32,power_law_1.2,0.08828288078308105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,32,power_law_1.2,2.3264178466796874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,32,power_law_1.2,0.07894015789031983
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,32,power_law_1.2,0.0879270362854004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,32,power_law_1.2,0.09315199851989746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,32,power_law_1.2,0.0914470386505127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,32,power_law_1.2,0.09213055610656738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,32,power_law_1.2,0.09912447929382324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,32,power_law_1.2,0.10268927574157716
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,32,power_law_1.2,0.10677248001098634
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,1,32,power_law_1.01,0.2012825584411621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,32,power_law_1.2,0.10765567779541016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,32,power_law_1.2,0.1088742446899414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,32,power_law_1.2,0.11871871948242188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,32,power_law_1.2,0.14802559852600097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,32,power_law_1.2,0.16428159713745116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,32,power_law_1.2,3.826329650878906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,32,power_law_1.2,0.1997542381286621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,32,power_law_1.2,0.231146240234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,32,power_law_1.2,0.09331199645996094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,32,power_law_1.2,0.3259648132324219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,32,power_law_1.2,0.5025113677978515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,32,power_law_1.2,0.6615309143066407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,1,32,power_law_1.2,0.05497344017028809
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,32,power_law_1.2,4.656510620117187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,32,power_law_1.2,0.7782720184326172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,1,32,power_law_1.2,0.054434561729431154
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,1,32,power_law_1.2,0.05376128196716309
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,1,32,power_law_1.2,0.054096641540527346
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,1,32,power_law_1.2,0.04678143978118897
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,32,power_law_1.2,1.2410982513427735
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,1,32,power_law_1.2,0.05090176105499268
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,1,32,power_law_1.2,0.05100031852722168
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,1,32,power_law_1.2,0.05156352043151855
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,1,32,power_law_1.2,0.05107840061187744
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,1,32,power_law_1.2,0.05163263797760009
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,1,32,power_law_1.2,0.05216896057128906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,32,power_law_1.2,1.7985752868652345
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,1,32,power_law_1.2,0.052296957969665527
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,1,32,power_law_1.2,0.05252480030059814
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,1,32,power_law_1.2,0.05371263980865478
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,1,32,power_law_1.2,0.05016704082489014
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,1,32,power_law_1.01,0.3059391975402832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,1,32,power_law_1.01,0.3656563186645508
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,1,32,power_law_1.2,0.05255167961120606
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,1,32,power_law_1.2,0.05811200141906738
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,1,32,power_law_1.2,0.06300543785095215
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,1,32,power_law_1.2,0.08
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,1,32,power_law_1.2,0.06957312107086182
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,16,balanced,0.04232192039489746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,16,balanced,0.04125311851501465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,16,balanced,0.06288512229919434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,16,balanced,0.0748198413848877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,16,balanced,0.09724287986755371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,16,balanced,0.1418278408050537
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,1,32,power_law_1.2,0.09783807754516602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,16,balanced,0.18811391830444335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,16,balanced,0.19538944244384765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,16,balanced,0.193621768951416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,16,balanced,0.191909122467041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,16,balanced,0.19643775939941405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,16,balanced,0.20142335891723634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,16,balanced,0.2059916877746582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,16,balanced,0.20004480361938476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,16,balanced,0.20502527236938475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,16,balanced,0.21125503540039064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,16,balanced,0.22042112350463866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,16,balanced,0.23727487564086913
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,16,balanced,0.2520115280151367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,16,balanced,0.2887436866760254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,16,balanced,0.32089088439941404
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,16,balanced,0.44693759918212894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,16,balanced,0.531863021850586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,16,balanced,0.7220953369140626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,16,balanced,0.9056665802001953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,16,balanced,1.3263052368164063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,16,balanced,0.03632512092590332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,16,balanced,0.036266241073608395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,16,balanced,1.729228820800781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,16,balanced,0.038743040561676025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,16,balanced,0.04648575782775879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,16,balanced,0.06453120231628418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,16,balanced,0.09092608451843262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,16,balanced,0.10882304191589356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,16,balanced,0.10114687919616698
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,16,balanced,0.10419967651367187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,16,balanced,0.1052006435394287
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,16,balanced,0.10716032028198241
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,16,balanced,0.11028991699218751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,16,balanced,0.1138035202026367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,16,balanced,0.10501503944396973
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,16,balanced,0.10792448043823241
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,16,balanced,0.11133952140808105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,16,balanced,0.11891967773437499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,16,balanced,0.12643199920654297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,16,balanced,0.14042495727539062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,16,balanced,0.16148096084594726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,16,balanced,0.18337791442871093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,16,balanced,0.227194881439209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,16,balanced,0.26490623474121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,16,balanced,0.38582782745361327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,16,balanced,0.4681856155395508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,16,balanced,0.6636019134521485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,16,balanced,0.8733439636230468
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,1,32,power_law_1.2,0.12090623855590821
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,1,32,power_law_1.2,0.15302528381347655
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,1,32,power_law_1.2,0.2256844711303711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,1,32,power_law_1.2,0.2861414337158203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,1,32,power_law_1.2,0.4456614303588867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,1,32,power_law_1.2,0.5765683364868164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,2,16,balanced,0.02185983896255493
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,2,16,balanced,0.021684479713439942
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,2,16,balanced,0.057396478652954094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,2,16,balanced,0.057160959243774415
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,2,16,balanced,0.05712255954742431
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,2,16,balanced,0.05727104187011719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,2,16,balanced,0.05723264217376709
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,2,16,balanced,0.05746304035186768
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,2,16,balanced,0.05736063957214356
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,2,16,balanced,0.05740287780761719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,2,16,balanced,0.05785215854644775
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,2,16,balanced,0.05949696063995361
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,2,16,balanced,0.05962368011474609
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,2,16,balanced,0.06193151950836182
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,2,16,balanced,0.06326528072357178
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,2,16,balanced,0.06613503932952881
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,2,16,balanced,0.06884479999542237
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,2,16,balanced,0.022429440021514892
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,2,16,balanced,0.025547521114349363
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,2,16,balanced,0.036652801036834715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,16,power_law_1.01,0.0963814353942871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,16,power_law_1.01,0.13789055824279786
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,16,power_law_1.01,0.13805184364318848
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,2,16,balanced,0.05671679973602295
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,2,16,balanced,0.056430082321166995
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,2,16,balanced,0.05653632164001465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,16,power_law_1.01,0.13796095848083495
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,16,power_law_1.01,0.14041855812072754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,16,power_law_1.01,0.1702707290649414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,16,power_law_1.01,0.1811097526550293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,16,power_law_1.01,0.1835468864440918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,16,power_law_1.01,0.18652160644531252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,16,power_law_1.01,0.18917247772216797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,16,power_law_1.01,0.19284223556518554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,16,power_law_1.01,0.20025600433349608
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,16,power_law_1.01,0.20025728225708006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,16,power_law_1.01,0.20166528701782224
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,16,power_law_1.01,0.20440448760986327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,16,power_law_1.01,0.2206502342224121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,16,power_law_1.01,0.24758783340454102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,16,power_law_1.01,0.2929190444946289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,16,power_law_1.01,0.38334720611572265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,16,power_law_1.01,0.2029657554626465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,16,power_law_1.01,0.469251823425293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,16,power_law_1.01,0.6459494018554688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,16,power_law_1.01,0.7590182495117188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,16,power_law_1.01,1.1433932495117187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,16,power_law_1.01,1.4301708984375001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,16,power_law_1.01,0.0880678367614746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,16,power_law_1.01,0.08769536018371582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,16,power_law_1.01,0.06128128051757813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,16,power_law_1.01,0.08705280303955079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,16,power_law_1.01,2.347440643310547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,16,power_law_1.01,0.09185152053833008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,16,power_law_1.01,0.08398719787597655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,16,power_law_1.01,0.08769408226013184
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,16,power_law_1.01,0.0903756809234619
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,16,power_law_1.01,0.09272704124450684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,16,power_law_1.01,0.09617535591125488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,16,power_law_1.01,0.0978816032409668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,16,power_law_1.01,2.8089086914062498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,16,power_law_1.01,0.10536064147949217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,16,power_law_1.01,0.11060992240905762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,16,power_law_1.01,0.10149760246276855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,16,power_law_1.01,0.10376704216003418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,16,power_law_1.01,0.10668288230895998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,16,power_law_1.01,0.11001600265502928
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,16,power_law_1.01,0.13100543975830078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,16,power_law_1.01,0.1466649627685547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,16,power_law_1.01,0.19877759933471678
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,16,power_law_1.01,0.23287296295166016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,16,power_law_1.01,0.3075008010864258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,16,power_law_1.01,0.3919014358520508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,16,power_law_1.01,0.5144883346557617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,16,power_law_1.01,0.6737868499755859
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,2,16,power_law_1.01,0.03803136110305786
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,2,16,power_law_1.01,0.05604608058929443
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,2,16,power_law_1.01,0.056055040359497064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,16,power_law_1.01,1.0161011505126953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,2,16,power_law_1.01,0.05515903949737548
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,2,16,power_law_1.01,0.056290559768676765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,16,power_law_1.01,1.1842546844482422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,2,16,power_law_1.01,0.04901887893676758
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,2,16,power_law_1.01,0.05113855838775635
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,2,16,power_law_1.01,0.05238016128540039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,2,16,power_law_1.01,0.052227840423583985
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,2,16,power_law_1.01,0.053089280128479
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,2,16,power_law_1.01,0.053521919250488284
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,2,16,power_law_1.01,0.053465600013732906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,2,16,power_law_1.01,0.054169597625732424
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,2,16,power_law_1.01,0.05596672058105469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,2,16,power_law_1.01,0.05534207820892334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,2,16,power_law_1.01,0.05586559772491455
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,2,16,power_law_1.01,0.05762432098388672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,2,16,balanced,0.09693440437316894
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,2,16,balanced,0.10257023811340331
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,2,16,balanced,0.13623295783996583
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,2,16,balanced,0.16749696731567382
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,2,16,power_law_1.01,0.0644262409210205
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,2,16,power_law_1.01,0.06960000038146973
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,2,16,power_law_1.01,0.09088128089904786
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,2,16,power_law_1.01,0.07836287975311279
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,16,power_law_1.2,0.09699968338012696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,16,power_law_1.2,0.13852928161621095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,16,power_law_1.2,0.12552063941955566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,16,power_law_1.2,0.1391411209106445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,16,power_law_1.2,0.14166655540466308
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,16,power_law_1.2,0.16587263107299804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,16,power_law_1.2,0.17839616775512696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,16,power_law_1.2,0.18527231216430665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,16,power_law_1.2,0.18779903411865234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,16,power_law_1.2,0.18542207717895506
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,16,power_law_1.2,0.1988876724243164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,2,16,power_law_1.01,0.11751296043395995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,16,power_law_1.2,0.19846912384033205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,16,power_law_1.2,0.20102655410766604
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,16,power_law_1.2,0.20543359756469726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,16,power_law_1.2,0.21141887664794923
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,16,power_law_1.2,0.23671167373657226
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,2,16,power_law_1.01,0.1364889621734619
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,16,power_law_1.2,0.269932804107666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,16,power_law_1.2,0.3165350341796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,16,power_law_1.2,0.4587251281738281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,16,power_law_1.2,0.19419647216796876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,16,power_law_1.2,0.5056959915161132
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,16,power_law_1.2,0.7096717071533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,16,power_law_1.2,0.8590656280517578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,16,power_law_1.2,1.326128692626953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,16,power_law_1.2,1.8403570556640623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,16,power_law_1.2,0.060712962150573736
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,2,16,power_law_1.01,0.1835353660583496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,16,power_law_1.2,0.08623231887817383
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,2,16,power_law_1.01,0.23844863891601564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,16,power_law_1.2,0.09093631744384766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,16,power_law_1.2,0.09052672386169433
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,16,power_law_1.2,0.08401535987854004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,16,power_law_1.2,2.616632385253906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,16,power_law_1.2,0.08937855720520019
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,16,power_law_1.2,0.0906048011779785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,16,power_law_1.2,0.0937177562713623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,16,power_law_1.2,0.10331775665283202
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,16,power_law_1.2,0.10619647979736328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,16,power_law_1.2,0.10978816032409669
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,16,power_law_1.2,0.10345215797424316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,16,power_law_1.2,0.0798412799835205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,16,power_law_1.2,0.10748543739318847
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,16,power_law_1.2,0.10921728134155273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,16,power_law_1.2,0.12468607902526856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,16,power_law_1.2,0.1491916847229004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,16,power_law_1.2,0.1704243278503418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,16,power_law_1.2,3.3110861206054687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,16,power_law_1.2,0.21132671356201174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,16,power_law_1.2,0.2488934326171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,16,power_law_1.2,0.09493120193481445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,16,power_law_1.2,0.3347622299194336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,16,power_law_1.2,0.4807167816162109
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,2,16,power_law_1.2,0.03807615995407104
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,16,power_law_1.2,0.6057408142089844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,2,16,power_law_1.2,0.05572864055633545
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,2,16,power_law_1.2,0.05140351772308349
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,16,power_law_1.2,0.7522278594970703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,2,16,power_law_1.2,0.055160322189331056
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,2,16,power_law_1.2,0.05653247833251953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,2,16,power_law_1.2,0.044035840034484866
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,2,16,power_law_1.2,0.04979328155517578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,16,power_law_1.2,1.0224832153320311
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,2,16,power_law_1.2,0.05164031982421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,2,16,power_law_1.2,0.051971840858459475
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,2,16,power_law_1.2,0.053209600448608396
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,2,16,power_law_1.2,0.05425663948059082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,2,16,power_law_1.2,0.05542143821716309
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,2,16,power_law_1.2,0.05591423988342285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,16,power_law_1.2,1.5136933898925782
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,2,16,power_law_1.2,0.05762688159942627
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,2,16,power_law_1.2,0.06131711959838867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,2,16,power_law_1.2,0.05009791851043701
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,2,16,power_law_1.2,0.054402561187744135
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,2,16,power_law_1.01,0.3353267288208008
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,2,16,power_law_1.01,0.44230911254882815
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,2,16,power_law_1.2,0.06777088165283203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,2,16,power_law_1.2,0.07690624237060548
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,2,16,power_law_1.2,0.08736767768859863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,8,balanced,0.0347763204574585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,8,balanced,0.03639935970306397
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,2,16,power_law_1.2,0.13820672035217285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,2,16,power_law_1.2,0.11587072372436524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,8,balanced,0.061806077957153316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,8,balanced,0.07648640155792237
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,8,balanced,0.097774076461792
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,8,balanced,0.14859135627746584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,2,16,power_law_1.2,0.17640192031860352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,8,balanced,0.18871295928955079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,8,balanced,0.19152000427246094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,8,balanced,0.19617408752441406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,8,balanced,0.20141567230224608
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,8,balanced,0.20034816741943356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,8,balanced,0.20328960418701172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,8,balanced,0.21245311737060546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,8,balanced,0.21747072219848632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,8,balanced,0.22808704376220704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,8,balanced,0.24229503631591798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,8,balanced,0.27360256195068355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,8,balanced,0.3044646453857422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,8,balanced,0.36810367584228515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,8,balanced,0.43139583587646485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,8,balanced,0.1941324806213379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,8,balanced,0.6082854461669922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,8,balanced,0.7474034881591797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,8,balanced,1.043545608520508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,8,balanced,0.039133439064025874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,8,balanced,1.3463845825195313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,8,balanced,0.03822207927703857
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,8,balanced,0.0448524808883667
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,8,balanced,0.05534719944000244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,8,balanced,0.07264639854431151
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,8,balanced,0.10361856460571288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,8,balanced,0.11583616256713866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,8,balanced,0.11672320365905761
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,8,balanced,1.9736550903320313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,8,balanced,0.1218393611907959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,8,balanced,0.1273292827606201
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,8,balanced,0.11535615921020508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,8,balanced,0.12041472434997558
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,8,balanced,0.12280447959899903
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,8,balanced,0.12782976150512695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,8,balanced,0.1295257568359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,8,balanced,0.13532544136047364
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,8,balanced,0.14695551872253418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,8,balanced,2.7358016967773438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,8,balanced,0.1662335968017578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,8,balanced,0.18594688415527344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,8,balanced,0.2215590476989746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,8,balanced,0.26167295455932615
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,8,balanced,0.34075904846191407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,8,balanced,0.41818111419677734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,8,balanced,0.6042700958251953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,8,balanced,0.7551052856445313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,8,balanced,1.1002547454833984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,8,balanced,1.4449305725097656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,2,16,power_law_1.2,0.27654912948608396
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,2,16,power_law_1.2,0.35040897369384766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,2,16,power_law_1.2,0.46722560882568354
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,2,16,power_law_1.2,0.6847756958007812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,4,8,balanced,0.02431999921798706
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,4,8,balanced,0.023880960941314696
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,4,8,balanced,0.02592895984649658
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,4,8,balanced,0.06487040042877197
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,4,8,balanced,0.06518015861511231
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,4,8,balanced,0.06474751949310302
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,4,8,balanced,0.06501760005950927
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,4,8,balanced,0.06549376010894775
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,4,8,balanced,0.0665011215209961
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,4,8,balanced,0.029785599708557133
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,4,8,balanced,0.06637951850891113
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,4,8,balanced,0.06768640041351318
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,4,8,balanced,0.0695039987564087
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,4,8,balanced,0.07054592132568359
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,4,8,balanced,0.07469183921813964
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,4,8,balanced,0.07669248104095458
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,4,8,balanced,0.08290687561035157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,4,8,balanced,0.0883187198638916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,4,8,balanced,0.06521471977233886
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,4,8,balanced,0.04313216209411621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,8,power_law_1.01,0.07548416137695313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,8,power_law_1.01,0.0945075225830078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,4,8,balanced,0.06408959865570069
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,8,power_law_1.01,0.10435968399047853
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,8,power_law_1.01,0.10764543533325195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,8,power_law_1.01,0.13170559883117677
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,8,power_law_1.01,0.12119296073913574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,8,power_law_1.01,0.17388671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,8,power_law_1.01,0.1783078384399414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,8,power_law_1.01,0.17879167556762696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,8,power_law_1.01,0.1870297622680664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,8,power_law_1.01,0.1936115264892578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,8,power_law_1.01,0.18830976486206055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,8,power_law_1.01,0.19222272872924803
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,8,power_law_1.01,0.19539455413818357
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,8,power_law_1.01,0.19776512145996095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,8,power_law_1.01,0.20279167175292967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,8,power_law_1.01,0.21884799957275392
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,8,power_law_1.01,0.2550144004821777
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,8,power_law_1.01,0.28245887756347654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,8,power_law_1.01,0.3677017593383789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,8,power_law_1.01,0.4764416122436524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,8,power_law_1.01,0.6212287902832031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,8,power_law_1.01,0.7936370849609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,8,power_law_1.01,1.0533042907714845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,8,power_law_1.01,1.3491981506347657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,8,power_law_1.01,1.9280281066894531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,8,power_law_1.01,0.0452569580078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,8,power_law_1.01,0.06663551807403564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,8,power_law_1.01,0.07458432197570801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,8,power_law_1.01,0.07531775951385497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,8,power_law_1.01,2.7070220947265624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,8,power_law_1.01,0.08538368225097656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,8,power_law_1.01,0.09904255867004394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,8,power_law_1.01,0.09526911735534668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,8,power_law_1.01,0.10340352058410644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,8,power_law_1.01,0.10732671737670899
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,8,power_law_1.01,0.11049471855163576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,8,power_law_1.01,0.12002559661865235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,8,power_law_1.01,0.11264127731323241
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,8,power_law_1.01,0.11164928436279298
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,4,8,balanced,0.06444287776947022
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,8,power_law_1.01,0.1138969612121582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,8,power_law_1.01,0.11685888290405273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,8,power_law_1.01,0.12388223648071288
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,4,8,balanced,0.06498303890228271
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,8,power_law_1.01,0.12814080238342285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,8,power_law_1.01,0.14951295852661134
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,8,power_law_1.01,0.16779008865356446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,8,power_law_1.01,0.21744768142700197
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,8,power_law_1.01,0.2652518463134766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,8,power_law_1.01,0.33463680267333984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,8,power_law_1.01,0.4400665664672852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,8,power_law_1.01,0.5855936050415039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,8,power_law_1.01,0.7286489868164063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,4,8,power_law_1.01,0.028893439769744872
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,4,8,power_law_1.01,0.04377855777740479
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,4,8,power_law_1.01,0.05241983890533447
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,8,power_law_1.01,1.0706893157958983
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,4,8,power_law_1.01,0.052779521942138675
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,4,8,power_law_1.01,0.05722496032714843
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,4,8,power_law_1.01,0.04967679977416992
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,4,8,power_law_1.01,0.06131584167480468
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,4,8,power_law_1.01,0.05848576068878174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,8,power_law_1.01,1.4069517517089845
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,4,8,power_law_1.01,0.06001920223236083
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,4,8,power_law_1.01,0.059520001411437984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,4,8,power_law_1.01,0.06162303924560547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,4,8,power_law_1.01,0.06173056125640869
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,4,8,power_law_1.01,0.06279551982879639
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,4,8,power_law_1.01,0.06540544033050537
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,4,8,power_law_1.01,0.0682700777053833
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,4,8,power_law_1.01,0.07268223762512208
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,4,8,power_law_1.01,0.06457215785980225
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,4,8,balanced,0.1272486400604248
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,4,8,balanced,0.13981568336486816
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,4,8,balanced,0.24185344696044925
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,4,8,balanced,0.19274368286132812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,4,8,power_law_1.01,0.07839488029479981
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,4,8,power_law_1.01,0.0833523178100586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,4,8,power_law_1.01,0.1041907215118408
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,4,8,power_law_1.01,0.1180992031097412
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,8,power_law_1.2,0.074999041557312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,8,power_law_1.2,0.09356032371520996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,8,power_law_1.2,0.0935206413269043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,8,power_law_1.2,0.11074687957763671
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,8,power_law_1.2,0.12698240280151368
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,8,power_law_1.2,0.11209343910217284
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,8,power_law_1.2,0.1741529655456543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,8,power_law_1.2,0.18035455703735354
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,8,power_law_1.2,0.17960832595825196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,8,power_law_1.2,0.18199296951293947
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,8,power_law_1.2,0.19370624542236328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,8,power_law_1.2,0.18983295440673828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,8,power_law_1.2,0.18922752380371094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,8,power_law_1.2,0.19453311920166017
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,8,power_law_1.2,0.20254207611083985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,8,power_law_1.2,0.2081126403808594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,8,power_law_1.2,0.2335296058654785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,8,power_law_1.2,0.255732479095459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,8,power_law_1.2,0.2990950393676758
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,4,8,power_law_1.01,0.14019583702087401
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,8,power_law_1.2,0.36724224090576174
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,4,8,power_law_1.01,0.18240896224975586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,8,power_law_1.2,0.4982553482055664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,8,power_law_1.2,0.6602713775634765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,8,power_law_1.2,0.8468838500976563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,8,power_law_1.2,1.164232940673828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,4,8,power_law_1.01,0.3219596862792969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,4,8,power_law_1.01,0.2635622406005859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,8,power_law_1.2,0.045669121742248533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,8,power_law_1.2,0.06691840171813965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,8,power_law_1.2,1.5502554321289064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,8,power_law_1.2,0.07459455966949463
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,8,power_law_1.2,0.07403776168823242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,8,power_law_1.2,0.08571136474609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,8,power_law_1.2,0.09657855987548827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,8,power_law_1.2,0.09875200271606446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,8,power_law_1.2,0.0869542407989502
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,8,power_law_1.2,0.10238719940185548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,8,power_law_1.2,0.10590592384338379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,8,power_law_1.2,0.11402112007141114
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,8,power_law_1.2,0.10964608192443848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,8,power_law_1.2,0.11607168197631837
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,8,power_law_1.2,2.2426048278808595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,8,power_law_1.2,0.1184166431427002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,8,power_law_1.2,0.12326272010803223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,8,power_law_1.2,0.1389913558959961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,8,power_law_1.2,0.15688063621520995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,8,power_law_1.2,0.18870784759521483
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,8,power_law_1.2,0.22929664611816408
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,8,power_law_1.2,0.2715750312805176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,8,power_law_1.2,0.37477760314941405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,8,power_law_1.2,0.11086208343505859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,8,power_law_1.2,3.038577880859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,8,power_law_1.2,0.45711360931396483
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,4,8,power_law_1.2,0.029635839462280274
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,4,8,power_law_1.2,0.042929921150207516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,4,8,power_law_1.2,0.04678143978118897
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,4,8,power_law_1.2,0.053857278823852536
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,8,power_law_1.2,0.8023999786376953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,4,8,power_law_1.2,0.055413761138916016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,4,8,power_law_1.2,0.054237442016601564
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,4,8,power_law_1.2,0.0565388822555542
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,8,power_law_1.2,1.1005606079101562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,4,8,power_law_1.2,0.05875072002410888
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,4,8,power_law_1.2,0.060056319236755376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,4,8,power_law_1.2,0.061560320854187014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,8,power_law_1.2,0.6558668518066406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,8,power_law_1.2,1.4812159729003906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,4,8,power_law_1.2,0.06504576206207276
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,4,8,power_law_1.2,0.06499839782714843
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,4,8,power_law_1.2,0.06678400039672852
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,4,8,power_law_1.2,0.06958079814910889
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,4,8,power_law_1.2,0.0769049596786499
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,4,8,power_law_1.2,0.058430719375610354
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,4,8,power_law_1.2,0.060426239967346196
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,4,8,power_law_1.01,0.4732672119140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,4,8,power_law_1.01,0.5649049758911133
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,4,8,power_law_1.2,0.09841535568237306
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,4,8,power_law_1.2,0.12184576034545898
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,4,8,power_law_1.2,0.13975680351257325
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,4,8,power_law_1.2,0.08608384132385254
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,4,8,power_law_1.2,0.21043712615966798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,8,4,balanced,0.039031040668487546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,8,4,balanced,0.03543936014175415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,8,4,balanced,0.061687040328979495
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,8,4,balanced,0.07859456062316894
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,4,8,power_law_1.2,0.24148992538452146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,8,4,balanced,0.09896960258483886
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,8,4,balanced,0.1499673557281494
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,8,4,balanced,0.14615936279296876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,8,4,balanced,0.1478656005859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,8,4,balanced,0.1965017509460449
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,8,4,balanced,0.19936384201049803
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,8,4,balanced,0.20350847244262696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,8,4,balanced,0.2069875144958496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,8,4,balanced,0.21239040374755858
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,8,4,balanced,0.2158412742614746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,8,4,balanced,0.22179456710815432
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,8,4,balanced,0.22885887145996095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,8,4,balanced,0.24319103240966794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,8,4,balanced,0.2754483222961426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,8,4,balanced,0.30404479980468746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,8,4,balanced,0.36183040618896484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,8,4,balanced,0.42346240997314455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,8,4,balanced,0.5985228729248047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,8,4,balanced,0.7296396636962891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,8,4,balanced,1.023683853149414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,8,4,balanced,0.03495424032211304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,8,4,balanced,1.3069798278808595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,8,4,balanced,0.03852544069290161
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,8,4,balanced,0.0460262393951416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,8,4,balanced,0.06609024047851562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,8,4,balanced,0.0863526439666748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,8,4,balanced,0.12194815635681153
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,8,4,balanced,0.12347519874572754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,8,4,balanced,0.12534399986267092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,8,4,balanced,1.9238604736328124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,8,4,balanced,0.13949952125549317
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,8,4,balanced,0.14135680198669434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,8,4,balanced,0.14246399879455568
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,8,4,balanced,0.14722687721252442
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,8,4,balanced,0.1499724769592285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,8,4,balanced,0.15460607528686524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,8,4,balanced,0.15528703689575196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,8,4,balanced,0.1616806411743164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,8,4,balanced,0.1701043128967285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,8,4,balanced,0.18936832427978517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,8,4,balanced,2.674194030761719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,8,4,balanced,0.2072537612915039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,8,4,balanced,0.24248064041137693
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,8,4,balanced,0.27830272674560547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,8,4,balanced,0.3478361511230469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,8,4,balanced,0.4224051284790039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,8,4,balanced,0.6307916641235352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,8,4,balanced,0.7715430450439453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,8,4,balanced,1.1259519958496094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,8,4,balanced,1.473289031982422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,4,8,power_law_1.2,0.38808830261230465
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,4,8,power_law_1.2,0.47757568359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,4,8,power_law_1.2,0.8299916839599609
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,4,8,power_law_1.2,1.0281932830810547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,8,4,balanced,0.03093247890472412
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,8,4,balanced,0.08428928375244141
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,8,4,balanced,0.08422911643981934
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,8,4,balanced,0.08457088470458984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,8,4,balanced,0.034967041015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,8,4,balanced,0.032084479331970214
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,8,4,balanced,0.08449664115905761
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,8,4,balanced,0.08520447731018066
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,8,4,balanced,0.08643712043762207
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,8,4,balanced,0.08699135780334473
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,8,4,balanced,0.08773119926452637
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,8,4,balanced,0.0422438383102417
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,8,4,balanced,0.08855551719665528
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,8,4,balanced,0.09240960121154786
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,8,4,balanced,0.09485568046569824
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,8,4,balanced,0.10086015701293945
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,8,4,balanced,0.106561279296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,8,4,balanced,0.11740415573120118
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,8,4,balanced,0.129683198928833
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,8,4,balanced,0.057123842239379885
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,8,4,power_law_1.01,0.06032896041870117
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,8,4,balanced,0.08376704216003418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,8,4,power_law_1.01,0.07531007766723632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,8,4,power_law_1.01,0.08464768409729004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,8,4,power_law_1.01,0.09413248062133789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,8,4,power_law_1.01,0.1088371181488037
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,8,4,power_law_1.01,0.11737728118896484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,8,4,power_law_1.01,0.1293452835083008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,8,4,power_law_1.01,0.13028096199035644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,8,4,power_law_1.01,0.18651647567749025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,8,4,power_law_1.01,0.19165439605712892
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,8,4,power_law_1.01,0.19254400253295897
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,8,4,power_law_1.01,0.19659135818481446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,8,4,power_law_1.01,0.19985919952392578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,8,4,power_law_1.01,0.2085491180419922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,8,4,power_law_1.01,0.21242240905761717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,8,4,power_law_1.01,0.22504831314086915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,8,4,power_law_1.01,0.24024063110351562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,8,4,power_law_1.01,0.26591360092163085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,8,4,power_law_1.01,0.2990233612060547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,8,4,power_law_1.01,0.37039104461669925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,8,4,power_law_1.01,0.45630718231201167
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,8,4,power_law_1.01,0.6364262390136719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,8,4,power_law_1.01,0.7741542053222656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,8,4,power_law_1.01,1.0705165100097656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,8,4,power_law_1.01,1.401070098876953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,8,4,power_law_1.01,2.044300842285156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,8,4,power_law_1.01,0.040396800041198735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,8,4,power_law_1.01,0.0530675220489502
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,8,4,power_law_1.01,0.06579967975616455
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,8,4,balanced,0.0840998363494873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,8,4,power_law_1.01,2.6761920166015623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,8,4,power_law_1.01,0.08008064270019531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,8,4,power_law_1.01,0.09174271583557128
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,8,4,balanced,0.08416255950927734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,8,4,power_law_1.01,0.1033523178100586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,8,4,power_law_1.01,0.10311552047729493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,8,4,power_law_1.01,0.10718336105346679
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,8,4,power_law_1.01,0.11920255661010741
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,8,4,power_law_1.01,0.12088319778442383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,8,4,power_law_1.01,0.12550016403198244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,8,4,power_law_1.01,0.1298201560974121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,8,4,power_law_1.01,0.13284607887268068
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,8,4,power_law_1.01,0.14150143623352052
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,8,4,power_law_1.01,0.14247039794921873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,8,4,power_law_1.01,0.15049856185913085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,8,4,power_law_1.01,0.16057855606079102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,8,4,power_law_1.01,0.1785856056213379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,8,4,power_law_1.01,0.20052480697631836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,8,4,power_law_1.01,0.2573478317260742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,8,4,power_law_1.01,0.3119436836242676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,8,4,power_law_1.01,0.3988608169555664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,8,4,power_law_1.01,0.477789421081543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,8,4,power_law_1.01,0.6616857910156251
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,8,4,power_law_1.01,0.03453824043273926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,8,4,power_law_1.01,0.8315033721923829
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,8,4,power_law_1.01,0.040733442306518555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,8,4,power_law_1.01,0.04795263767242432
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,8,4,power_law_1.01,0.05455615997314454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,8,4,power_law_1.01,1.1865805053710938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,8,4,power_law_1.01,0.060312318801879886
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,8,4,power_law_1.01,0.06324863910675049
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,8,4,power_law_1.01,0.07138688087463378
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,8,4,power_law_1.01,0.07695104122161865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,8,4,power_law_1.01,1.5898701477050783
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,8,4,power_law_1.01,0.07456255912780761
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,8,4,power_law_1.01,0.07603072166442872
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,8,4,power_law_1.01,0.07776383876800537
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,8,4,power_law_1.01,0.07974400043487548
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,8,4,power_law_1.01,0.07966207981109619
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,8,4,power_law_1.01,0.08361599922180177
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,8,4,power_law_1.01,0.08713855743408203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,8,4,power_law_1.01,0.09124480247497559
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,8,4,power_law_1.01,0.09256064414978027
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,8,4,balanced,0.1931558418273926
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,8,4,balanced,0.39139072418212895
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,8,4,balanced,0.3069721603393555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,8,4,balanced,0.21875839233398436
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,8,4,power_law_1.01,0.106179838180542
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,8,4,power_law_1.01,0.17456127166748048
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,8,4,power_law_1.01,0.11576064109802246
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,8,4,power_law_1.01,0.1410086441040039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,8,4,power_law_1.2,0.06052095890045166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,8,4,power_law_1.2,0.07360640048980713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,8,4,power_law_1.2,0.07745535850524903
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,8,4,power_law_1.2,0.09306495666503907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,8,4,power_law_1.2,0.10051967620849608
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,8,4,power_law_1.2,0.11268863677978516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,8,4,power_law_1.2,0.12677632331848146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,8,4,power_law_1.2,0.13293824195861817
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,8,4,power_law_1.2,0.1865203285217285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,8,4,power_law_1.2,0.18846208572387696
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,8,4,power_law_1.01,0.22977279663085937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,8,4,power_law_1.2,0.1956211280822754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,8,4,power_law_1.2,0.1990399932861328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,8,4,power_law_1.01,0.2605824089050293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,8,4,power_law_1.2,0.2017689514160156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,8,4,power_law_1.2,0.20917503356933595
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,8,4,power_law_1.01,0.43459457397460943
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,8,4,power_law_1.2,0.2140428733825684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,8,4,power_law_1.2,0.22584575653076172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,8,4,power_law_1.2,0.23953535079956056
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,8,4,power_law_1.2,0.272861442565918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,8,4,power_law_1.2,0.30650495529174804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,8,4,power_law_1.2,0.3766592025756836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,8,4,power_law_1.2,0.47505535125732423
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,8,4,power_law_1.01,0.5238579177856446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,8,4,power_law_1.2,0.6520473480224609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,8,4,power_law_1.2,0.8253427124023437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,8,4,power_law_1.2,0.039851520061492916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,8,4,power_law_1.2,0.04900991916656494
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,8,4,power_law_1.2,1.1366796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,8,4,power_law_1.2,1.4279653930664062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,8,4,power_law_1.2,0.07660672187805176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,8,4,power_law_1.2,0.08681344032287598
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,8,4,power_law_1.2,0.10310784339904785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,8,4,power_law_1.2,0.10579711914062499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,8,4,power_law_1.2,0.12032511711120605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,8,4,power_law_1.2,2.126254119873047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,8,4,power_law_1.2,0.05940351963043213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,8,4,power_law_1.2,0.12196991920471192
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,8,4,power_law_1.2,0.12981375694274902
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,8,4,power_law_1.2,0.13315072059631347
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,8,4,power_law_1.2,0.14168831825256348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,8,4,power_law_1.2,0.09857024192810059
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,8,4,power_law_1.2,0.14343296051025392
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,8,4,power_law_1.2,0.15230719566345213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,8,4,power_law_1.2,0.162685432434082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,8,4,power_law_1.2,0.1865945625305176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,8,4,power_law_1.2,0.20939136505126954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,8,4,power_law_1.2,0.26945535659790043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,8,4,power_law_1.2,0.1319923210144043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,8,4,power_law_1.2,0.31632896423339846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,8,4,power_law_1.2,2.7710821533203123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,8,4,power_law_1.2,0.3970572662353516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,8,4,power_law_1.2,0.03422976016998291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,8,4,power_law_1.2,0.4868723297119141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,8,4,power_law_1.2,0.6925132751464844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,8,4,power_law_1.2,0.04124671936035156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,8,4,power_law_1.2,0.050007038116455085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,8,4,power_law_1.2,0.8760499572753906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,8,4,power_law_1.2,0.06321280002593994
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,8,4,power_law_1.2,0.06433919906616212
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,8,4,power_law_1.2,0.07164031982421874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,8,4,power_law_1.2,1.2360294342041016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,8,4,power_law_1.2,0.0749183988571167
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,8,4,power_law_1.2,0.0753766393661499
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,8,4,power_law_1.2,0.07818111896514893
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,8,4,power_law_1.2,1.5492594909667967
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,8,4,power_law_1.2,0.08006400108337403
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,8,4,power_law_1.2,0.04362624168395996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,8,4,power_law_1.2,0.08298879623413086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,8,4,power_law_1.2,0.0852735996246338
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,8,4,power_law_1.2,0.08787327766418457
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,8,4,power_law_1.2,0.09602944374084472
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,8,4,power_law_1.2,0.13542655944824217
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,8,4,power_law_1.2,0.07321983814239502
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,8,4,power_law_1.01,0.7317657470703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,8,4,power_law_1.01,1.0938406372070313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,8,4,power_law_1.2,0.10413311958312987
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,8,4,power_law_1.2,0.12185215950012207
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,8,4,power_law_1.2,0.2148518371582031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,8,4,power_law_1.2,0.18480255126953124
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,8,4,power_law_1.2,0.31337215423583986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,16,2,balanced,0.037474560737609866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,16,2,balanced,0.06315135955810547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,8,4,power_law_1.2,0.4153753662109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,16,2,balanced,0.08063360214233398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,16,2,balanced,0.10821632385253906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,16,2,balanced,0.15861503601074217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,16,2,balanced,0.16333440780639646
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,16,2,balanced,0.16065664291381837
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,16,2,balanced,0.1655731201171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,16,2,balanced,0.16550912857055664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,16,2,balanced,0.034204161167144774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,16,2,balanced,0.24642688751220704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,16,2,balanced,0.24925312042236328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,16,2,balanced,0.2576332855224609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,16,2,balanced,0.2614080047607422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,16,2,balanced,0.2694272041320801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,16,2,balanced,0.28240896224975587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,16,2,balanced,0.1590873622894287
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,16,2,balanced,0.31108736038208007
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,16,2,balanced,0.33886592864990234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,16,2,balanced,0.3963737487792969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,16,2,balanced,0.45473793029785153
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,16,2,balanced,0.6348006439208984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,16,2,balanced,0.7573017883300781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,16,2,balanced,1.0527040100097655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,16,2,balanced,0.035427839756011964
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,16,2,balanced,1.3302540588378906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,16,2,balanced,0.039927039146423336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,16,2,balanced,0.049141759872436526
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,16,2,balanced,0.0755673599243164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,16,2,balanced,0.11731840133666993
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,16,2,balanced,0.17192064285278322
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,16,2,balanced,1.957848358154297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,16,2,balanced,0.16802047729492187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,16,2,balanced,0.16952959060668946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,16,2,balanced,0.17179904937744142
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,16,2,balanced,0.17385856628417967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,16,2,balanced,0.175601921081543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,16,2,balanced,0.1990719985961914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,16,2,balanced,0.2020966339111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,16,2,balanced,0.2077132797241211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,16,2,balanced,0.2070425605773926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,16,2,balanced,0.21346431732177734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,16,2,balanced,0.2197478485107422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,16,2,balanced,0.2373043251037598
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,16,2,balanced,0.2538150405883789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,16,2,balanced,0.2879052734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,16,2,balanced,0.32108287811279296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,16,2,balanced,0.3919295883178711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,16,2,balanced,0.4566476821899414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,16,2,balanced,2.6765081787109373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,16,2,balanced,0.6971916961669922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,16,2,balanced,0.8314406585693359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,16,2,balanced,1.2144435119628905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,16,2,balanced,1.5999539184570313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,8,4,power_law_1.2,0.5818073654174805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,8,4,power_law_1.2,0.8117798614501954
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,8,4,power_law_1.2,1.3017613220214845
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,8,4,power_law_1.2,1.730040283203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,16,2,balanced,0.045698561668396
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,16,2,balanced,0.12579839706420898
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,16,2,balanced,0.12548864364624024
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,16,2,balanced,0.12685952186584473
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,16,2,balanced,0.04692992210388184
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,16,2,balanced,0.12750847816467287
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,16,2,balanced,0.1276979160308838
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,16,2,balanced,0.12921216011047362
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,16,2,balanced,0.1312076759338379
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,16,2,balanced,0.1329740810394287
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,16,2,balanced,0.13519359588623048
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,16,2,balanced,0.14208767890930177
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,16,2,balanced,0.14809856414794922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,16,2,balanced,0.052596478462219244
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,16,2,balanced,0.16929792404174804
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,16,2,balanced,0.19445247650146485
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,16,2,balanced,0.21942399978637694
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,16,2,balanced,0.06384511947631835
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,16,2,balanced,0.15772031784057616
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,16,2,balanced,0.12380800247192383
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,16,2,balanced,0.08295807838439942
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,16,2,power_law_1.01,0.03618560075759888
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,16,2,power_law_1.01,0.059795198440551756
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,16,2,power_law_1.01,0.07509632110595703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,16,2,power_law_1.01,0.08962047576904297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,16,2,power_law_1.01,0.1087168025970459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,16,2,power_law_1.01,0.12486271858215332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,16,2,power_law_1.01,0.13805952072143554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,16,2,power_law_1.01,0.13775360107421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,16,2,power_law_1.01,0.1414732837677002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,16,2,power_law_1.01,0.14566143989562988
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,16,2,power_law_1.01,0.15343359947204588
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,16,2,power_law_1.01,0.23833343505859378
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,16,2,power_law_1.01,0.24040447235107423
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,16,2,power_law_1.01,0.24971391677856447
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,16,2,power_law_1.01,0.2531942367553711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,16,2,power_law_1.01,0.2667110443115234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,16,2,power_law_1.01,0.2797171211242676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,16,2,power_law_1.01,0.3104537582397461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,16,2,power_law_1.01,0.3430476760864258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,16,2,power_law_1.01,0.4066944122314453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,16,2,power_law_1.01,0.47768703460693357
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,16,2,power_law_1.01,0.6839116668701172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,16,2,power_law_1.01,0.8134156799316405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,16,2,power_law_1.01,1.1177600097656248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,16,2,power_law_1.01,1.4197184753417968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,16,2,power_law_1.01,2.0169984436035158
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,16,2,power_law_1.01,0.039677441120147705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,16,2,power_law_1.01,0.046449918746948246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,16,2,power_law_1.01,0.060824317932128905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,16,2,power_law_1.01,2.618464050292969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,16,2,balanced,0.12437888145446778
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,16,2,power_law_1.01,0.08222847938537597
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,16,2,power_law_1.01,0.11691519737243652
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,16,2,power_law_1.01,0.13575296401977538
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,16,2,power_law_1.01,0.14216320037841795
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,16,2,balanced,0.12451071739196777
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,16,2,power_law_1.01,0.1523750400543213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,16,2,power_law_1.01,0.1578611183166504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,16,2,power_law_1.01,0.175863037109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,16,2,power_law_1.01,0.17978368759155275
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,16,2,power_law_1.01,0.18635520935058594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,16,2,power_law_1.01,0.14636927604675293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,16,2,power_law_1.01,0.1498483180999756
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,16,2,power_law_1.01,0.1998297691345215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,16,2,power_law_1.01,0.20903039932250977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,16,2,power_law_1.01,0.2284326362609863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,16,2,power_law_1.01,0.25361536026000975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,16,2,power_law_1.01,0.18634111404418946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,16,2,power_law_1.01,0.2946892738342285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,16,2,power_law_1.01,0.3625574493408203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,16,2,power_law_1.01,0.45434112548828126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,16,2,power_law_1.01,0.546616325378418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,16,2,power_law_1.01,0.7410086059570313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,16,2,power_law_1.01,0.04786943912506104
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,16,2,power_law_1.01,0.9252275085449219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,16,2,power_law_1.01,0.05294847965240479
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,16,2,power_law_1.01,0.05950719833374023
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,16,2,power_law_1.01,0.06787712097167968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,16,2,power_law_1.01,0.0762227201461792
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,16,2,power_law_1.01,1.3142489624023437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,16,2,power_law_1.01,0.09211135864257812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,16,2,power_law_1.01,0.10418944358825684
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,16,2,power_law_1.01,0.10692864418029786
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,16,2,power_law_1.01,1.7183923339843752
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,16,2,power_law_1.01,0.10805760383605958
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,16,2,power_law_1.01,0.11088000297546388
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,16,2,power_law_1.01,0.1155020809173584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,16,2,power_law_1.01,0.1192576026916504
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,16,2,power_law_1.01,0.12133119583129884
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,16,2,power_law_1.01,0.12650495529174804
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,16,2,power_law_1.01,0.13277695655822755
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,16,2,power_law_1.01,0.14139904022216795
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,16,2,power_law_1.01,0.14941951751708984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,16,2,balanced,0.3844211196899414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,16,2,balanced,0.5502272033691407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,16,2,balanced,0.33244415283203127
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,16,2,balanced,0.7096806335449218
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,16,2,power_law_1.01,0.17602304458618162
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,16,2,power_law_1.01,0.20034559249877928
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,16,2,power_law_1.01,0.25272832870483397
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,16,2,power_law_1.2,0.03861504077911377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,16,2,power_law_1.2,0.055974397659301754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,16,2,power_law_1.2,0.07203968048095703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,16,2,power_law_1.2,0.08750847816467286
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,16,2,power_law_1.2,0.10803968429565429
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,16,2,power_law_1.2,0.12131967544555664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,16,2,power_law_1.01,0.28653951644897463
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,16,2,power_law_1.2,0.13647104263305665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,16,2,power_law_1.2,0.1379852771759033
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,16,2,power_law_1.01,0.4214438247680664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,16,2,power_law_1.2,0.1390079975128174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,16,2,power_law_1.2,0.14580991744995117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,16,2,power_law_1.2,0.15279616355895995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,16,2,power_law_1.2,0.23852928161621093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,16,2,power_law_1.01,0.7146047973632813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,16,2,power_law_1.2,0.24102912902832033
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,16,2,power_law_1.01,0.5186956787109376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,16,2,power_law_1.2,0.2501171112060547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,16,2,power_law_1.2,0.2549324798583984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,16,2,power_law_1.2,0.2669440078735351
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,16,2,power_law_1.2,0.2825856018066406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,16,2,power_law_1.2,0.3463411331176758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,16,2,power_law_1.2,0.3152563285827637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,16,2,power_law_1.2,0.41074558258056637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,16,2,power_law_1.2,0.49325313568115237
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,16,2,power_law_1.01,1.018238754272461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,16,2,power_law_1.2,0.6879283142089844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,16,2,power_law_1.2,0.816684799194336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,16,2,power_law_1.2,0.039118080139160155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,16,2,power_law_1.2,1.1324569702148437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,16,2,power_law_1.2,0.04567552089691162
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,16,2,power_law_1.2,0.058417921066284184
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,16,2,power_law_1.2,0.08069120407104492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,16,2,power_law_1.2,0.10595199584960938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,16,2,power_law_1.2,1.4395738220214844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,16,2,power_law_1.2,0.12271360397338867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,16,2,power_law_1.2,0.14067584037780761
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,16,2,power_law_1.2,0.14851584434509277
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,16,2,power_law_1.2,0.15090432167053222
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,16,2,power_law_1.2,0.15852031707763672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,16,2,power_law_1.2,0.1768320083618164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,16,2,power_law_1.2,0.18082048416137694
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,16,2,power_law_1.2,2.028738555908203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,16,2,power_law_1.2,0.18485887527465822
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,16,2,power_law_1.2,0.18677631378173828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,16,2,power_law_1.2,0.1996326446533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,16,2,power_law_1.2,0.21733888626098632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,16,2,power_law_1.2,0.2351910400390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,16,2,power_law_1.2,0.2636787223815918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,16,2,power_law_1.2,2.6952832031250002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,16,2,power_law_1.2,0.14354047775268555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,16,2,power_law_1.2,0.3308812713623047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,16,2,power_law_1.2,0.3764121627807617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,16,2,power_law_1.2,0.46762111663818356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,16,2,power_law_1.2,0.5556415939331055
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,16,2,power_law_1.2,0.04772480010986328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,16,2,power_law_1.2,0.7646822357177735
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,16,2,power_law_1.2,0.052773118019104004
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,16,2,power_law_1.2,0.057116160392761225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,16,2,power_law_1.2,0.9571033477783203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,16,2,power_law_1.2,0.06696959972381591
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,16,2,power_law_1.2,0.07459712028503418
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,16,2,power_law_1.2,0.07859072208404541
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,16,2,power_law_1.2,0.10266240119934082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,16,2,power_law_1.2,0.10610176086425782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,16,2,power_law_1.2,1.3470745849609376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,16,2,power_law_1.2,0.10773119926452637
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,16,2,power_law_1.2,0.11123583793640138
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,16,2,power_law_1.2,0.11574272155761718
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,16,2,power_law_1.2,0.12038016319274902
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,16,2,power_law_1.2,0.12485247611999513
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,16,2,power_law_1.2,0.13168895721435547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,16,2,power_law_1.2,0.1387775993347168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,16,2,power_law_1.2,1.7439334106445312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,16,2,power_law_1.2,0.15260543823242187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,16,2,power_law_1.2,0.24185087203979494
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,16,2,power_law_1.01,1.5378866577148438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,16,2,power_law_1.01,1.9931085205078127
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,16,2,power_law_1.2,0.1637183952331543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,16,2,power_law_1.2,0.20403072357177737
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,16,2,power_law_1.2,0.39961856842041016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,16,2,power_law_1.2,0.5676006317138672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,32,1,balanced,0.04345600128173828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,16,2,power_law_1.2,0.3226380920410156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,32,1,balanced,0.07803391933441162
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,32,1,balanced,0.09822400093078613
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,32,1,balanced,0.13553855895996095
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,16,2,power_law_1.2,0.745902099609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,32,1,balanced,0.20847423553466796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,32,1,balanced,0.20969343185424805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,32,1,balanced,0.21254528045654295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,32,1,balanced,0.21350080490112305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,32,1,balanced,0.21434879302978516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,32,1,balanced,0.2186137580871582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,32,1,balanced,0.054079360961914055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,32,1,balanced,0.2333363151550293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,32,1,balanced,0.23717119216918947
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,32,1,balanced,0.3581817626953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,32,1,balanced,0.3647046279907227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,32,1,balanced,0.22298879623413087
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,32,1,balanced,0.37982078552246096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,32,1,balanced,0.4077619171142578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,32,1,balanced,0.43354110717773436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,32,1,balanced,0.487006721496582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,32,1,balanced,0.545673599243164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,32,1,balanced,0.7267871856689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,32,1,balanced,0.8251513671874999
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,32,1,balanced,1.1486886596679686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,32,1,balanced,0.04633344173431396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,32,1,balanced,1.436373748779297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,32,1,balanced,0.054561281204223634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,32,1,balanced,0.06566336154937744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,32,1,balanced,0.09886783599853516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,32,1,balanced,0.15692352294921874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,32,1,balanced,0.27865472793579105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,32,1,balanced,0.2739833641052246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,32,1,balanced,0.27637567520141604
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,32,1,balanced,2.084062042236328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,32,1,balanced,0.2785964775085449
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,32,1,balanced,0.28081663131713863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,32,1,balanced,0.28548479080200195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,32,1,balanced,0.28944704055786136
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,32,1,balanced,0.2926905632019043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,32,1,balanced,0.3006118392944336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,32,1,balanced,0.352305908203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,32,1,balanced,2.8051904296875003
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,32,1,balanced,0.35753662109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,32,1,balanced,0.36494335174560544
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,32,1,balanced,0.37918270111083985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,32,1,balanced,0.3953011322021484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,32,1,balanced,0.4253561782836914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,32,1,balanced,0.4598028945922851
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,32,1,balanced,0.5246572875976563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,32,1,balanced,0.5863020706176758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,32,1,balanced,0.9458739471435547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,32,1,balanced,1.0776659393310548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,32,1,balanced,1.5690284729003907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,32,1,balanced,2.0539974975585937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,16,2,power_law_1.2,1.0929216003417968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,16,2,power_law_1.2,1.484145965576172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,16,2,power_law_1.2,2.4280601501464845
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,16,2,power_law_1.2,3.2056192016601566
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,32,1,balanced,0.07344319820404052
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,32,1,balanced,0.07652480125427245
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,32,1,balanced,0.08522815704345703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,32,1,balanced,0.2044300842285156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,32,1,balanced,0.20482816696166992
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,32,1,balanced,0.10323007583618164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,32,1,balanced,0.20604671478271483
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,32,1,balanced,0.20798528671264646
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,32,1,balanced,0.209737606048584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,32,1,balanced,0.21266304016113283
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,32,1,balanced,0.21459135055541992
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,32,1,balanced,0.2179961585998535
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,32,1,balanced,0.2232192039489746
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,32,1,balanced,0.24621183395385743
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,32,1,balanced,0.23460607528686522
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,32,1,balanced,0.26936128616333005
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,32,1,balanced,0.2941785621643066
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,32,1,balanced,0.34402111053466794
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,32,1,balanced,0.39311168670654295
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,32,1,balanced,0.13366080284118653
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,32,1,power_law_1.01,0.04351103782653808
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,32,1,power_law_1.01,0.04805439949035645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,32,1,power_law_1.01,0.0799622392654419
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,32,1,power_law_1.01,0.10026368141174316
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,32,1,balanced,0.2012614440917969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,32,1,power_law_1.01,0.12512639999389646
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,32,1,power_law_1.01,0.14842432022094726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,32,1,power_law_1.01,0.17512639999389648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,32,1,power_law_1.01,0.18325311660766602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,32,1,power_law_1.01,0.18831552505493163
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,32,1,power_law_1.01,0.18918783187866212
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,32,1,power_law_1.01,0.19860544204711914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,32,1,power_law_1.01,0.21259584426879882
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,32,1,power_law_1.01,0.2195871925354004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,32,1,power_law_1.01,0.2413350486755371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,32,1,power_law_1.01,0.3494432067871094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,32,1,power_law_1.01,0.3577753448486328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,32,1,power_law_1.01,0.37369537353515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,32,1,power_law_1.01,0.4073446273803711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,32,1,power_law_1.01,0.4387027359008789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,32,1,power_law_1.01,0.5034207916259765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,32,1,power_law_1.01,0.5663462448120117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,32,1,power_law_1.01,0.7406816101074218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,32,1,power_law_1.01,0.9136544036865235
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,32,1,balanced,0.2030476760864258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,32,1,power_law_1.01,1.210479965209961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,32,1,power_law_1.01,1.5284967041015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,32,1,balanced,0.20356096267700194
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,32,1,power_law_1.01,0.04763840198516846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,32,1,power_law_1.01,0.05419712066650391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,32,1,power_law_1.01,2.151263427734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,32,1,power_law_1.01,0.07470911979675293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,32,1,power_law_1.01,0.10107071876525879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,32,1,power_law_1.01,0.1443072032928467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,32,1,power_law_1.01,0.20105344772338868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,32,1,power_law_1.01,0.23142719268798828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,32,1,power_law_1.01,0.23826623916625977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,32,1,power_law_1.01,2.7836767578124997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,32,1,power_law_1.01,0.24331392288208006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,32,1,power_law_1.01,0.24530176162719725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,32,1,power_law_1.01,0.2550873565673828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,32,1,power_law_1.01,0.26270336151123047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,32,1,power_law_1.01,0.26683391571044923
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,32,1,power_law_1.01,0.2745305633544922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,32,1,power_law_1.01,0.32637630462646483
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,32,1,power_law_1.01,0.3310483169555664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,32,1,power_law_1.01,0.34392704010009767
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,32,1,power_law_1.01,0.367116813659668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,32,1,power_law_1.01,0.39646015167236326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,32,1,power_law_1.01,0.4526764678955078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,32,1,power_law_1.01,0.5240620803833008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,32,1,power_law_1.01,0.6330265426635743
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,32,1,power_law_1.01,0.7402617645263672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,32,1,power_law_1.01,0.9748480224609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,32,1,power_law_1.01,1.2228620910644532
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,32,1,power_law_1.01,0.07466303825378418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,32,1,power_law_1.01,1.7031513977050783
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,32,1,power_law_1.01,0.07716095924377442
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,32,1,power_law_1.01,2.1803654479980468
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,32,1,power_law_1.01,0.10239168167114258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,32,1,power_law_1.01,0.0877625560760498
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,32,1,power_law_1.01,0.11913663864135741
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,32,1,power_law_1.01,0.13503935813903808
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,32,1,power_law_1.01,0.16679231643676756
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,32,1,power_law_1.01,0.173253116607666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,32,1,power_law_1.01,0.1763974380493164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,32,1,power_law_1.01,0.1797587203979492
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,32,1,power_law_1.01,0.1884480094909668
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,32,1,power_law_1.01,0.19438016891479493
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,32,1,power_law_1.01,0.1995756721496582
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,32,1,balanced,0.6129433441162109
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,32,1,power_law_1.01,0.20743040084838865
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,32,1,power_law_1.01,0.22045888900756835
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,32,1,power_law_1.01,0.22757631301879883
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,32,1,power_law_1.01,0.24444416046142575
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,32,1,balanced,0.7118745422363281
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,32,1,power_law_1.01,0.28619775772094724
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,32,1,balanced,1.0342598724365235
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,32,1,balanced,1.34623291015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,32,1,power_law_1.01,0.325970573425293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,32,1,power_law_1.01,0.41522945404052736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,32,1,power_law_1.2,0.04353663921356201
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,32,1,power_law_1.01,0.5070495986938477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,32,1,power_law_1.2,0.047618560791015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,32,1,power_law_1.2,0.0793120002746582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,32,1,power_law_1.2,0.09489983558654785
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,32,1,power_law_1.01,0.6943360137939453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,32,1,power_law_1.2,0.11873151779174804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,32,1,power_law_1.2,0.1457478427886963
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,32,1,power_law_1.01,0.8887648010253907
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,32,1,power_law_1.01,1.2769996643066406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,32,1,power_law_1.2,0.18086719512939453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,32,1,power_law_1.2,0.17206399917602538
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,32,1,power_law_1.2,0.18610048294067383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,32,1,power_law_1.2,0.1907948875427246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,32,1,power_law_1.2,0.2129792022705078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,32,1,power_law_1.2,0.22376127243041993
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,32,1,power_law_1.01,1.6826751708984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,32,1,power_law_1.2,0.1983884811401367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,32,1,power_law_1.2,0.24602176666259767
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,32,1,power_law_1.2,0.3515039825439453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,32,1,power_law_1.2,0.3621907043457031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,32,1,power_law_1.2,0.3779302215576172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,32,1,power_law_1.2,0.44437374114990236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,32,1,power_law_1.2,0.4108601760864258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,32,1,power_law_1.2,0.5092505645751954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,32,1,power_law_1.2,0.5785113525390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,32,1,power_law_1.2,0.7485132598876953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,32,1,power_law_1.2,0.9232198333740234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,32,1,power_law_1.2,0.04636735916137695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,32,1,power_law_1.01,2.4752000427246097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,32,1,power_law_1.2,0.05457856178283691
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,32,1,power_law_1.2,1.217031021118164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,32,1,power_law_1.2,0.07323264122009278
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,32,1,power_law_1.2,0.09795519828796387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,32,1,power_law_1.2,1.5413977050781251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,32,1,power_law_1.2,0.13303104400634766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,32,1,power_law_1.2,0.19616384506225587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,32,1,power_law_1.2,0.2279372787475586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,32,1,power_law_1.2,0.2408518409729004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,32,1,power_law_1.2,0.23699840545654297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,32,1,power_law_1.2,0.24568639755249025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,32,1,power_law_1.2,0.25258623123168944
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,32,1,power_law_1.2,2.167376708984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,32,1,power_law_1.2,0.2601446342468262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,32,1,power_law_1.2,0.26679487228393556
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,32,1,power_law_1.2,0.2785196876525879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,32,1,power_law_1.2,0.3297689437866211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,32,1,power_law_1.2,0.33811134338378906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,32,1,power_law_1.2,0.3501715087890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,32,1,power_law_1.2,0.3808051300048828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,32,1,power_law_1.2,0.4085529708862305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,32,1,power_law_1.2,2.7980078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,32,1,power_law_1.2,0.48243457794189454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,32,1,power_law_1.2,0.5379750442504883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,32,1,power_law_1.2,0.6500876617431641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,32,1,power_law_1.2,0.7835174560546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,32,1,power_law_1.2,0.9878489685058593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,32,1,power_law_1.2,1.2387372589111327
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,32,1,power_law_1.2,0.07435776233673096
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,32,1,power_law_1.2,0.07747456073760986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,32,1,power_law_1.2,1.7207551574707032
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,32,1,power_law_1.2,0.0870342445373535
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,32,1,power_law_1.2,0.09979328155517578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,32,1,power_law_1.2,0.11574463844299317
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,32,1,power_law_1.2,2.1983815002441407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,32,1,power_law_1.2,0.13111616134643556
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,32,1,power_law_1.2,0.1646444892883301
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,32,1,power_law_1.2,0.17233919143676757
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,32,1,power_law_1.2,0.17553472518920898
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,32,1,power_law_1.2,0.18150976181030273
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,32,1,power_law_1.2,0.18912511825561523
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,32,1,power_law_1.2,0.19618560791015624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,32,1,power_law_1.2,0.20341888427734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,32,1,power_law_1.2,0.21843904495239258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,32,1,power_law_1.2,0.2298284721374512
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,32,1,power_law_1.2,0.24330560684204103
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,32,1,power_law_1.2,0.2757792091369629
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,32,1,power_law_1.01,3.425621643066406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,32,1,power_law_1.2,0.33465854644775395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,32,1,power_law_1.2,0.3991667175292969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,32,1,power_law_1.2,0.6649171447753905
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,32,1,power_law_1.2,0.5372063827514648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,32,balanced,0.03750911951065063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,32,balanced,0.04933248043060302
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,32,balanced,0.036599040031433105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,32,balanced,0.07074304103851319
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,32,balanced,0.08529536247253418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,32,balanced,0.12303487777709961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,32,balanced,0.16235263824462892
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,32,balanced,0.16496896743774414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,32,balanced,0.1651519966125488
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,32,1,power_law_1.2,0.9774732971191407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,32,balanced,0.1681216049194336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,32,balanced,0.16765184402465821
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,32,1,power_law_1.2,1.309015655517578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,32,balanced,0.17060863494873046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,32,balanced,0.17274112701416017
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,32,balanced,0.17874303817749024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,32,balanced,0.18076671600341795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,32,balanced,0.181212158203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,32,balanced,0.20043392181396485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,32,balanced,0.21639423370361327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,32,balanced,0.24059263229370118
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,32,balanced,0.38059391021728517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,32,balanced,0.3895999908447266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,32,balanced,0.18964351654052733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,32,balanced,0.5177305603027345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,32,balanced,0.6308761596679687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,32,balanced,0.24286336898803712
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,32,1,power_law_1.2,2.009771575927734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,32,balanced,0.03581696033477783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,32,balanced,0.03666815996170044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,32,balanced,0.036993279457092285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,32,balanced,0.05483007907867432
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,32,balanced,0.9112576293945314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,32,balanced,0.03912064075469971
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,32,balanced,0.07951615810394287
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,32,balanced,0.08712960243225097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,32,balanced,0.09121791839599609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,32,balanced,0.08813695907592774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,32,balanced,0.08837632179260255
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,32,balanced,0.08930303573608397
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,32,balanced,0.09032064437866211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,32,balanced,0.09213952064514161
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,32,balanced,0.09561471939086914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,32,balanced,0.09798015594482422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,32,balanced,0.10025728225708008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,32,balanced,0.09388287544250488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,32,balanced,0.10118528366088868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,32,balanced,0.1074892807006836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,32,balanced,0.1152128028869629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,32,balanced,0.12809727668762208
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,32,balanced,0.15622015953063967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,32,balanced,0.17595903396606444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,32,balanced,0.2505855941772461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,32,balanced,0.2944243240356445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,32,balanced,1.1715494537353517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,32,balanced,0.41074176788330075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,32,balanced,0.5352473449707031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,32,1,power_law_1.2,2.719360046386719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,1,32,balanced,0.021032960414886476
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,1,32,balanced,0.0483622407913208
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,1,32,balanced,0.02073983907699585
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,1,32,balanced,0.02145024061203003
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,1,32,balanced,0.030113279819488525
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,1,32,balanced,0.04833920001983642
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,1,32,balanced,0.04847104072570801
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,1,32,balanced,0.048198399543762205
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,1,32,balanced,0.02103935956954956
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,1,32,balanced,0.04828671932220459
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,1,32,balanced,0.048519678115844726
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,1,32,balanced,0.04874623775482177
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,1,32,balanced,0.04870143890380859
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,1,32,balanced,0.04902016162872315
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,1,32,balanced,0.04933631896972656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,1,32,balanced,0.04899968147277832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,1,32,balanced,0.049571838378906254
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,1,32,balanced,0.04936575889587402
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,1,32,balanced,0.049688320159912105
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,1,32,balanced,0.05173120021820068
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,1,32,balanced,0.052225279808044436
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,1,32,balanced,0.05402368068695068
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,32,power_law_1.01,0.12132608413696289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,32,power_law_1.01,0.12092415809631347
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,32,power_law_1.01,0.1230399990081787
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,32,power_law_1.01,0.1224563217163086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,32,power_law_1.01,0.15730560302734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,32,power_law_1.01,0.15285887718200683
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,32,power_law_1.01,0.15464320182800292
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,32,power_law_1.01,0.16162559509277344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,32,power_law_1.01,0.16311168670654297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,32,power_law_1.01,0.1671449661254883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,32,power_law_1.01,0.1719820785522461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,32,power_law_1.01,0.17410303115844727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,32,power_law_1.01,0.17503488540649415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,32,power_law_1.01,0.18715007781982423
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,32,power_law_1.01,0.181844482421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,32,power_law_1.01,0.18409088134765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,32,power_law_1.01,0.21498367309570315
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,32,power_law_1.01,0.2304524803161621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,32,power_law_1.01,0.3113715171813965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,32,power_law_1.01,0.40357246398925783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,32,power_law_1.01,0.44708225250244144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,32,power_law_1.01,0.6100172805786133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,32,power_law_1.01,0.821802215576172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,32,power_law_1.01,1.034150390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,1,32,balanced,0.06362112045288086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,32,power_law_1.01,1.5407359313964843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,32,power_law_1.01,2.0534375000000002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,32,power_law_1.01,0.07833983898162841
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,32,power_law_1.01,0.07973375797271728
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,32,power_law_1.01,2.574065856933594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,32,power_law_1.01,0.07972735881805419
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,32,power_law_1.01,0.07926015853881836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,32,power_law_1.01,0.07890304088592529
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,32,power_law_1.01,0.0844979190826416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,32,power_law_1.01,0.08570752143859864
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,32,power_law_1.01,0.0918489646911621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,32,power_law_1.01,0.08765312194824218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,32,power_law_1.01,0.09136896133422852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,32,power_law_1.01,0.09168767929077148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,32,power_law_1.01,0.09158271789550781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,32,power_law_1.01,0.09330559730529785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,32,power_law_1.01,0.10155136108398437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,32,power_law_1.01,0.10288000106811523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,32,power_law_1.01,0.09691776275634766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,32,power_law_1.01,0.10502528190612792
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,32,power_law_1.01,0.11119487762451172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,32,power_law_1.01,0.14493439674377442
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,32,power_law_1.01,0.19135103225708008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,32,power_law_1.01,0.21335935592651367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,32,power_law_1.01,0.2606028747558594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,32,power_law_1.01,0.32673408508300783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,32,power_law_1.01,0.4315315246582031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,32,power_law_1.01,0.6199744033813477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,32,power_law_1.01,0.8292774200439453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,1,32,power_law_1.01,0.04850175857543945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,32,power_law_1.01,1.0545152282714843
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,1,32,power_law_1.01,0.04794623851776123
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,1,32,power_law_1.01,0.04786816120147705
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,1,32,balanced,0.0754636812210083
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,1,32,power_law_1.01,0.0453107213973999
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,1,32,power_law_1.01,0.0427788782119751
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,1,32,power_law_1.01,0.046045441627502445
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,1,32,power_law_1.01,0.046465277671813965
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,1,32,power_law_1.01,0.046577920913696294
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,1,32,balanced,0.07861760139465332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,1,32,power_law_1.01,0.04803455829620361
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,1,32,balanced,0.12059264183044434
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,1,32,power_law_1.01,0.04772736072540283
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,1,32,power_law_1.01,0.04882304191589355
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,1,32,power_law_1.01,0.047470078468322754
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,1,32,power_law_1.01,0.047165441513061526
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,1,32,balanced,0.10060416221618654
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,1,32,power_law_1.01,0.04650496006011963
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,1,32,power_law_1.01,0.05017471790313721
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,1,32,power_law_1.01,0.04822527885437012
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,1,32,power_law_1.01,0.04900864124298095
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,1,32,power_law_1.01,0.06845312118530274
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,1,32,power_law_1.01,0.05641344070434571
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,1,32,power_law_1.01,0.07066368103027344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,1,32,power_law_1.01,0.05257599830627442
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,1,32,power_law_1.01,0.09192447662353516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,32,power_law_1.2,0.11994239807128906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,32,power_law_1.2,0.11973888397216796
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,1,32,power_law_1.01,0.11378047943115235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,32,power_law_1.2,0.12171775817871093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,32,power_law_1.2,0.122161922454834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,32,power_law_1.2,0.15739520072937013
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,32,power_law_1.2,0.14788607597351072
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,32,power_law_1.2,0.16089216232299805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,32,power_law_1.2,0.16017919540405273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,32,power_law_1.2,0.1646771240234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,32,power_law_1.2,0.16175104141235352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,32,power_law_1.2,0.1690457534790039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,32,power_law_1.2,0.16915456771850584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,32,power_law_1.2,0.1696793556213379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,32,power_law_1.2,0.1833241653442383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,32,power_law_1.2,0.1834611129760742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,32,power_law_1.2,0.2040115165710449
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,32,power_law_1.2,0.22579328536987306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,32,power_law_1.2,0.2872140884399414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,32,power_law_1.2,0.3230809783935547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,32,power_law_1.2,0.43077247619628906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,32,power_law_1.2,0.4895283126831055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,32,power_law_1.2,0.6882137298583985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,32,power_law_1.2,0.918951644897461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,32,power_law_1.2,1.4066047668457031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,32,power_law_1.2,1.892737274169922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,32,power_law_1.2,3.1162060546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,32,power_law_1.2,0.07811200141906738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,32,power_law_1.2,0.07881984233856201
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,32,power_law_1.2,0.07933695793151856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,32,power_law_1.2,0.07959807872772216
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,32,power_law_1.2,4.064288024902344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,32,power_law_1.2,0.0693068790435791
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,32,power_law_1.2,0.08291071891784668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,32,power_law_1.2,0.08301183700561524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,32,power_law_1.2,0.07754111766815186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,32,power_law_1.2,0.08685952186584472
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,32,power_law_1.2,0.08985856056213379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,32,power_law_1.2,0.09107071876525878
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,32,power_law_1.2,0.09738240242004395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,32,power_law_1.2,0.09903743743896484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,32,power_law_1.2,0.09729536056518555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,32,power_law_1.2,0.10126720428466798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,32,power_law_1.2,0.10308223724365236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,32,power_law_1.2,0.10510975837707519
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,32,power_law_1.2,0.13494784355163575
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,32,power_law_1.2,0.1596671962738037
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,32,power_law_1.2,0.20971263885498045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,32,power_law_1.2,0.2419571113586426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,32,power_law_1.2,0.3251443099975586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,32,power_law_1.2,0.39679615020751957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,32,power_law_1.2,0.6374489593505859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,32,power_law_1.2,0.7411980438232422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,32,power_law_1.2,1.1906905364990235
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,1,32,power_law_1.2,0.04810880184173584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,1,32,power_law_1.2,0.04822400093078613
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,32,power_law_1.2,1.7695462036132814
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,1,32,power_law_1.2,0.048508157730102544
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,1,32,power_law_1.2,0.043278079032897945
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,1,32,power_law_1.2,0.03976063966751099
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,1,32,power_law_1.2,0.045512962341308597
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,1,32,power_law_1.2,0.04434688091278076
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,1,32,power_law_1.2,0.04818431854248047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,1,32,power_law_1.2,0.04393983840942382
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,1,32,power_law_1.2,0.04443647861480713
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,1,32,power_law_1.2,0.044519681930541996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,1,32,power_law_1.2,0.04466303825378418
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,1,32,power_law_1.2,0.04595071792602539
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,1,32,power_law_1.2,0.04586880207061768
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,1,32,power_law_1.2,0.04616831779479981
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,1,32,power_law_1.2,0.048866558074951175
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,1,32,power_law_1.01,0.2784537506103516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,1,32,power_law_1.2,0.06561920166015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,1,32,power_law_1.01,0.14821632385253908
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,1,32,power_law_1.01,0.19314048767089845
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,1,32,power_law_1.01,0.3706099319458008
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,1,32,power_law_1.2,0.05111167907714844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,1,32,power_law_1.2,0.06015999794006348
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,1,32,power_law_1.2,0.07803391933441162
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,1,32,power_law_1.2,0.08631296157836914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,2,16,balanced,0.03526911973953247
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,2,16,balanced,0.032943360805511475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,2,16,balanced,0.03501568078994751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,2,16,balanced,0.06615295886993408
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,2,16,balanced,0.08535167694091797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,1,32,power_law_1.2,0.1351116752624512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,2,16,balanced,0.12741503715515137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,2,16,balanced,0.17346176147460937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,1,32,power_law_1.2,0.10659584045410156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,2,16,balanced,0.16632192611694335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,2,16,balanced,0.1715225601196289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,2,16,balanced,0.17058303833007812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,2,16,balanced,0.17437183380126953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,2,16,balanced,0.1783500862121582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,2,16,balanced,0.17535488128662108
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,2,16,balanced,0.1817740821838379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,2,16,balanced,0.18555263519287107
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,2,16,balanced,0.19458431243896485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,2,16,balanced,0.2095244789123535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,2,16,balanced,0.22461183547973632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,2,16,balanced,0.25679487228393555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,2,16,balanced,0.2871667289733887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,2,16,balanced,0.3978713607788086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,2,16,balanced,0.18348800659179687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,2,16,balanced,0.6386713409423829
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,2,16,balanced,0.8102336120605468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,2,16,balanced,1.1709030151367188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,2,16,balanced,0.47243518829345704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,2,16,balanced,0.0333951997756958
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,2,16,balanced,0.03418879985809326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,2,16,balanced,0.03504767894744873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,2,16,balanced,0.03956608057022094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,2,16,balanced,0.05861120223999024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,2,16,balanced,0.08522368431091308
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,2,16,balanced,0.09282943725585938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,2,16,balanced,0.09675007820129394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,2,16,balanced,1.5279833984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,2,16,balanced,0.09280384063720704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,2,16,balanced,0.0926796817779541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,2,16,balanced,0.09528575897216797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,2,16,balanced,0.09847423553466797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,2,16,balanced,0.10225919723510742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,2,16,balanced,0.09348352432250977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,2,16,balanced,0.09650431632995606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,2,16,balanced,0.10042752265930174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,2,16,balanced,0.1059455966949463
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,2,16,balanced,0.11173760414123535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,2,16,balanced,0.1257536029815674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,2,16,balanced,0.14410112380981446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,2,16,balanced,0.16474239349365233
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,2,16,balanced,0.20307968139648436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,2,16,balanced,0.23919616699218751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,2,16,balanced,0.348716812133789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,2,16,balanced,0.42262142181396484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,2,16,balanced,0.5966259384155274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,2,16,balanced,0.783587875366211
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,1,32,power_law_1.2,0.19988224029541016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,1,32,power_law_1.2,0.2437504005432129
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,1,32,power_law_1.2,0.4208729553222656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,1,32,power_law_1.2,0.5495129776000977
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,2,16,balanced,0.02082047939300537
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,2,16,balanced,0.05082367897033692
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,2,16,balanced,0.02108288049697876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,2,16,balanced,0.021675519943237305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,2,16,balanced,0.05071872234344482
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,2,16,balanced,0.05088640213012695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,2,16,balanced,0.05157504081726074
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,2,16,balanced,0.051375360488891596
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,2,16,balanced,0.05115520000457764
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,2,16,balanced,0.024158720970153806
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,2,16,balanced,0.05170432090759277
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,2,16,balanced,0.05154304027557373
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,2,16,balanced,0.05199488162994385
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,2,16,balanced,0.05385087966918946
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,2,16,balanced,0.05380735874176026
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,2,16,balanced,0.0560038423538208
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,2,16,balanced,0.058699522018432614
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,2,16,balanced,0.06139520168304443
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,2,16,balanced,0.05181056022644043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,2,16,balanced,0.03023103952407837
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,2,16,balanced,0.04927872180938721
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,2,16,balanced,0.05108736038208008
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,2,16,balanced,0.051146240234374994
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,2,16,power_law_1.01,0.12152959823608397
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,2,16,power_law_1.01,0.12164608001708985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,2,16,power_law_1.01,0.1199295997619629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,2,16,power_law_1.01,0.16061952590942383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,2,16,power_law_1.01,0.16378751754760743
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,2,16,power_law_1.01,0.16366975784301757
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,2,16,power_law_1.01,0.16603647232055666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,2,16,power_law_1.01,0.16402559280395507
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,2,16,power_law_1.01,0.07958655834197999
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,2,16,power_law_1.01,0.17541120529174806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,2,16,power_law_1.01,0.174520320892334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,2,16,power_law_1.01,0.18518911361694337
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,2,16,power_law_1.01,0.17957120895385742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,2,16,power_law_1.01,0.18820224761962892
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,2,16,power_law_1.01,0.19436288833618165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,2,16,power_law_1.01,0.2185369682312012
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,2,16,power_law_1.01,0.17090431213378907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,2,16,power_law_1.01,0.2387648010253906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,2,16,power_law_1.01,0.32795646667480466
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,2,16,power_law_1.01,0.41491329193115234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,2,16,power_law_1.01,0.5141247940063477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,2,16,power_law_1.01,0.6907545471191406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,2,16,power_law_1.01,0.9212172698974609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,2,16,power_law_1.01,1.2399462127685548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,2,16,power_law_1.01,0.05713024139404297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,2,16,power_law_1.01,0.07878015995025635
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,2,16,power_law_1.01,1.8405030822753907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,2,16,power_law_1.01,0.07943039894104004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,2,16,power_law_1.01,0.08243840217590333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,2,16,power_law_1.01,0.08393343925476074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,2,16,power_law_1.01,2.5399090576171877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,2,16,power_law_1.01,0.0884825611114502
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,2,16,power_law_1.01,0.08617216110229492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,2,16,power_law_1.01,0.08081024169921876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,2,16,power_law_1.01,0.08578304290771485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,2,16,power_law_1.01,0.09056896209716797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,2,16,power_law_1.01,0.09207296371459961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,2,16,power_law_1.01,0.09718144416809082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,2,16,power_law_1.01,0.12194047927856447
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,2,16,power_law_1.01,0.10021120071411134
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,2,16,power_law_1.01,0.09372799873352052
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,2,16,power_law_1.01,0.0969382381439209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,2,16,power_law_1.01,0.10214400291442871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,2,16,power_law_1.01,0.1005452823638916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,2,16,power_law_1.01,0.11847935676574708
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,2,16,power_law_1.01,0.14077312469482423
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,2,16,power_law_1.01,0.17767423629760742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,2,16,power_law_1.01,0.19206655502319336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,2,16,power_law_1.01,0.2632064056396485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,2,16,power_law_1.01,0.3387366485595703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,2,16,power_law_1.01,0.4737241744995117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,2,16,power_law_1.01,0.6211840057373047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,2,16,power_law_1.01,0.03258879899978638
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,2,16,power_law_1.01,0.8448960113525391
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,2,16,power_law_1.01,0.05071743965148926
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,2,16,power_law_1.01,0.050943999290466314
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,2,16,power_law_1.01,0.05059840202331543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,2,16,power_law_1.01,1.1533734130859377
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,2,16,power_law_1.01,0.050405120849609374
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,2,16,power_law_1.01,0.04566271781921387
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,2,16,power_law_1.01,0.048903679847717284
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,2,16,power_law_1.01,0.047957758903503414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,2,16,power_law_1.01,0.047609601020812985
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,2,16,balanced,0.0908556842803955
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,2,16,balanced,0.08608384132385254
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,2,16,power_law_1.01,0.047298560142517085
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,2,16,power_law_1.01,0.047513599395751956
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,2,16,power_law_1.01,0.04778880119323731
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,2,16,power_law_1.01,0.04877056121826172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,2,16,power_law_1.01,0.05107967853546143
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,2,16,power_law_1.01,0.052761597633361815
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,2,16,balanced,0.11932928085327148
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,2,16,balanced,0.1456716823577881
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,2,16,power_law_1.01,0.049489917755126955
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,2,16,power_law_1.01,0.051423997879028314
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,2,16,power_law_1.01,0.058617601394653315
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,2,16,power_law_1.01,0.06295807838439942
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,2,16,power_law_1.01,0.07573631763458252
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,2,16,power_law_1.01,0.07987711906433106
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,2,16,power_law_1.2,0.08209024429321289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,2,16,power_law_1.2,0.11850751876831053
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,2,16,power_law_1.01,0.20235263824462893
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,2,16,power_law_1.2,0.10611328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,2,16,power_law_1.01,0.09922687530517578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,2,16,power_law_1.2,0.1222374439239502
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,2,16,power_law_1.01,0.12069760322570802
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,2,16,power_law_1.2,0.15797120094299316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,2,16,power_law_1.2,0.15920639991760255
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,2,16,power_law_1.2,0.16417407989501953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,2,16,power_law_1.2,0.1679257583618164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,2,16,power_law_1.01,0.16687231063842772
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,2,16,power_law_1.2,0.16729984283447266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,2,16,power_law_1.2,0.1754240036010742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,2,16,power_law_1.2,0.17910272598266602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,2,16,power_law_1.2,0.1805286407470703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,2,16,power_law_1.2,0.17596160888671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,2,16,power_law_1.2,0.1829849624633789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,2,16,power_law_1.2,0.18803327560424804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,2,16,power_law_1.2,0.20426624298095702
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,2,16,power_law_1.2,0.12759296417236327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,2,16,power_law_1.2,0.2245849609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,2,16,power_law_1.2,0.2834252738952637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,2,16,power_law_1.2,0.3724812698364258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,2,16,power_law_1.2,0.441082878112793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,2,16,power_law_1.2,0.6146355056762696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,2,16,power_law_1.2,0.7920140838623047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,2,16,power_law_1.2,0.05638400077819824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,2,16,power_law_1.2,0.07498623847961425
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,2,16,power_law_1.2,0.07778175830841064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,2,16,power_law_1.2,1.1708338928222655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,2,16,power_law_1.2,0.07726208209991456
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,2,16,power_law_1.2,0.08220928192138673
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,2,16,power_law_1.2,0.08832768440246583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,2,16,power_law_1.2,1.4922023010253906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,2,16,power_law_1.2,0.08457216262817382
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,2,16,power_law_1.2,0.08902655601501465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,2,16,power_law_1.2,0.08898048400878907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,2,16,power_law_1.2,0.09462400436401366
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,2,16,power_law_1.2,0.0989247989654541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,2,16,power_law_1.2,0.10099072456359863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,2,16,power_law_1.2,2.228491516113281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,2,16,power_law_1.2,0.09921024322509767
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,2,16,power_law_1.2,0.0970905590057373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,2,16,power_law_1.2,0.10133248329162599
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,2,16,power_law_1.2,0.10905344009399413
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,2,16,power_law_1.2,0.12216959953308107
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,2,16,power_law_1.2,0.1520691204071045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,2,16,power_law_1.2,2.7233203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,2,16,power_law_1.2,0.1950489616394043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,2,16,power_law_1.2,0.24297727584838866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,2,16,power_law_1.2,0.3029529571533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,2,16,power_law_1.2,0.43114879608154294
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,2,16,power_law_1.2,0.5495769500732421
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,2,16,power_law_1.2,0.032412159442901614
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,2,16,power_law_1.2,0.7000768280029297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,2,16,power_law_1.2,0.04991360187530518
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,2,16,power_law_1.2,0.047642879486083985
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,2,16,power_law_1.2,0.049582080841064455
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,2,16,power_law_1.2,0.050170879364013675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,2,16,power_law_1.2,1.0697216033935546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,2,16,power_law_1.2,0.08793855667114257
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,2,16,power_law_1.2,0.047551999092102046
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,2,16,power_law_1.2,0.04849279880523681
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,2,16,power_law_1.2,0.04943871974945068
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,2,16,power_law_1.2,1.5448204040527345
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,2,16,power_law_1.2,0.048037118911743164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,2,16,power_law_1.2,0.048533759117126464
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,2,16,power_law_1.2,0.05009280204772949
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,2,16,power_law_1.2,0.05074304103851318
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,2,16,power_law_1.2,0.052391681671142575
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,2,16,power_law_1.2,0.03752063989639282
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,2,16,power_law_1.2,0.04720640182495117
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,2,16,power_law_1.2,0.049422078132629395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,2,16,power_law_1.01,0.2875916862487793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,2,16,power_law_1.01,0.4036019134521484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,2,16,power_law_1.2,0.0690124797821045
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,2,16,power_law_1.2,0.05626751899719238
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,2,16,power_law_1.2,0.06399871826171874
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,2,16,power_law_1.2,0.0871615982055664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,4,8,balanced,0.03171583890914917
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,4,8,balanced,0.03304575920104981
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,4,8,balanced,0.0557478380203247
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,2,16,power_law_1.2,0.09497856140136719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,4,8,balanced,0.0680460786819458
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,4,8,balanced,0.08605952262878418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,4,8,balanced,0.13238143920898438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,4,8,balanced,0.16525056838989258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,4,8,balanced,0.17059455871582033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,4,8,balanced,0.1731532859802246
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,2,16,power_law_1.2,0.11845503807067873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,4,8,balanced,0.17762943267822268
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,4,8,balanced,0.1707097625732422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,4,8,balanced,0.17594751358032226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,4,8,balanced,0.17879039764404298
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,4,8,balanced,0.18665088653564452
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,4,8,balanced,0.19116416931152344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,4,8,balanced,0.1990336036682129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,4,8,balanced,0.21241727828979493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,4,8,balanced,0.24245632171630857
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,4,8,balanced,0.2667635154724121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,4,8,balanced,0.3218035125732422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,4,8,balanced,0.38050304412841796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,4,8,balanced,0.5384371185302734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,4,8,balanced,0.6605401611328124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,4,8,balanced,0.9170880126953126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,4,8,balanced,1.1795021057128907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,4,8,balanced,0.03299200057983399
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,4,8,balanced,0.03499392032623291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,4,8,balanced,0.041173758506774905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,4,8,balanced,1.7286834716796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,4,8,balanced,0.0485427188873291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,4,8,balanced,0.06526976108551026
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,4,8,balanced,0.09725055694580079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,4,8,balanced,0.10504960060119628
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,4,8,balanced,0.10643839836120605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,4,8,balanced,0.11030015945434571
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,4,8,balanced,0.11315711975097656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,4,8,balanced,2.2827186584472656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,4,8,balanced,0.10652159690856933
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,4,8,balanced,0.10854528427124024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,4,8,balanced,0.11357312202453614
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,4,8,balanced,0.11510016441345214
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,4,8,balanced,0.11945088386535645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,4,8,balanced,0.1309657573699951
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,4,8,balanced,0.14779775619506835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,4,8,balanced,0.16673791885375977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,4,8,balanced,0.1999513626098633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,4,8,balanced,0.10391296386718749
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,4,8,balanced,0.23436031341552735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,4,8,balanced,0.3058252716064453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,4,8,balanced,0.3734579086303711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,4,8,balanced,0.5408959960937499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,4,8,balanced,0.6748057556152344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,4,8,balanced,0.9805414581298828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,4,8,balanced,1.2863795471191406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,2,16,power_law_1.2,0.16279935836791992
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,2,16,power_law_1.2,0.21629568099975588
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,2,16,power_law_1.2,0.2875993537902832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,2,16,power_law_1.2,0.4010150527954101
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,2,16,power_law_1.2,0.6466944122314453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,4,8,balanced,0.023128321170806883
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,4,8,balanced,0.057740797996521
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,4,8,balanced,0.023673601150512695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,4,8,balanced,0.058118400573730467
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,4,8,balanced,0.058456320762634274
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,4,8,balanced,0.05862656116485596
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,4,8,balanced,0.05855616092681885
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,4,8,balanced,0.05876607894897461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,4,8,balanced,0.05935359954833984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,4,8,balanced,0.05907072067260742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,4,8,balanced,0.06043392181396484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,4,8,balanced,0.061710081100463866
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,4,8,balanced,0.06420735836029053
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,4,8,balanced,0.024700160026550295
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,4,8,balanced,0.06606847763061523
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,4,8,balanced,0.06911744117736816
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,4,8,balanced,0.07416704177856445
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,4,8,balanced,0.07883008003234863
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,4,8,balanced,0.0268505597114563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,4,8,balanced,0.03779072046279907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,4,8,power_law_1.01,0.0658790397644043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,4,8,power_law_1.01,0.08343680381774903
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,4,8,power_law_1.01,0.08554880142211914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,4,8,power_law_1.01,0.10284159660339356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,4,8,power_law_1.01,0.11431936264038085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,4,8,power_law_1.01,0.1117363166809082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,4,8,power_law_1.01,0.15555583953857421
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,4,8,power_law_1.01,0.15923711776733399
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,4,8,power_law_1.01,0.16367488861083984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,4,8,balanced,0.05761919975280762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,4,8,power_law_1.01,0.1669798469543457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,4,8,power_law_1.01,0.17124351501464843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,4,8,power_law_1.01,0.1723366355895996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,4,8,power_law_1.01,0.1687539291381836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,4,8,power_law_1.01,0.1729075241088867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,4,8,power_law_1.01,0.17759744644165038
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,4,8,power_law_1.01,0.1796236801147461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,4,8,power_law_1.01,0.1957094383239746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,4,8,power_law_1.01,0.22785791397094726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,4,8,power_law_1.01,0.2545023918151855
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,4,8,balanced,0.057519359588623045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,4,8,power_law_1.01,0.3121766471862793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,4,8,power_law_1.01,0.3754073715209961
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,4,8,balanced,0.058172159194946296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,4,8,power_law_1.01,0.5185343933105468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,4,8,power_law_1.01,0.65670654296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,4,8,power_law_1.01,0.9275430297851562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,4,8,power_law_1.01,1.1525593566894532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,4,8,power_law_1.01,0.06012544155120849
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,4,8,power_law_1.01,0.03921024084091186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,4,8,power_law_1.01,1.718085174560547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,4,8,power_law_1.01,0.06852992057800293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,4,8,power_law_1.01,0.07497344017028809
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,4,8,power_law_1.01,0.08373503684997559
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,4,8,power_law_1.01,0.09061632156372071
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,4,8,power_law_1.01,0.09602304458618163
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,4,8,power_law_1.01,2.2433164978027342
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,4,8,power_law_1.01,0.09558143615722656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,4,8,power_law_1.01,0.09790080070495606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,4,8,power_law_1.01,0.10970111846923829
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,4,8,power_law_1.01,0.1061747169494629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,4,8,power_law_1.01,0.10129407882690429
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,4,8,power_law_1.01,0.1039743995666504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,4,8,power_law_1.01,0.10566783905029298
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,4,8,power_law_1.01,0.10701696395874023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,4,8,power_law_1.01,0.11303168296813966
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,4,8,power_law_1.01,0.13257216453552245
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,4,8,power_law_1.01,0.14116991996765135
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,4,8,power_law_1.01,0.0937343978881836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,4,8,power_law_1.01,0.19148544311523436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,4,8,power_law_1.01,0.22845695495605467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,4,8,power_law_1.01,0.2917939186096191
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,4,8,power_law_1.01,0.38516223907470704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,4,8,power_law_1.01,0.508144645690918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,4,8,power_law_1.01,0.6336012649536132
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,4,8,power_law_1.01,0.027304959297180176
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,4,8,power_law_1.01,0.039307520389556885
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,4,8,power_law_1.01,0.045416960716247554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,4,8,power_law_1.01,0.8960781097412109
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,4,8,power_law_1.01,0.04861824035644531
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,4,8,power_law_1.01,0.054424319267272946
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,4,8,power_law_1.01,0.05658239841461181
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,4,8,power_law_1.01,0.0532966423034668
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,4,8,power_law_1.01,0.05260416030883789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,4,8,power_law_1.01,1.2537586975097657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,4,8,power_law_1.01,0.05304319858551025
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,4,8,power_law_1.01,0.0545139217376709
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,4,8,power_law_1.01,0.05543424129486084
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,4,8,power_law_1.01,0.05680895805358886
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,4,8,power_law_1.01,0.05648128032684326
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,4,8,power_law_1.01,0.056942081451416014
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,4,8,power_law_1.01,0.05908991813659668
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,4,8,power_law_1.01,0.060348157882690434
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,4,8,power_law_1.01,0.06304512023925782
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,4,8,balanced,0.11305472373962402
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,4,8,balanced,0.16882944107055664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,4,8,balanced,0.12223487854003907
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,4,8,balanced,0.21182207107543943
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,4,8,power_law_1.01,0.08859264373779296
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,4,8,power_law_1.01,0.06972032070159913
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,4,8,power_law_1.01,0.10270591735839844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,4,8,power_law_1.2,0.06560256004333495
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,4,8,power_law_1.2,0.08297087669372558
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,4,8,power_law_1.2,0.07791615962982178
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,4,8,power_law_1.01,0.07710464000701904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,4,8,power_law_1.2,0.09847423553466797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,4,8,power_law_1.2,0.11693440437316896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,4,8,power_law_1.2,0.12028672218322753
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,4,8,power_law_1.2,0.15486080169677735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,4,8,power_law_1.2,0.16247552871704102
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,4,8,power_law_1.01,0.14086015701293947
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,4,8,power_law_1.2,0.16252159118652343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,4,8,power_law_1.2,0.1657676887512207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,4,8,power_law_1.2,0.1705344009399414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,4,8,power_law_1.2,0.1700441551208496
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,4,8,power_law_1.01,0.15720831871032714
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,4,8,power_law_1.2,0.17528064727783205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,4,8,power_law_1.2,0.1769843292236328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,4,8,power_law_1.2,0.18467967987060546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,4,8,power_law_1.2,0.19776512145996095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,4,8,power_law_1.2,0.235150089263916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,4,8,power_law_1.2,0.2650227165222168
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,4,8,power_law_1.01,0.23599359512329104
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,4,8,power_law_1.2,0.3510950469970703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,4,8,power_law_1.2,0.1677836799621582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,4,8,power_law_1.2,0.4478720092773438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,4,8,power_law_1.2,0.562026252746582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,4,8,power_law_1.2,0.7314739227294922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,4,8,power_law_1.2,1.0110924530029297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,4,8,power_law_1.2,1.1994509124755859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,4,8,power_law_1.2,0.03986943960189819
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,4,8,power_law_1.2,0.06105984210968017
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,4,8,power_law_1.2,0.06379263877868653
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,4,8,power_law_1.2,0.07048575878143311
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,4,8,power_law_1.2,0.08528127670288085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,4,8,power_law_1.2,1.8571200561523438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,4,8,power_law_1.2,0.08465279579162598
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,4,8,power_law_1.2,0.09691391944885254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,4,8,power_law_1.2,0.09888256072998047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,4,8,power_law_1.2,0.09876607894897461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,4,8,power_law_1.2,0.10140800476074219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,4,8,power_law_1.2,0.10845184326171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,4,8,power_law_1.2,0.10585856437683105
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,4,8,power_law_1.01,0.2892966461181641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,4,8,power_law_1.2,0.1011673641204834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,4,8,power_law_1.2,0.10683008193969727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,4,8,power_law_1.2,0.10596991539001466
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,4,8,power_law_1.2,0.10741632461547852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,4,8,power_law_1.2,2.528875579833984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,4,8,power_law_1.2,0.13869824409484863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,4,8,power_law_1.2,0.16910079956054686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,4,8,power_law_1.2,0.19797119140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,4,8,power_law_1.2,0.2446233558654785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,4,8,power_law_1.2,0.32811904907226563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,4,8,power_law_1.2,0.41636478424072265
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,4,8,power_law_1.2,0.026355199813842774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,4,8,power_law_1.2,0.7362086486816406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,4,8,power_law_1.2,0.037607679367065425
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,4,8,power_law_1.2,0.11628031730651855
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,4,8,power_law_1.2,0.04108416080474854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,4,8,power_law_1.2,1.0499571228027345
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,4,8,power_law_1.2,0.05441023826599121
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,4,8,power_law_1.2,0.048117761611938474
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,4,8,power_law_1.2,0.05255807876586914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,4,8,power_law_1.2,0.05472640037536621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,4,8,power_law_1.2,1.37876220703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,4,8,power_law_1.2,0.053580799102783204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,4,8,power_law_1.2,0.5304140853881836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,4,8,power_law_1.2,0.053374719619750974
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,4,8,power_law_1.2,0.0557696008682251
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,4,8,power_law_1.2,0.05656320095062256
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,4,8,power_law_1.2,0.05830016136169434
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,4,8,power_law_1.2,0.05958784103393555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,4,8,power_law_1.2,0.06074368000030518
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,4,8,power_law_1.2,0.06261760234832764
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,4,8,power_law_1.2,0.0673804807662964
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,4,8,power_law_1.2,0.0476364803314209
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,4,8,power_law_1.2,0.09801216125488281
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,4,8,power_law_1.01,0.3889548873901367
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,4,8,power_law_1.01,0.5524492645263672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,4,8,power_law_1.2,0.07442431926727296
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,4,8,power_law_1.2,0.09224575996398926
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,4,8,power_law_1.2,0.1156710433959961
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,4,8,power_law_1.2,0.3267302322387695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,8,4,balanced,0.03134975910186767
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,4,8,power_law_1.2,0.17874431610107422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,8,4,balanced,0.03871360063552857
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,8,4,balanced,0.05477503776550293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,8,4,balanced,0.07064064025878906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,8,4,balanced,0.08678144454956055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,8,4,balanced,0.13440511703491212
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,8,4,balanced,0.1291660785675049
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,8,4,balanced,0.1297651195526123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,8,4,balanced,0.17196159362792968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,4,8,power_law_1.2,0.21816959381103515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,8,4,balanced,0.17397247314453126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,8,4,balanced,0.1778598403930664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,8,4,balanced,0.18084991455078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,8,4,balanced,0.1862835121154785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,8,4,balanced,0.1883852767944336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,8,4,balanced,0.19894655227661134
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,8,4,balanced,0.21175296783447264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,8,4,balanced,0.24044672012329102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,8,4,balanced,0.26436864852905273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,8,4,balanced,0.31758975982666016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,8,4,balanced,0.37171585083007813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,8,4,balanced,0.5259635162353515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,8,4,balanced,0.6373555374145508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,8,4,balanced,0.19257856369018556
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,8,4,balanced,0.8945484924316407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,8,4,balanced,0.031744000911712644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,8,4,balanced,0.03495039939880371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,8,4,balanced,0.04159232139587402
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,8,4,balanced,1.1409907531738281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,8,4,balanced,0.05656832218170166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,8,4,balanced,0.07892096042633057
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,8,4,balanced,0.11543935775756835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,8,4,balanced,0.11026944160461424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,8,4,balanced,0.11102592468261718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,8,4,balanced,0.12380160331726073
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,8,4,balanced,0.12453887939453126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,8,4,balanced,0.1266419219970703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,8,4,balanced,1.6777676391601564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,8,4,balanced,0.13106559753417968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,8,4,balanced,0.13229056358337404
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,8,4,balanced,0.13813247680664062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,8,4,balanced,0.1423731231689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,8,4,balanced,0.15233792304992674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,8,4,balanced,0.16794240951538086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,8,4,balanced,2.197831726074219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,8,4,balanced,0.18371583938598632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,8,4,balanced,0.21610240936279296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,8,4,balanced,0.2470783996582031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,8,4,balanced,0.3094745635986328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,8,4,balanced,0.3759628677368164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,8,4,balanced,0.13651328086853026
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,8,4,balanced,0.5583257675170898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,8,4,balanced,0.6843276977539062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,8,4,balanced,0.9954662322998047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,8,4,balanced,1.3049267578125001
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,4,8,power_law_1.2,0.3789324951171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,4,8,power_law_1.2,0.6567616271972656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,4,8,power_law_1.2,0.8449766540527344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,8,4,balanced,0.031322879791259764
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,8,4,balanced,0.07581183910369874
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,8,4,balanced,0.030115840435028074
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,8,4,balanced,0.03253119945526123
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,8,4,balanced,0.03799936056137085
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,8,4,balanced,0.0759065580368042
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,8,4,balanced,0.07601151943206788
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,8,4,balanced,0.07606272220611572
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,8,4,balanced,0.07666816234588622
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,8,4,balanced,0.0773414421081543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,8,4,balanced,0.0779366397857666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,8,4,balanced,0.07883776187896728
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,8,4,balanced,0.08029824256896972
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,8,4,balanced,0.08219776153564454
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,8,4,balanced,0.08539008140563965
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,8,4,balanced,0.05209472179412842
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,8,4,balanced,0.09035519599914552
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,8,4,balanced,0.09501184463500977
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,8,4,balanced,0.10518272399902344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,8,4,balanced,0.11405823707580567
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,8,4,power_law_1.01,0.03947135925292969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,8,4,power_law_1.01,0.0665715217590332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,8,4,power_law_1.01,0.07069824218750001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,8,4,power_law_1.01,0.08472831726074219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,8,4,power_law_1.01,0.0948953628540039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,8,4,power_law_1.01,0.09982848167419434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,8,4,power_law_1.01,0.11932671546936036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,8,4,power_law_1.01,0.1192204761505127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,8,4,power_law_1.01,0.1650137519836426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,8,4,power_law_1.01,0.16612735748291016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,8,4,power_law_1.01,0.1715545654296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,8,4,power_law_1.01,0.17623935699462892
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,8,4,power_law_1.01,0.1782067108154297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,8,4,power_law_1.01,0.18449920654296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,8,4,power_law_1.01,0.1875391960144043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,8,4,power_law_1.01,0.19312383651733397
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,8,4,power_law_1.01,0.20918527603149414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,8,4,power_law_1.01,0.23646976470947267
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,8,4,power_law_1.01,0.2618073654174805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,8,4,balanced,0.07519231796264648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,8,4,power_law_1.01,0.32366336822509767
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,8,4,power_law_1.01,0.41554176330566406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,8,4,power_law_1.01,0.5506060791015626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,8,4,power_law_1.01,0.6674329376220702
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,8,4,power_law_1.01,0.9460582733154297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,8,4,balanced,0.07550720214843751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,8,4,power_law_1.01,1.23884033203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,8,4,power_law_1.01,0.037236480712890624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,8,4,power_law_1.01,0.04286208152770996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,8,4,power_law_1.01,1.7136703491210938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,8,4,power_law_1.01,0.05821951866149903
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,8,4,power_law_1.01,0.07474944114685059
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,8,4,power_law_1.01,0.08735615730285644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,8,4,power_law_1.01,0.09210368156433105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,8,4,power_law_1.01,0.09608192443847656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,8,4,power_law_1.01,2.34076416015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,8,4,power_law_1.01,0.10019328117370604
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,8,4,power_law_1.01,0.10996479988098144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,8,4,power_law_1.01,0.11371775627136231
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,8,4,power_law_1.01,0.11351552009582519
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,8,4,power_law_1.01,0.11633664131164552
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,8,4,power_law_1.01,0.11796223640441894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,8,4,power_law_1.01,0.12862591743469237
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,8,4,power_law_1.01,0.12888575553894044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,8,4,power_law_1.01,0.1323456001281738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,8,4,power_law_1.01,0.14263808250427246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,8,4,power_law_1.01,0.15724287986755373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,8,4,power_law_1.01,0.18095359802246094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,8,4,power_law_1.01,0.22854272842407228
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,8,4,power_law_1.01,0.27023103713989255
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,8,4,power_law_1.01,0.33772926330566405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,8,4,power_law_1.01,0.4272358322143555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,8,4,power_law_1.01,0.5883123016357421
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,8,4,power_law_1.01,0.7320870208740234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,8,4,power_law_1.01,0.03299072027206421
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,8,4,power_law_1.01,0.0370739197731018
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,8,4,power_law_1.01,1.0108710479736327
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,8,4,power_law_1.01,0.043668479919433595
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,8,4,power_law_1.01,0.049743361473083496
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,8,4,power_law_1.01,0.055564799308776856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,8,4,power_law_1.01,1.354455108642578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,8,4,power_law_1.01,0.061478400230407716
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,8,4,power_law_1.01,0.06850944042205812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,8,4,power_law_1.01,0.07081088066101074
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,8,4,power_law_1.01,0.07243008136749267
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,8,4,power_law_1.01,0.07206528186798096
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,8,4,power_law_1.01,0.07114880084991455
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,8,4,power_law_1.01,0.07271552085876465
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,8,4,power_law_1.01,0.07410304069519043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,8,4,power_law_1.01,0.07834752082824707
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,8,4,power_law_1.01,0.08063615798950195
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,8,4,power_law_1.01,0.08299008369445801
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,8,4,power_law_1.01,0.0836019229888916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,8,4,balanced,0.07603839874267579
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,8,4,balanced,0.2682547187805176
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,8,4,balanced,0.1684774398803711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,8,4,balanced,0.19115903854370117
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,8,4,balanced,0.34027904510498047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,8,4,power_law_1.01,0.10910592079162598
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,8,4,power_law_1.01,0.0937177562713623
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,8,4,power_law_1.01,0.12722304344177246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,8,4,power_law_1.2,0.03955199956893921
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,8,4,power_law_1.01,0.1621683120727539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,8,4,power_law_1.2,0.06706687927246094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,8,4,power_law_1.2,0.07144192218780518
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,8,4,power_law_1.2,0.08251392364501953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,8,4,power_law_1.2,0.09415424346923829
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,8,4,power_law_1.2,0.1049779224395752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,8,4,power_law_1.2,0.11904895782470704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,8,4,power_law_1.2,0.12467328071594239
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,8,4,power_law_1.2,0.16597503662109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,8,4,power_law_1.2,0.1677503967285156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,8,4,power_law_1.2,0.17181440353393557
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,8,4,power_law_1.2,0.17347967147827148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,8,4,power_law_1.2,0.1757619285583496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,8,4,power_law_1.2,0.18339071273803711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,8,4,power_law_1.2,0.1890982437133789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,8,4,power_law_1.2,0.19223552703857422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,8,4,power_law_1.2,0.21171583175659178
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,8,4,power_law_1.2,0.24075008392333985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,8,4,power_law_1.2,0.2682137680053711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,8,4,power_law_1.2,0.33766529083251956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,8,4,power_law_1.2,0.4153881454467774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,8,4,power_law_1.2,0.5669491195678711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,8,4,power_law_1.2,0.7043059539794922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,8,4,power_law_1.01,0.1971072006225586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,8,4,power_law_1.01,0.24903039932250975
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,8,4,power_law_1.01,0.34562049865722655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,8,4,power_law_1.2,1.2798489379882814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,8,4,power_law_1.2,0.03746304035186768
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,8,4,power_law_1.01,0.45326206207275394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,8,4,power_law_1.2,0.043057918548583984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,8,4,power_law_1.2,0.054744319915771486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,8,4,power_law_1.2,0.9842047882080077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,8,4,power_law_1.2,0.07270912170410157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,8,4,power_law_1.2,0.0836070442199707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,8,4,power_law_1.2,0.10124544143676757
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,8,4,power_law_1.2,0.09785856246948242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,8,4,power_law_1.2,0.1108351993560791
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,8,4,power_law_1.2,1.7363673400878905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,8,4,power_law_1.2,0.11165311813354492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,8,4,power_law_1.2,0.11617664337158203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,8,4,power_law_1.2,0.11731200218200684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,8,4,power_law_1.2,0.12209535598754884
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,8,4,power_law_1.2,0.1277875232696533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,8,4,power_law_1.2,0.12982144355773925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,8,4,power_law_1.2,0.13364864349365235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,8,4,power_law_1.2,0.14730239868164063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,8,4,power_law_1.2,0.16832639694213866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,8,4,power_law_1.2,2.3532275390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,8,4,power_law_1.2,0.19153024673461913
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,8,4,power_law_1.2,0.09186944007873535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,8,4,power_law_1.2,0.2433395195007324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,8,4,power_law_1.2,0.26946943283081054
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,8,4,power_law_1.2,0.35697921752929684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,8,4,power_law_1.2,0.43225086212158204
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,8,4,power_law_1.2,0.03288831949234009
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,8,4,power_law_1.2,0.5985728073120117
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,8,4,power_law_1.2,0.03845376014709473
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,8,4,power_law_1.2,0.04026879787445069
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,8,4,power_law_1.2,0.759205093383789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,8,4,power_law_1.2,0.047655677795410155
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,8,4,power_law_1.2,0.054007039070129395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,8,4,power_law_1.2,0.05749375820159912
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,8,4,power_law_1.2,0.06980991840362549
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,8,4,power_law_1.2,1.1043110656738282
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,8,4,power_law_1.2,0.06980735778808593
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,8,4,power_law_1.2,0.06926080226898193
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,8,4,power_law_1.2,0.07030144214630127
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,8,4,power_law_1.2,0.07145215988159179
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,8,4,power_law_1.2,0.07993216037750243
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,8,4,power_law_1.2,0.07523839950561524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,8,4,power_law_1.2,1.4473869323730468
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,8,4,power_law_1.2,0.08359552383422851
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,8,4,power_law_1.2,0.08666367530822754
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,8,4,power_law_1.2,0.09055359840393067
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,8,4,power_law_1.2,0.1232051181793213
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,8,4,power_law_1.2,0.07373439788818359
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,8,4,power_law_1.01,0.6214003372192383
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,8,4,power_law_1.01,1.0083353424072266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,8,4,power_law_1.2,0.10489855766296388
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,8,4,power_law_1.2,0.2001510429382324
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,8,4,power_law_1.2,0.36755455017089844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,8,4,power_law_1.2,0.27290367126464843
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,8,4,power_law_1.2,0.15916288375854493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,16,2,balanced,0.03134720087051392
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,16,2,balanced,0.056825599670410155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,16,2,balanced,0.03255295991897583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,16,2,balanced,0.07059967994689942
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,16,2,balanced,0.09727359771728515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,16,2,balanced,0.1385523223876953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,16,2,balanced,0.1392691230773926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,16,2,balanced,0.14131711959838866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,16,2,balanced,0.14431360244750976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,16,2,balanced,0.14400768280029297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,16,2,balanced,0.214147834777832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,16,2,balanced,0.21680128097534182
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,16,2,balanced,0.2227840042114258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,16,2,balanced,0.2273855972290039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,8,4,power_law_1.2,0.5637171173095703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,16,2,balanced,0.23299455642700195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,16,2,balanced,0.24564224243164062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,16,2,balanced,0.2708543968200684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,16,2,balanced,0.29616384506225585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,16,2,balanced,0.34517120361328124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,16,2,balanced,0.14157567977905275
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,16,2,balanced,0.39514881134033203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,16,2,balanced,0.5551308822631836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,16,2,balanced,0.6651404571533204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,16,2,balanced,0.03381247997283936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,16,2,balanced,0.917422103881836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,16,2,balanced,0.037084159851074217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,16,2,balanced,0.04450816154479981
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,16,2,balanced,0.0654860782623291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,16,2,balanced,0.10510080337524415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,16,2,balanced,0.1533683204650879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,16,2,balanced,0.14866815567016603
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,16,2,balanced,1.6938444519042968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,16,2,balanced,0.14947456359863281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,16,2,balanced,0.15026559829711914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,16,2,balanced,0.15174528121948244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,16,2,balanced,0.154333438873291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,16,2,balanced,0.17509632110595702
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,16,2,balanced,0.1773401641845703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,16,2,balanced,0.18135551452636717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,16,2,balanced,1.1614105224609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,16,2,balanced,0.18188415527343751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,16,2,balanced,0.1855936050415039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,16,2,balanced,0.19422592163085936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,16,2,balanced,0.20759679794311525
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,16,2,balanced,0.2243814468383789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,16,2,balanced,0.2525580787658691
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,16,2,balanced,0.2832192039489746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,16,2,balanced,0.34522750854492185
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,16,2,balanced,0.405689582824707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,16,2,balanced,0.6159040069580077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,16,2,balanced,2.215747833251953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,16,2,balanced,0.7328217315673828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,16,2,balanced,1.0680127716064454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,16,2,balanced,1.4036146545410157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,8,4,power_law_1.2,0.7059750366210937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,8,4,power_law_1.2,1.054228515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,8,4,power_law_1.2,1.625804748535156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,16,2,balanced,0.04916736125946045
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,16,2,balanced,0.045694718360900874
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,16,2,balanced,0.11156736373901369
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,16,2,balanced,0.11199104309082031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,16,2,balanced,0.05922304153442383
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,16,2,balanced,0.11308416366577148
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,16,2,balanced,0.11322879791259766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,16,2,balanced,0.11392255783081055
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,16,2,balanced,0.1153536033630371
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,16,2,balanced,0.07640704154968261
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,16,2,balanced,0.11684351921081544
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,16,2,balanced,0.121146879196167
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,16,2,balanced,0.12657535552978516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,16,2,balanced,0.14084223747253416
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,16,2,balanced,0.1500096035003662
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,16,2,balanced,0.17183231353759765
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,16,2,balanced,0.04444799900054931
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,16,2,balanced,0.19235328674316404
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,16,2,balanced,0.13144319534301757
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,16,2,balanced,0.11064576148986816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,16,2,power_law_1.01,0.032494080066680905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,16,2,power_law_1.01,0.03658495903015137
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,16,2,balanced,0.11136896133422851
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,16,2,power_law_1.01,0.0629414415359497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,16,2,power_law_1.01,0.09224703788757324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,16,2,power_law_1.01,0.10103808403015138
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,16,2,power_law_1.01,0.11892864227294922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,16,2,power_law_1.01,0.12165504455566407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,16,2,power_law_1.01,0.0785484790802002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,16,2,power_law_1.01,0.1249510383605957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,16,2,power_law_1.01,0.12619263648986817
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,16,2,power_law_1.01,0.1334272003173828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,16,2,power_law_1.01,0.2076518440246582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,16,2,power_law_1.01,0.2119705581665039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,16,2,power_law_1.01,0.21934463500976564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,16,2,power_law_1.01,0.2234752082824707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,16,2,power_law_1.01,0.2283417510986328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,16,2,power_law_1.01,0.24310655593872071
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,16,2,power_law_1.01,0.27052160263061525
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,16,2,power_law_1.01,0.2981683158874512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,16,2,power_law_1.01,0.3545971298217773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,16,2,power_law_1.01,0.4254182434082031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,16,2,power_law_1.01,0.5908851242065429
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,16,2,power_law_1.01,0.7029325103759765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,16,2,power_law_1.01,0.974145278930664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,16,2,power_law_1.01,1.2303961944580077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,16,2,power_law_1.01,1.7737907409667968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,16,2,power_law_1.01,0.036170239448547366
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,16,2,power_law_1.01,0.04183807849884033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,16,2,power_law_1.01,0.055048961639404294
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,16,2,power_law_1.01,2.2772671508789064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,16,2,power_law_1.01,0.07470719814300537
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,16,2,power_law_1.01,0.10011903762817383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,16,2,power_law_1.01,0.12310015678405761
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,16,2,power_law_1.01,0.12648063659667969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,16,2,balanced,0.11098752021789551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,16,2,power_law_1.01,0.1299225616455078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,16,2,power_law_1.01,0.13236864089965822
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,16,2,power_law_1.01,0.13518336296081543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,16,2,power_law_1.01,0.14002431869506834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,16,2,power_law_1.01,0.15677696228027344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,16,2,power_law_1.01,0.15838591575622557
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,16,2,power_law_1.01,0.16492927551269532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,16,2,power_law_1.01,0.16802560806274414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,16,2,power_law_1.01,0.1732441520690918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,16,2,power_law_1.01,0.1831283187866211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,16,2,power_law_1.01,0.20224767684936523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,16,2,power_law_1.01,0.22462591171264648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,16,2,power_law_1.01,0.26684543609619144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,16,2,power_law_1.01,0.3209369659423828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,16,2,power_law_1.01,0.399153938293457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,16,2,power_law_1.01,0.48131969451904294
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,16,2,power_law_1.01,0.6321971130371093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,16,2,power_law_1.01,0.8088179016113282
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,16,2,power_law_1.01,0.04646656036376953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,16,2,power_law_1.01,1.1471577453613282
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,16,2,power_law_1.01,0.04946176052093506
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,16,2,power_law_1.01,0.05629183769226074
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,16,2,power_law_1.01,0.06374144077301025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,16,2,power_law_1.01,1.4811698913574218
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,16,2,power_law_1.01,0.0725875186920166
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,16,2,power_law_1.01,0.07959296226501464
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,16,2,power_law_1.01,0.09593600273132324
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,16,2,power_law_1.01,0.09657855987548827
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,16,2,power_law_1.01,0.10072575569152833
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,16,2,power_law_1.01,0.10199935913085938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,16,2,power_law_1.01,0.10591615676879881
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,16,2,power_law_1.01,0.10741632461547852
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,16,2,power_law_1.01,0.11154175758361817
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,16,2,power_law_1.01,0.11485183715820313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,16,2,power_law_1.01,0.11795583724975586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,16,2,power_law_1.01,0.12279168128967285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,16,2,power_law_1.01,0.13144319534301757
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,16,2,balanced,0.2908787155151367
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,16,2,balanced,0.47803905487060544
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,16,2,balanced,0.3340428924560547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,16,2,balanced,0.6159219360351562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,16,2,power_law_1.01,0.15365759849548338
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,16,2,power_law_1.01,0.17547136306762695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,16,2,power_law_1.01,0.2158399963378906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,16,2,power_law_1.01,0.26702592849731444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,16,2,power_law_1.2,0.03516544103622436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,16,2,power_law_1.2,0.036669440269470215
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,16,2,power_law_1.01,0.376495361328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,16,2,power_law_1.2,0.06252927780151367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,16,2,power_law_1.2,0.07411839962005615
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,16,2,power_law_1.2,0.09168512344360351
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,16,2,power_law_1.2,0.1012607955932617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,16,2,power_law_1.2,0.11766016006469726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,16,2,power_law_1.2,0.12383744239807129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,16,2,power_law_1.2,0.12287615776062011
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,16,2,power_law_1.01,0.7493325042724609
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,16,2,power_law_1.01,0.4684428787231445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,16,2,power_law_1.2,0.12902527809143066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,16,2,power_law_1.2,0.13624064445495604
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,16,2,power_law_1.2,0.20625152587890624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,16,2,power_law_1.2,0.21230848312377928
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,16,2,power_law_1.2,0.21861631393432618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,16,2,power_law_1.2,0.22203903198242186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,16,2,power_law_1.2,0.2294924736022949
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,16,2,power_law_1.2,0.24571647644042968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,16,2,power_law_1.2,0.2727884864807129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,16,2,power_law_1.2,0.30005632400512694
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,16,2,power_law_1.2,0.36271102905273434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,16,2,power_law_1.2,0.4413849639892578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,16,2,power_law_1.2,0.5995967864990235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,16,2,power_law_1.2,0.7070195007324218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,16,2,power_law_1.2,0.990882568359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,16,2,power_law_1.2,0.03627392053604126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,16,2,power_law_1.2,1.2447923278808593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,16,2,power_law_1.2,0.041588478088378907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,16,2,power_law_1.2,0.053442559242248534
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,16,2,power_law_1.2,0.0707200002670288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,16,2,power_law_1.2,0.09728256225585938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,16,2,power_law_1.2,1.8004992675781248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,16,2,power_law_1.2,0.10838272094726562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,16,2,power_law_1.2,0.12763392448425293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,16,2,power_law_1.2,0.12857728004455565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,16,2,power_law_1.2,2.3044557189941406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,16,2,power_law_1.2,0.13164928436279297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,16,2,power_law_1.2,0.1357875156402588
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,16,2,power_law_1.2,0.14017151832580566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,16,2,power_law_1.2,0.15786879539489745
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,16,2,power_law_1.2,0.16075904846191408
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,16,2,power_law_1.2,0.16678783416748047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,16,2,power_law_1.2,0.16938112258911134
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,16,2,power_law_1.2,0.1746329689025879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,16,2,power_law_1.2,0.18915327072143556
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,16,2,power_law_1.2,0.21042560577392577
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,16,2,power_law_1.2,0.2312153625488281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,16,2,power_law_1.2,0.2907571220397949
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,16,2,power_law_1.2,0.32512767791748043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,16,2,power_law_1.2,0.40589569091796873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,16,2,power_law_1.2,0.48865665435791017
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,16,2,power_law_1.2,0.65275390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,16,2,power_law_1.2,0.046488318443298336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,16,2,power_law_1.2,0.8292774200439453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,16,2,power_law_1.2,0.049803519248962404
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,16,2,power_law_1.2,1.1626547241210938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,16,2,power_law_1.2,0.05484799861907959
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,16,2,power_law_1.2,0.061539840698242185
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,16,2,power_law_1.2,0.07057024002075195
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,16,2,power_law_1.2,0.07896575927734376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,16,2,power_law_1.2,0.0945305633544922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,16,2,power_law_1.2,1.5046630859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,16,2,power_law_1.2,0.0971878433227539
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,16,2,power_law_1.2,0.09935872077941894
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,16,2,power_law_1.2,0.1016921615600586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,16,2,power_law_1.2,0.10822527885437012
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,16,2,power_law_1.2,0.10641792297363281
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,16,2,power_law_1.2,0.11166208267211915
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,16,2,power_law_1.2,0.11626751899719238
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,16,2,power_law_1.2,0.1227353572845459
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,16,2,power_law_1.2,0.12896639823913575
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,16,2,power_law_1.2,0.17086463928222656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,16,2,power_law_1.01,0.8632704162597656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,16,2,power_law_1.01,1.3824678039550782
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,16,2,power_law_1.01,1.9890713500976562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,16,2,power_law_1.2,0.20135551452636719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,16,2,power_law_1.2,0.1457420825958252
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,16,2,power_law_1.2,0.330959358215332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,16,2,power_law_1.2,0.266431999206543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,32,1,balanced,0.045158400535583496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,32,1,balanced,0.044132480621337886
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,32,1,balanced,0.07053055763244628
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,32,1,balanced,0.0883296012878418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,32,1,balanced,0.11873472213745118
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,32,1,balanced,0.18213888168334963
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,32,1,balanced,0.18501567840576172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,32,1,balanced,0.18555776596069334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,32,1,balanced,0.18672063827514648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,32,1,balanced,0.1887238311767578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,32,1,balanced,0.19159231185913086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,32,1,balanced,0.19538047790527344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,32,1,balanced,0.20245439529418946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,32,1,balanced,0.20822528839111326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,32,1,balanced,0.3109312057495117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,32,1,balanced,0.3164076805114746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,32,1,balanced,0.33024192810058595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,32,1,balanced,0.35275009155273435
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,32,1,balanced,0.37673919677734374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,32,1,balanced,0.4228505706787109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,32,1,balanced,0.4742022323608398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,32,1,balanced,0.6345606231689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,16,2,power_law_1.2,0.5220723342895508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,32,1,balanced,0.7406572723388672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,16,2,power_law_1.2,0.6406873321533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,32,1,balanced,0.9970598602294922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,16,2,power_law_1.2,0.9248703765869142
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,32,1,balanced,1.2482675170898436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,32,1,balanced,0.0435590410232544
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,32,1,balanced,0.05079999923706054
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,32,1,balanced,0.061526398658752444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,32,1,balanced,0.08672896385192871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,32,1,balanced,0.14619711875915525
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,32,1,balanced,0.24023296356201174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,32,1,balanced,1.8016709899902346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,32,1,balanced,0.2404185676574707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,32,1,balanced,0.24367679595947264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,32,1,balanced,0.24494400024414062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,32,1,balanced,0.24617727279663085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,32,1,balanced,0.2498489570617676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,32,1,balanced,0.2534899139404297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,32,1,balanced,0.2568902397155762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,32,1,balanced,0.2636044883728027
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,32,1,balanced,0.3084767913818359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,32,1,balanced,2.3545074462890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,32,1,balanced,0.31204736709594727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,32,1,balanced,0.32002624511718747
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,32,1,balanced,0.33205310821533207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,32,1,balanced,0.3458137512207031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,32,1,balanced,0.3737548828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,32,1,balanced,0.4028531265258789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,32,1,balanced,0.46014846801757814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,32,1,balanced,0.5168633651733399
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,32,1,balanced,0.8266207885742187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,32,1,balanced,0.9417043304443359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,32,1,balanced,1.3677113342285157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,32,1,balanced,1.7872499084472657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,16,2,power_law_1.2,1.4388684082031251
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,16,2,power_law_1.2,2.2267237854003907
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,16,2,power_law_1.2,2.8152178955078124
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,32,1,balanced,0.07473279953002929
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,32,1,balanced,0.07255360126495361
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,32,1,balanced,0.08303104400634766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,32,1,balanced,0.1829190444946289
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,32,1,balanced,0.12296192169189453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,32,1,balanced,0.1830406379699707
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,32,1,balanced,0.1844735908508301
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,32,1,balanced,0.18627071380615234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,32,1,balanced,0.18901695251464842
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,32,1,balanced,0.1865990447998047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,32,1,balanced,0.19479679107666015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,32,1,balanced,0.19163328170776367
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,32,1,balanced,0.19942527770996094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,32,1,balanced,0.2087673568725586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,32,1,balanced,0.09483839988708495
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,32,1,balanced,0.21814655303955077
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,32,1,balanced,0.2589804840087891
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,32,1,balanced,0.23928384780883788
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,32,1,balanced,0.30253120422363283
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,32,1,balanced,0.3429190444946289
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,32,1,balanced,0.1801356887817383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,32,1,power_law_1.01,0.041892480850219724
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,32,1,power_law_1.01,0.07005184173583985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,32,1,power_law_1.01,0.08715904235839844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,32,1,power_law_1.01,0.10664640426635744
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,32,1,balanced,0.18073408126831053
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,32,1,power_law_1.01,0.13189311981201174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,32,1,power_law_1.01,0.15353343963623048
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,32,1,power_law_1.01,0.16069120407104492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,32,1,power_law_1.01,0.16483455657958984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,32,1,power_law_1.01,0.16540672302246093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,32,1,balanced,0.1805843162536621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,32,1,power_law_1.01,0.17386816024780274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,32,1,power_law_1.01,0.18553344726562498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,32,1,power_law_1.01,0.19163263320922852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,32,1,power_law_1.01,0.21133823394775392
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,32,1,power_law_1.01,0.30418560028076175
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,32,1,power_law_1.01,0.3116966438293457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,32,1,power_law_1.01,0.3284307098388672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,32,1,power_law_1.01,0.3542911911010742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,32,1,power_law_1.01,0.38272895812988283
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,32,1,power_law_1.01,0.43932224273681636
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,32,1,power_law_1.01,0.4955801773071289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,32,1,power_law_1.01,0.6452845001220704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,32,1,power_law_1.01,0.7979225921630859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,32,1,power_law_1.01,1.0538009643554687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,32,1,power_law_1.01,1.3316064453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,32,1,power_law_1.01,0.0435475206375122
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,32,1,power_law_1.01,0.049480319023132324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,32,1,power_law_1.01,1.8661932373046874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,32,1,power_law_1.01,0.063787522315979
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,32,1,power_law_1.01,0.09282688140869141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,32,1,power_law_1.01,0.12462656021118164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,32,1,power_law_1.01,0.17970176696777346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,32,1,power_law_1.01,2.408106231689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,32,1,power_law_1.01,0.20148096084594727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,32,1,power_law_1.01,0.21061439514160157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,32,1,power_law_1.01,0.21291072845458983
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,32,1,power_law_1.01,0.2163852882385254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,32,1,power_law_1.01,0.2236083221435547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,32,1,power_law_1.01,0.2293497657775879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,32,1,power_law_1.01,0.234085750579834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,32,1,power_law_1.01,0.2413420867919922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,32,1,power_law_1.01,0.28581823348999025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,32,1,power_law_1.01,0.28917760848999025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,32,1,power_law_1.01,0.3022867202758789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,32,1,power_law_1.01,0.32076545715332033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,32,1,power_law_1.01,0.34732990264892577
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,32,1,power_law_1.01,0.39563262939453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,32,1,power_law_1.01,0.037765760421752934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,32,1,power_law_1.01,0.4569068908691406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,32,1,power_law_1.01,0.5516223907470703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,32,1,power_law_1.01,0.647795181274414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,32,1,power_law_1.01,0.8519238281250001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,32,1,power_law_1.01,1.06837890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,32,1,power_law_1.01,1.485518035888672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,32,1,power_law_1.01,0.07362815856933594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,32,1,power_law_1.01,0.07581888198852539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,32,1,power_law_1.01,1.9050534057617188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,32,1,power_law_1.01,0.09723456382751465
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,32,1,power_law_1.01,0.08498751640319824
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,32,1,power_law_1.01,0.11158335685729981
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,32,1,power_law_1.01,0.12400320053100586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,32,1,power_law_1.01,0.1515071964263916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,32,1,power_law_1.01,0.1578451156616211
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,32,1,power_law_1.01,0.16052928924560547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,32,1,power_law_1.01,0.16324224472045898
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,32,1,power_law_1.01,0.16955583572387695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,32,1,power_law_1.01,0.17550336837768554
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,32,1,power_law_1.01,0.1798361587524414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,32,1,power_law_1.01,0.18618112564086914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,32,1,power_law_1.01,0.19623104095458985
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,32,1,power_law_1.01,0.2210099220275879
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,32,1,power_law_1.01,0.20432960510253909
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,32,1,balanced,0.5341798400878905
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,32,1,balanced,0.8996896362304687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,32,1,balanced,1.1714982604980468
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,32,1,balanced,0.6187750244140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,32,1,power_law_1.01,0.2581644821166992
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,32,1,power_law_1.01,0.43677120208740233
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,32,1,power_law_1.01,0.2950611114501953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,32,1,power_law_1.01,0.36809024810791013
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,32,1,power_law_1.2,0.03992959976196289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,32,1,power_law_1.2,0.043263998031616215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,32,1,power_law_1.2,0.08493184089660645
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,32,1,power_law_1.01,0.5927276611328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,32,1,power_law_1.01,0.748806381225586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,32,1,power_law_1.2,0.10455615997314453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,32,1,power_law_1.2,0.15118271827697755
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,32,1,power_law_1.2,0.13385727882385254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,32,1,power_law_1.2,0.15939392089843749
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,32,1,power_law_1.2,0.16350080490112304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,32,1,power_law_1.2,0.1668351936340332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,32,1,power_law_1.2,0.173939208984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,32,1,power_law_1.2,0.18635839462280274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,32,1,power_law_1.2,0.19612863540649414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,32,1,power_law_1.01,1.1316722869873046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,32,1,power_law_1.2,0.3065881538391113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,32,1,power_law_1.2,0.31363840103149415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,32,1,power_law_1.2,0.32901184082031254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,32,1,power_law_1.2,0.357446403503418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,32,1,power_law_1.2,0.3864646530151367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,32,1,power_law_1.2,0.21398719787597656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,32,1,power_law_1.2,0.44316673278808594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,32,1,power_law_1.01,1.4499200439453124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,32,1,power_law_1.2,0.5066668701171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,32,1,power_law_1.2,0.6500985717773438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,32,1,power_law_1.2,0.8028998565673827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,32,1,power_law_1.2,0.04350080013275147
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,32,1,power_law_1.2,1.0590406036376954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,32,1,power_law_1.2,0.049687042236328124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,32,1,power_law_1.2,0.06554880142211914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,32,1,power_law_1.2,1.3386540222167969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,32,1,power_law_1.2,0.07076352119445802
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,32,1,power_law_1.2,0.08652607917785644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,32,1,power_law_1.2,0.11920384407043456
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,32,1,power_law_1.2,0.17042623519897462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,32,1,power_law_1.2,0.20039615631103516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,32,1,power_law_1.2,1.880485076904297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,32,1,power_law_1.2,0.20800640106201174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,32,1,power_law_1.2,0.21119808197021483
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,32,1,power_law_1.2,0.21622976303100586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,32,1,power_law_1.2,0.22840576171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,32,1,power_law_1.2,0.23383359909057616
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,32,1,power_law_1.2,0.2442092704772949
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,32,1,power_law_1.2,2.4171519470214844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,32,1,power_law_1.2,0.2883577537536621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,32,1,power_law_1.2,0.2948422431945801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,32,1,power_law_1.2,0.3064575958251953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,32,1,power_law_1.2,0.33276287078857425
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,32,1,power_law_1.2,0.35654846191406253
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,32,1,power_law_1.2,0.22131519317626952
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,32,1,power_law_1.2,0.4212697601318359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,32,1,power_law_1.2,0.4689766311645508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,32,1,power_law_1.2,0.567070083618164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,32,1,power_law_1.2,0.6848748779296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,32,1,power_law_1.2,0.8622406768798829
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,32,1,power_law_1.2,1.0818905639648437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,32,1,power_law_1.2,0.07299903869628907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,32,1,power_law_1.2,1.5002418518066407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,32,1,power_law_1.2,0.07607679843902589
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,32,1,power_law_1.2,0.0844428825378418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,32,1,power_law_1.2,1.919384307861328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,32,1,power_law_1.2,0.09509696006774902
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,32,1,power_law_1.2,0.10799103736877443
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,32,1,power_law_1.2,0.12144512176513671
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,32,1,power_law_1.2,0.14942912101745603
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,32,1,power_law_1.2,0.15959168434143067
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,32,1,power_law_1.2,0.1556486415863037
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,32,1,power_law_1.2,0.16470207214355467
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,32,1,power_law_1.2,0.17135040283203123
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,32,1,power_law_1.2,0.1773427200317383
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,32,1,power_law_1.2,0.18419456481933594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,32,1,power_law_1.2,0.1948627281188965
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,32,1,power_law_1.2,0.20523391723632814
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,32,1,power_law_1.2,0.21629119873046876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,32,1,power_law_1.01,2.170019226074219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,32,1,power_law_1.01,2.868177795410156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,32,1,power_law_1.2,0.29343423843383787
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,32,1,power_law_1.2,0.24402496337890628
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,32,1,power_law_1.2,0.46353984832763667
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,32,1,power_law_1.2,0.34491710662841796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,32,balanced,0.02184959888458252
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,32,1,power_law_1.2,0.8379891204833985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,32,balanced,0.021856000423431395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,32,balanced,0.019786239862442018
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,32,1,power_law_1.2,0.5826086425781251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,32,balanced,0.02227583885192871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,32,balanced,0.034193921089172366
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,32,balanced,0.03300479888916016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,32,balanced,0.0349401593208313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,32,balanced,0.03665920019149781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,32,balanced,0.03544831991195679
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,32,balanced,0.038297600746154785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,32,balanced,0.0383078408241272
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,32,balanced,0.03928191900253296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,32,balanced,0.04075647830963135
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,32,balanced,0.04199679851531982
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,32,balanced,0.04357888221740723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,32,balanced,0.04221824169158935
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,32,balanced,0.046329598426818844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,32,balanced,0.04821375846862793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,32,balanced,0.020309760570526122
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,32,balanced,0.058291201591491704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,32,balanced,0.06184192180633545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,32,balanced,0.08693632125854492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,32,balanced,0.10421248435974122
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,32,balanced,0.1376294422149658
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,32,balanced,0.17103744506835938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,32,balanced,0.24555007934570314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,32,balanced,0.024384000301361085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,32,balanced,0.32261119842529296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,32,balanced,0.024401919841766356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,32,balanced,0.023508479595184324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,32,balanced,0.024687359333038332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,32,balanced,0.02519808053970337
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,32,balanced,0.030453760623931885
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,32,balanced,0.03173631906509399
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,32,balanced,0.03089920043945312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,32,balanced,0.03297280073165894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,32,balanced,0.03214463949203491
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,32,balanced,0.03296128034591675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,32,balanced,0.034200320243835444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,32,balanced,0.03533695936203003
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,32,balanced,0.0296614408493042
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,32,balanced,0.036366078853607174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,32,balanced,0.03216255903244018
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,32,balanced,0.03442559957504272
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,32,balanced,0.036843519210815426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,32,balanced,0.041547517776489254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,32,balanced,0.04471295833587646
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,32,balanced,0.05630591869354248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,32,balanced,0.03257983922958374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,32,balanced,0.06592127799987793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,32,balanced,0.08764032363891602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,32,balanced,0.10854911804199219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,32,balanced,0.15510656356811522
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,32,balanced,0.20443775177001955
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,32,1,power_law_1.2,1.131475830078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,32,1,power_law_1.2,1.7071603393554688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,1,32,balanced,0.015130879878997801
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,1,32,balanced,0.01529088020324707
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,1,32,balanced,0.015720959901809693
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,1,32,balanced,0.015357439517974853
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,1,32,balanced,0.015697920322418214
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,1,32,balanced,0.015331840515136719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,1,32,balanced,0.015362559556961058
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,1,32,balanced,0.015353599786758423
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,1,32,balanced,0.01578495979309082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,1,32,balanced,0.015735039710998537
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,1,32,balanced,0.015325440168380739
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,1,32,balanced,0.0156441605091095
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,1,32,balanced,0.01550336003303528
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,1,32,balanced,0.015802880525588987
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,1,32,balanced,0.0157695996761322
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,1,32,balanced,0.015778559446334838
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,1,32,balanced,0.016104960441589357
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,1,32,balanced,0.01658112049102783
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,1,32,balanced,0.016343040466308596
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,32,1,power_law_1.2,2.4221125793457032
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,1,32,balanced,0.015344640016555786
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,1,32,balanced,0.015370240211486816
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,1,32,balanced,0.018170880079269407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,32,power_law_1.01,0.02194688081741333
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,1,32,balanced,0.018663680553436278
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,32,power_law_1.01,0.021456639766693115
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,32,power_law_1.01,0.021960959434509278
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,32,power_law_1.01,0.03250432014465332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,32,power_law_1.01,0.03382143974304199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,32,power_law_1.01,0.03460736036300659
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,32,power_law_1.01,0.03708287954330444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,32,power_law_1.01,0.03586303949356079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,32,power_law_1.01,0.03852799892425537
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,32,power_law_1.01,0.03949440002441406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,32,power_law_1.01,0.039777278900146484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,32,power_law_1.01,0.0416755199432373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,32,power_law_1.01,0.04460927963256836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,32,power_law_1.01,0.04892928123474121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,32,power_law_1.01,0.04545279979705811
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,1,32,balanced,0.025213439464569092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,32,power_law_1.01,0.048958721160888674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,32,power_law_1.01,0.054260478019714356
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,1,32,balanced,0.02214143991470337
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,32,power_law_1.01,0.0665561580657959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,32,power_law_1.01,0.08104703903198243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,32,power_law_1.01,0.10061440467834473
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,32,power_law_1.01,0.13508352279663086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,32,power_law_1.01,0.1702694320678711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,32,power_law_1.01,0.2281203269958496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,32,power_law_1.01,0.37057281494140626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,32,power_law_1.01,0.5315865707397461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,32,power_law_1.01,0.02473344087600708
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,32,power_law_1.01,0.02476799964904785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,32,power_law_1.01,0.7032742309570312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,32,power_law_1.01,0.888727035522461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,32,power_law_1.01,0.02514048099517822
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,32,power_law_1.01,0.03077759981155396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,32,power_law_1.01,0.03053440093994141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,32,power_law_1.01,0.033002240657806395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,32,power_law_1.01,0.03179519891738892
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,32,power_law_1.01,0.03293056011199951
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,32,power_law_1.01,0.034149119853973384
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,32,power_law_1.01,0.03505408048629761
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,32,power_law_1.01,0.03743232011795044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,32,power_law_1.01,0.036590080261230465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,32,power_law_1.01,0.034214398860931396
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,1,32,balanced,0.029510400295257568
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,32,power_law_1.01,0.03303679943084717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,32,power_law_1.01,0.03581056118011475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,32,power_law_1.01,0.040757761001586915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,32,power_law_1.01,0.05255680084228516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,32,power_law_1.01,0.06766719818115234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,32,power_law_1.01,0.029693439006805423
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,32,power_law_1.01,0.07834112167358398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,32,power_law_1.01,0.11065216064453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,32,power_law_1.01,0.1670591926574707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,32,power_law_1.01,0.18410367965698243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,32,power_law_1.01,0.30114303588867186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,32,power_law_1.01,0.38084606170654295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,32,power_law_1.01,0.044464640617370606
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,1,32,power_law_1.01,0.015343359708786012
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,1,32,power_law_1.01,0.015772160291671753
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,1,32,power_law_1.01,0.015356160402297975
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,1,32,power_law_1.01,0.015356160402297975
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,1,32,power_law_1.01,0.015430400371551514
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,1,32,power_law_1.01,0.01578495979309082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,1,32,power_law_1.01,0.015377919673919677
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,1,32,power_law_1.01,0.015741440057754515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,32,power_law_1.01,0.6483058929443359
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,1,32,power_law_1.01,0.015368959903717042
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,1,32,power_law_1.01,0.0157478404045105
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,1,32,power_law_1.01,0.01566208004951477
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,1,32,power_law_1.01,0.01571328043937683
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,1,32,power_law_1.01,0.01619328022003174
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,1,32,power_law_1.01,0.01577087998390198
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,1,32,power_law_1.01,0.016989439725875854
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,1,32,balanced,0.03503999948501587
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,1,32,power_law_1.01,0.015380480289459228
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,1,32,power_law_1.01,0.016665600538253784
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,1,32,power_law_1.01,0.02145279884338379
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,1,32,power_law_1.01,0.01850111961364746
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,1,32,power_law_1.01,0.02649087905883789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,1,32,power_law_1.01,0.019598079919815065
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,1,32,power_law_1.01,0.031741440296173096
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,1,32,power_law_1.01,0.04791168212890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,32,power_law_1.2,0.021546239852905276
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,1,32,power_law_1.01,0.0325708794593811
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,1,32,power_law_1.01,0.06476672172546387
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,1,32,power_law_1.01,0.08105216026306153
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,32,power_law_1.2,0.034206719398498536
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,32,power_law_1.2,0.021856000423431395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,32,power_law_1.2,0.036990718841552736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,32,power_law_1.2,0.0382860803604126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,32,power_law_1.2,0.03828223943710327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,32,power_law_1.2,0.04028672218322754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,32,power_law_1.2,0.042388482093811033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,32,power_law_1.2,0.021875200271606447
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,32,power_law_1.2,0.04382080078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,32,power_law_1.2,0.04568448066711426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,32,power_law_1.2,0.04734848022460937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,32,power_law_1.2,0.046933760643005376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,32,power_law_1.2,0.03256831884384155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,32,power_law_1.2,0.05217152118682862
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,32,power_law_1.2,0.06049920082092285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,1,32,power_law_1.01,0.1106879997253418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,32,power_law_1.2,0.07923711776733398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,32,power_law_1.2,0.09528063774108887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,32,power_law_1.2,0.034630401134490965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,32,power_law_1.2,0.12311936378479005
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,32,power_law_1.2,0.15699328422546385
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,32,power_law_1.2,0.26092800140380856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,32,power_law_1.2,0.02562175989151001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,32,power_law_1.2,0.025100800991058346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,32,power_law_1.2,0.2914265632629395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,32,power_law_1.2,0.025189120769500733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,32,power_law_1.2,0.029867520332336427
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,32,power_law_1.2,0.030092799663543703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,32,power_law_1.2,0.031563520431518555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,32,power_law_1.2,0.03158400058746338
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,32,power_law_1.2,0.032629759311676027
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,32,power_law_1.2,0.03255295991897583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,32,power_law_1.2,0.47012863159179685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,32,power_law_1.2,0.03329792022705078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,32,power_law_1.2,0.03449088096618652
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,32,power_law_1.2,0.03625087976455689
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,32,power_law_1.2,0.03739775896072388
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,32,power_law_1.2,0.03591552019119263
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,32,power_law_1.2,0.036062719821929934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,32,power_law_1.2,0.037868800163269045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,32,power_law_1.2,0.039211521148681636
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,32,power_law_1.2,0.052701439857482914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,32,power_law_1.2,0.061560320854187014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,32,power_law_1.2,0.07873280048370361
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,32,power_law_1.2,0.08905088424682617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,32,power_law_1.2,0.194069766998291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,32,power_law_1.2,0.14916735649108886
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,32,power_law_1.2,0.281844482421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,32,power_law_1.2,0.39839103698730466
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,1,32,power_law_1.2,0.01572991967201233
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,1,32,power_law_1.2,0.015371520519256592
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,1,32,power_law_1.2,0.015327999591827393
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,1,32,power_law_1.2,0.015329279899597167
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,1,32,power_law_1.2,0.015438079833984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,1,32,power_law_1.2,0.015664639472961425
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,1,32,power_law_1.2,0.01534719944000244
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,1,32,power_law_1.2,0.015678720474243162
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,1,32,power_law_1.2,0.01573632001876831
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,1,32,power_law_1.2,0.01575935959815979
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,1,32,power_law_1.2,0.015752960443496705
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,1,32,power_law_1.2,0.01569664001464844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,1,32,power_law_1.2,0.0157478404045105
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,1,32,power_law_1.2,0.015731199979782104
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,32,power_law_1.2,0.5512614440917969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,32,power_law_1.2,0.5738150405883788
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,1,32,power_law_1.2,0.01701120018959045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,32,power_law_1.2,0.9221606445312499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,32,power_law_1.2,0.7357657623291016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,32,power_law_1.2,1.1724889373779297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,1,32,power_law_1.2,0.019043840169906616
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,1,32,power_law_1.2,0.016915199756622316
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,1,32,power_law_1.2,0.017766400575637817
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,1,32,power_law_1.2,0.02069119930267334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,1,32,power_law_1.2,0.024769279956817627
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,1,32,power_law_1.2,0.02717952013015747
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,16,balanced,0.019404799938201905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,16,balanced,0.019007999897003174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,16,balanced,0.01906559944152832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,16,balanced,0.02185472011566162
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,16,balanced,0.03255935907363892
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,16,balanced,0.034115839004516604
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,16,balanced,0.03543296098709107
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,16,balanced,0.03576319932937622
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,16,balanced,0.03676543951034546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,16,balanced,0.038223359584808346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,16,balanced,0.039946238994598385
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,16,balanced,0.04167935848236084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,16,balanced,0.03796479940414429
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,16,balanced,0.040387840270996095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,16,balanced,0.041452798843383785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,16,balanced,0.044779520034790035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,16,balanced,0.05116415977478027
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,16,balanced,0.05671552181243896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,16,balanced,0.07834368228912354
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,16,balanced,0.09005311965942384
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,16,balanced,0.11807359695434569
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,16,balanced,0.14532992362976074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,16,balanced,0.19749631881713867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,16,balanced,0.2525503921508789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,16,balanced,0.3709952163696289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,16,balanced,0.48457214355468753
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,16,balanced,0.0227455997467041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,16,balanced,0.023060479164123536
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,16,balanced,0.02348927974700928
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,16,balanced,0.023912959098815918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,16,balanced,0.024311039447784424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,16,balanced,0.029651200771331786
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,16,balanced,0.030456318855285647
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,16,balanced,0.03087104082107544
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,16,balanced,0.030926079750061036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,16,balanced,0.03135616064071655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,16,balanced,0.0329036808013916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,16,balanced,0.03374847888946533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,16,balanced,0.03504512071609497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,16,balanced,0.030872321128845213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,16,balanced,0.03307264089584351
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,1,32,power_law_1.2,0.04844543933868408
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,16,balanced,0.03412735939025879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,16,balanced,0.036240639686584475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,16,balanced,0.03700736045837402
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,16,balanced,0.018981120586395263
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,16,balanced,0.042456321716308594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,16,balanced,0.04928256034851074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,16,balanced,0.0811558437347412
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,16,balanced,0.057007360458374026
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,16,balanced,0.09516544342041015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,16,balanced,0.13486335754394532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,16,balanced,0.167825927734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,16,balanced,0.2461542320251465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,16,balanced,0.3285145568847656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,1,32,power_law_1.2,0.03586816072463989
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,1,32,power_law_1.2,0.06703487873077393
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,1,32,power_law_1.2,0.08202879905700684
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,1,32,power_law_1.2,0.11414143562316895
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,2,16,balanced,0.016131839752197265
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,2,16,balanced,0.01615615963935852
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,2,16,balanced,0.01658751964569092
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,2,16,balanced,0.016579840183258057
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,2,16,balanced,0.016930559873580935
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,2,16,balanced,0.0169868803024292
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,2,16,balanced,0.01693439960479736
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,2,16,balanced,0.01703935980796814
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,2,16,balanced,0.017176320552825926
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,2,16,balanced,0.016993279457092288
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,2,16,balanced,0.01692800045013428
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,2,16,balanced,0.017328640222549437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,2,16,balanced,0.015737600326538086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,2,16,balanced,0.01774464011192322
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,2,16,balanced,0.015717120170593263
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,2,16,balanced,0.01821951985359192
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,2,16,balanced,0.01782912015914917
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,2,16,balanced,0.01846783995628357
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,2,16,balanced,0.016940799951553346
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,2,16,balanced,0.016632319688796998
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,1,32,power_law_1.2,0.13237119674682618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,16,power_law_1.01,0.021108479499816896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,16,power_law_1.01,0.022105600833892822
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,16,power_law_1.01,0.022288639545440674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,16,power_law_1.01,0.022726399898529055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,16,power_law_1.01,0.03342848062515259
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,16,power_law_1.01,0.03583616018295288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,16,power_law_1.01,0.037876479625701905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,16,power_law_1.01,0.0366425609588623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,16,power_law_1.01,0.03656447887420654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,16,power_law_1.01,0.03753727912902832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,16,power_law_1.01,0.0396940803527832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,16,power_law_1.01,0.04246143817901611
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,16,power_law_1.01,0.04345856189727783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,16,power_law_1.01,0.04394752025604248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,16,power_law_1.01,0.04665728092193604
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,16,power_law_1.01,0.04820096015930176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,16,power_law_1.01,0.05182975769042969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,16,power_law_1.01,0.06835584163665771
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,16,power_law_1.01,0.08047871589660645
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,2,16,balanced,0.017023999691009522
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,16,power_law_1.01,0.10650495529174804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,16,power_law_1.01,0.12444671630859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,16,power_law_1.01,0.16582784652709961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,16,power_law_1.01,0.19526399612426756
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,16,power_law_1.01,0.35688961029052735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,16,power_law_1.01,0.44386302947998046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,16,power_law_1.01,0.5854489517211914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,16,power_law_1.01,0.024011518955230713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,16,power_law_1.01,0.024299519062042235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,16,power_law_1.01,0.865132827758789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,16,power_law_1.01,0.024039680957794192
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,16,power_law_1.01,0.025137920379638672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,16,power_law_1.01,0.030113279819488525
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,16,power_law_1.01,0.03134848117828369
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,16,power_law_1.01,0.033314559459686276
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,16,power_law_1.01,0.03133568048477173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,16,power_law_1.01,0.03168384075164795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,16,power_law_1.01,0.03299072027206421
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,16,power_law_1.01,0.034955520629882816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,16,power_law_1.01,0.03625855922698974
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,16,power_law_1.01,0.03369856119155884
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,16,power_law_1.01,0.033137919902801516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,16,power_law_1.01,0.035000319480896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,16,power_law_1.01,0.03642368078231811
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,16,power_law_1.01,0.03746432065963745
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,2,16,balanced,0.021184000968933105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,16,power_law_1.01,0.04577023983001709
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,2,16,balanced,0.022264320850372315
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,16,power_law_1.01,0.05493760108947754
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,2,16,balanced,0.03215872049331665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,16,power_law_1.01,0.08130304336547851
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,16,power_law_1.01,0.14277376174926756
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,16,power_law_1.01,0.21815168380737307
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,2,16,balanced,0.026808319091796873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,16,power_law_1.01,0.2875456047058106
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,2,16,power_law_1.01,0.016702719926834107
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,2,16,power_law_1.01,0.01702911972999573
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,16,power_law_1.01,0.39911937713623047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,2,16,power_law_1.01,0.01699839949607849
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,2,16,power_law_1.01,0.01705471992492676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,16,power_law_1.01,0.5809920120239258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,2,16,power_law_1.01,0.016576000452041627
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,2,16,power_law_1.01,0.01698815941810608
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,2,16,power_law_1.01,0.01695744037628174
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,2,16,power_law_1.01,0.01696895956993103
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,2,16,power_law_1.01,0.017044479846954348
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,2,16,power_law_1.01,0.0170035195350647
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,2,16,power_law_1.01,0.017008639574050903
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,2,16,power_law_1.01,0.01745919942855835
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,2,16,power_law_1.01,0.017004799842834473
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,2,16,power_law_1.01,0.017375999689102174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,16,power_law_1.01,0.10012672424316407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,2,16,power_law_1.01,0.02048896074295044
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,2,16,balanced,0.04100736141204834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,16,power_law_1.01,0.07322751998901367
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,2,16,balanced,0.049976320266723634
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,2,16,power_law_1.01,0.016618239879608154
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,2,16,power_law_1.01,0.019384319782257083
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,2,16,power_law_1.01,0.021841919422149657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,2,16,power_law_1.01,0.018394880294799805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,2,16,power_law_1.01,0.024359679222106932
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,16,power_law_1.2,0.021145598888397215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,16,power_law_1.2,0.02067456007003784
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,2,16,power_law_1.01,0.02919552087783813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,16,power_law_1.2,0.021178879737854005
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,16,power_law_1.2,0.021425919532775877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,16,power_law_1.2,0.032975358963012694
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,16,power_law_1.2,0.033681919574737554
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,2,16,power_law_1.01,0.03340159893035889
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,16,power_law_1.2,0.03540607929229737
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,16,power_law_1.2,0.03626240015029907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,16,power_law_1.2,0.037529599666595456
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,16,power_law_1.2,0.03907455921173096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,16,power_law_1.2,0.04236671924591064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,16,power_law_1.2,0.04364799976348877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,16,power_law_1.2,0.0392576003074646
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,16,power_law_1.2,0.04159359931945801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,16,power_law_1.2,0.04434432029724121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,16,power_law_1.2,0.04611968040466309
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,16,power_law_1.2,0.055493121147155766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,16,power_law_1.2,0.07387775897979736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,16,power_law_1.2,0.0877017593383789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,16,power_law_1.2,0.10930944442749024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,16,power_law_1.2,0.14198528289794923
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,16,power_law_1.2,0.19160959243774414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,2,16,power_law_1.01,0.04147200107574463
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,16,power_law_1.2,0.26371456146240235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,16,power_law_1.2,0.3762393569946289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,16,power_law_1.2,0.5063961410522462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,16,power_law_1.2,0.023582720756530763
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,16,power_law_1.2,0.023943679332733152
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,16,power_law_1.2,0.02394495964050293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,16,power_law_1.2,0.028811519145965574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,16,power_law_1.2,0.029898240566253665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,16,power_law_1.2,0.030522880554199217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,16,power_law_1.2,0.03130111932754517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,16,power_law_1.2,0.03206655979156494
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,16,power_law_1.2,0.03257472038269043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,16,power_law_1.2,0.03499775886535644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,16,power_law_1.2,0.035116798877716064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,16,power_law_1.2,0.03659647941589356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,16,power_law_1.2,0.03415424108505249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,16,power_law_1.2,0.036599040031433105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,16,power_law_1.2,0.03707776069641113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,16,power_law_1.2,0.039127039909362796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,16,power_law_1.2,0.04908031940460205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,16,power_law_1.2,0.0577177619934082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,16,power_law_1.2,0.07554431915283202
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,16,power_law_1.2,0.09185152053833008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,16,power_law_1.2,0.13461248397827147
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,16,power_law_1.2,0.16395135879516604
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,16,power_law_1.2,0.23667839050292966
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,16,power_law_1.2,0.024785919189453127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,16,power_law_1.2,0.33931774139404297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,2,16,power_law_1.01,0.05104767799377441
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,2,16,power_law_1.2,0.01674623966217041
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,2,16,power_law_1.2,0.01658751964569092
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,2,16,power_law_1.01,0.06990335941314697
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,2,16,power_law_1.2,0.016925439834594727
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,2,16,power_law_1.01,0.08403583526611327
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,2,16,power_law_1.2,0.016599040031433105
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,2,16,power_law_1.2,0.01656191945075989
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,2,16,power_law_1.2,0.016925439834594727
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,2,16,power_law_1.2,0.01693439960479736
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,2,16,power_law_1.2,0.016929279565811157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,2,16,power_law_1.2,0.01702783942222595
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,2,16,power_law_1.2,0.016944639682769776
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,2,16,power_law_1.2,0.017021440267562866
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,2,16,power_law_1.2,0.016925439834594727
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,2,16,power_law_1.01,0.12786687850952147
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,2,16,power_law_1.2,0.017020159959793092
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,2,16,power_law_1.2,0.018558720350265505
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,2,16,power_law_1.2,0.018718719482421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,16,power_law_1.2,0.751646728515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,16,power_law_1.2,0.9577996826171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,16,power_law_1.2,0.6678272247314453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,16,power_law_1.2,0.5010867309570313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,2,16,power_law_1.2,0.01908095955848694
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,2,16,power_law_1.2,0.02165888071060181
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,2,16,power_law_1.2,0.023541760444641114
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,2,16,power_law_1.2,0.018407679796218872
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,2,16,power_law_1.2,0.026040320396423337
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,4,8,balanced,0.019855359792709353
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,4,8,balanced,0.01896191954612732
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,8,balanced,0.019467519521713255
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,8,balanced,0.021003520488739012
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,8,balanced,0.02354048013687134
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,8,balanced,0.036224000453948975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,8,balanced,0.037482240200042725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,8,balanced,0.03864831924438476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,8,balanced,0.04034304141998291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,8,balanced,0.041561598777771
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,8,balanced,0.037470719814300536
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,8,balanced,0.03953536033630371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,8,balanced,0.04157440185546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,8,balanced,0.04395135879516602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,8,balanced,0.046786561012268066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,8,balanced,0.04966911792755127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,8,balanced,0.05327104091644287
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,2,16,power_law_1.2,0.031267840862274174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,8,balanced,0.070731520652771
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,8,balanced,0.08439552307128906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,8,balanced,0.11092608451843262
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,2,16,power_law_1.2,0.044049921035766604
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,8,balanced,0.18033792495727538
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,8,balanced,0.2267302322387695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,8,balanced,0.3222886276245117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,8,balanced,0.4166963195800781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,8,balanced,0.13465855598449708
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,8,balanced,0.6191680145263672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,8,balanced,0.022661120891571046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,8,balanced,0.02274816036224365
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,8,balanced,0.023915519714355467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,8,balanced,0.8110809326171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,8,balanced,0.02472064018249512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,8,balanced,0.027896320819854735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,8,balanced,0.03248768091201783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,8,balanced,0.03439743995666504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,8,balanced,0.03543936014175415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,8,balanced,0.037064960002899175
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,8,balanced,0.037793281078338625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,8,balanced,0.03337856054306031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,8,balanced,0.0353600001335144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,8,balanced,0.03711103916168213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,8,balanced,0.03875711917877197
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,8,balanced,0.03951488018035888
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,8,balanced,0.042339839935302735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,8,balanced,0.047863039970397946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,8,balanced,0.05707007884979248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,8,balanced,0.07760767936706543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,2,16,power_law_1.2,0.054432001113891605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,8,balanced,0.09182208061218262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,8,balanced,0.12999679565429687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,8,balanced,0.16130048751831055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,8,balanced,0.22898687362670897
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,8,balanced,0.2989580726623535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,8,balanced,0.03874176025390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,8,balanced,0.44609535217285157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,8,balanced,0.5767884826660156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,2,16,power_law_1.2,0.09410688400268555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,2,16,power_law_1.2,0.19716480255126953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,2,16,power_law_1.2,0.0770201587677002
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,4,8,balanced,0.01818112015724182
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,4,8,balanced,0.017487360239028932
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,4,8,balanced,0.0202841591835022
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,4,8,balanced,0.020264959335327147
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,4,8,balanced,0.020483839511871337
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,4,8,balanced,0.020695040225982668
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,4,8,balanced,0.02067456007003784
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,4,8,balanced,0.02066431999206543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,4,8,balanced,0.020718080997467043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,4,8,balanced,0.017091840505599976
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,4,8,balanced,0.020921599864959717
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,4,8,balanced,0.02062079906463623
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,4,8,balanced,0.019399679899215698
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,4,8,balanced,0.02087167978286743
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,4,8,balanced,0.020702719688415527
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,4,8,balanced,0.021081600189208984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,4,8,balanced,0.022716159820556643
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,4,8,balanced,0.02267904043197632
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,2,16,power_law_1.2,0.14023807525634766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,4,8,balanced,0.020247039794921876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,4,8,balanced,0.02027008056640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,4,8,balanced,0.020247039794921876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,4,8,power_law_1.01,0.020407040119171143
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,4,8,power_law_1.01,0.0202841591835022
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,8,power_law_1.01,0.022023680210113524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,8,power_law_1.01,0.02167423963546753
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,8,power_law_1.01,0.02472064018249512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,8,power_law_1.01,0.03489408016204834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,8,power_law_1.01,0.03623680114746094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,8,power_law_1.01,0.03825023889541626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,8,power_law_1.01,0.03867136001586914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,8,power_law_1.01,0.039879679679870605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,8,power_law_1.01,0.04256896018981934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,8,power_law_1.01,0.041428480148315426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,8,power_law_1.01,0.04009215831756592
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,8,power_law_1.01,0.04344319820404053
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,8,power_law_1.01,0.04585984230041504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,8,power_law_1.01,0.05173888206481934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,8,power_law_1.01,0.05875584125518799
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,8,power_law_1.01,0.07167232036590576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,8,power_law_1.01,0.08227456092834473
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,8,power_law_1.01,0.10750464439392089
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,4,8,balanced,0.0296396803855896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,8,power_law_1.01,0.12953472137451172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,8,power_law_1.01,0.1630284881591797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,8,power_law_1.01,0.19807104110717771
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,4,8,balanced,0.027630081176757814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,8,power_law_1.01,0.3023616027832031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,8,power_law_1.01,0.40428417205810546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,8,power_law_1.01,0.02388223886489868
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,4,8,balanced,0.03836416006088257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,8,power_law_1.01,0.5770419311523438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,8,power_law_1.01,0.025173759460449217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,8,power_law_1.01,0.025139200687408447
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,4,8,balanced,0.04606336116790771
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,8,power_law_1.01,0.033054718971252436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,8,power_law_1.01,0.03178623914718628
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,8,power_law_1.01,0.7578752136230469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,8,power_law_1.01,0.034218239784240725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,8,power_law_1.01,0.03462143898010254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,8,power_law_1.01,0.03657984018325806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,8,power_law_1.01,0.03582207918167114
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,8,power_law_1.01,0.03454591989517212
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,8,power_law_1.01,0.03661695957183838
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,8,power_law_1.01,0.03746047973632812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,8,power_law_1.01,0.03870719909667969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,8,power_law_1.01,0.04078080177307129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,8,power_law_1.01,0.049259519577026366
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,8,power_law_1.01,0.056314878463745124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,8,power_law_1.01,0.025585920810699464
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,8,power_law_1.01,0.07325568199157714
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,8,power_law_1.01,0.09202688217163087
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,8,power_law_1.01,0.12666367530822753
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,8,power_law_1.01,0.1534515190124512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,8,power_law_1.01,0.24075904846191407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,8,power_law_1.01,0.2819545555114746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,8,power_law_1.01,0.45225982666015624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,4,8,power_law_1.01,0.018216960430145264
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,4,8,power_law_1.01,0.019397120475769043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,8,power_law_1.01,0.5381977462768555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,4,8,power_law_1.01,0.018988800048828126
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,4,8,power_law_1.01,0.020271360874176025
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,4,8,power_law_1.01,0.018988800048828126
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,4,8,power_law_1.01,0.019866880178451538
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,4,8,power_law_1.01,0.020281600952148437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,4,8,power_law_1.01,0.02025984048843384
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,4,8,power_law_1.01,0.020209920406341553
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,4,8,power_law_1.01,0.020264959335327147
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,4,8,power_law_1.01,0.020198400020599368
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,4,8,power_law_1.01,0.02029952049255371
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,4,8,power_law_1.01,0.0202675199508667
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,4,8,power_law_1.01,0.02061311960220337
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,4,8,power_law_1.01,0.020611839294433592
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,4,8,power_law_1.01,0.019763200283050536
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,8,power_law_1.01,0.038886399269104005
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,4,8,balanced,0.07305984020233154
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,4,8,balanced,0.06155903816223145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,8,power_law_1.01,0.026708478927612307
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,4,8,power_law_1.01,0.027663359642028807
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,4,8,power_law_1.01,0.024796159267425538
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,4,8,power_law_1.01,0.022339839935302735
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,4,8,power_law_1.01,0.03333120107650757
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,4,8,power_law_1.2,0.019392000436782835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,4,8,power_law_1.2,0.020371201038360594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,4,8,power_law_1.01,0.036962559223175054
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,8,power_law_1.2,0.02235136032104492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,8,power_law_1.2,0.023946239948272705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,8,power_law_1.2,0.03714303970336914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,8,power_law_1.2,0.037061119079589845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,8,power_law_1.2,0.03854975938796997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,8,power_law_1.2,0.02068095922470093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,4,8,power_law_1.01,0.04477312088012696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,8,power_law_1.2,0.03927167892456055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,8,power_law_1.2,0.04031487941741944
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,8,power_law_1.2,0.04403327941894532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,8,power_law_1.2,0.04064767837524414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,8,power_law_1.2,0.03996031999588013
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,8,power_law_1.2,0.04403327941894532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,8,power_law_1.2,0.04590208053588867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,4,8,power_law_1.01,0.055278081893920895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,8,power_law_1.2,0.06034175872802734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,8,power_law_1.2,0.0732966423034668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,8,power_law_1.2,0.08620287895202636
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,8,power_law_1.2,0.11208576202392577
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,8,power_law_1.2,0.1329638385772705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,8,power_law_1.2,0.18414592742919922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,8,power_law_1.2,0.23979135513305666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,8,power_law_1.2,0.04787839889526367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,8,power_law_1.2,0.32663169860839847
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,8,power_law_1.2,0.023938560485839845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,8,power_law_1.2,0.02396415948867798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,8,power_law_1.2,0.02560256004333496
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,4,8,power_law_1.01,0.0740390396118164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,8,power_law_1.2,0.026380798816680907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,8,power_law_1.2,0.4459891128540039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,8,power_law_1.2,0.026039040088653563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,8,power_law_1.2,0.034190080165863036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,8,power_law_1.2,0.0354150390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,8,power_law_1.2,0.036587519645690916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,8,power_law_1.2,0.03876991987228394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,8,power_law_1.2,0.03658623933792114
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,8,power_law_1.2,0.03177855968475342
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,8,power_law_1.2,0.036358399391174315
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,8,power_law_1.2,0.0370086407661438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,8,power_law_1.2,0.0392742395401001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,8,power_law_1.2,0.043033599853515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,8,power_law_1.2,0.05132031917572022
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,8,power_law_1.2,0.06040703773498535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,8,power_law_1.2,0.07694079875946044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,8,power_law_1.2,0.04029183864593506
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,8,power_law_1.2,0.09337087631225585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,8,power_law_1.2,0.131843843460083
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,8,power_law_1.2,0.16438528060913088
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,8,power_law_1.2,0.03296639919281006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,8,power_law_1.2,0.32198913574218746
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,4,8,power_law_1.01,0.09849344253540039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,4,8,power_law_1.2,0.01822335958480835
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,4,8,power_law_1.01,0.1418943977355957
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,4,8,power_law_1.2,0.01857408046722412
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,4,8,power_law_1.2,0.019366400241851805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,4,8,power_law_1.2,0.019367680549621583
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,4,8,power_law_1.2,0.01944192051887512
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,4,8,power_law_1.2,0.020209920406341553
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,8,power_law_1.2,0.25633663177490235
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,4,8,power_law_1.2,0.020222721099853517
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,4,8,power_law_1.2,0.019879679679870605
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,4,8,power_law_1.2,0.02021888017654419
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,4,8,power_law_1.2,0.020225279331207276
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,4,8,power_law_1.2,0.020312321186065675
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,4,8,power_law_1.2,0.02023168087005615
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,4,8,power_law_1.01,0.15625472068786622
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,4,8,power_law_1.2,0.020225279331207276
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,4,8,power_law_1.2,0.020679678916931152
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,4,8,power_law_1.2,0.0215116810798645
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,4,8,power_law_1.2,0.021863679885864257
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,4,8,power_law_1.2,0.026580479145050052
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,8,power_law_1.2,0.647842559814453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,8,power_law_1.2,0.6411929321289063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,8,power_law_1.2,0.4689267349243164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,8,power_law_1.2,0.8846578979492188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,4,8,power_law_1.2,0.023735039234161377
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,4,8,power_law_1.2,0.03177983999252319
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,4,8,power_law_1.2,0.038305280208587644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,32,balanced,0.030949120521545408
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,32,balanced,0.03128959894180298
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,4,8,power_law_1.2,0.04281087875366211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,32,balanced,0.03136895895004273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,32,balanced,0.05779712200164795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,32,balanced,0.09263232231140137
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,4,8,power_law_1.2,0.060670719146728516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,32,balanced,0.09422847747802734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,32,balanced,0.09742591857910157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,32,balanced,0.09759488105773925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,32,balanced,0.10199295997619628
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,32,balanced,0.09939200401306152
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,32,balanced,0.10127872467041014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,32,balanced,0.1041433620452881
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,32,balanced,0.10769280433654785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,32,balanced,0.11142271995544432
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,32,balanced,0.11411328315734863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,32,balanced,0.10839808464050293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,32,balanced,0.11660160064697264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,32,balanced,0.12348031997680664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,32,balanced,0.13756287574768067
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,32,balanced,0.15001983642578126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,32,balanced,0.18065408706665037
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,32,balanced,0.2160576057434082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,32,balanced,0.3169830322265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,32,balanced,0.3877836990356445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,32,balanced,0.5453081512451172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,32,balanced,0.029770240783691403
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,32,balanced,0.7061478424072265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,32,balanced,0.03871232032775879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,32,balanced,0.030097920894622803
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,32,balanced,0.031823360919952394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,32,balanced,0.03054464101791382
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,32,balanced,0.03179392099380493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,32,balanced,0.038801920413970944
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,32,balanced,0.03898623943328857
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,32,balanced,0.04307583808898926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,32,balanced,0.040352001190185546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,32,balanced,0.04093696117401123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,32,balanced,0.048509440422058105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,32,balanced,0.041242880821228026
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,32,balanced,0.04188543796539307
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,32,balanced,0.04631552219390869
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,32,balanced,0.046822400093078614
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,32,balanced,0.0494271993637085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,32,balanced,0.048126721382141115
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,32,balanced,0.055979518890380865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,32,balanced,0.06063744068145752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,32,balanced,0.06788352012634277
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,32,balanced,0.07547135829925537
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,32,balanced,0.09799551963806152
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,32,balanced,0.11468799591064453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,32,balanced,0.15859711647033692
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,32,balanced,0.2014143943786621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,32,balanced,0.2770150375366211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,32,balanced,0.3644403076171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,4,8,power_law_1.2,0.06822656154632568
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,4,8,power_law_1.2,0.10394368171691895
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,4,8,power_law_1.2,0.11709568023681641
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,4,8,power_law_1.2,0.22836223602294922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,1,32,balanced,0.017853440046310426
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,1,32,balanced,0.018899199962615968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,1,32,balanced,0.018152960538864137
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,1,32,balanced,0.019040000438690186
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,1,32,balanced,0.01897472023963928
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,1,32,balanced,0.01905151963233948
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,1,32,balanced,0.019006719589233396
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,1,32,balanced,0.01903615951538086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,1,32,balanced,0.019040000438690186
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,1,32,balanced,0.019408639669418335
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,1,32,balanced,0.019423999786376954
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,1,32,balanced,0.019454720020294188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,1,32,balanced,0.020193281173706053
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,1,32,balanced,0.02023296117782593
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,1,32,balanced,0.021410560607910155
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,1,32,balanced,0.01815935969352722
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,1,32,balanced,0.02102400064468384
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,1,32,balanced,0.018302719593048095
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,4,8,power_law_1.2,0.19802112579345704
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,1,32,balanced,0.019078400135040283
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,1,32,balanced,0.01921280026435852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,32,power_law_1.01,0.05795839786529541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,32,power_law_1.01,0.05879551887512206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,32,power_law_1.01,0.057944321632385255
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,32,power_law_1.01,0.09160575866699219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,32,power_law_1.01,0.08960384368896485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,32,power_law_1.01,0.09439871788024903
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,32,power_law_1.01,0.09752448081970215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,32,power_law_1.01,0.09831040382385255
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,32,power_law_1.01,0.10041600227355958
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,32,power_law_1.01,0.10348928451538086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,32,power_law_1.01,0.10582655906677245
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,32,power_law_1.01,0.11228799819946289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,32,power_law_1.01,0.11706496238708497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,32,power_law_1.01,0.11503104209899902
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,32,power_law_1.01,0.114202880859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,32,power_law_1.01,0.11830528259277344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,32,power_law_1.01,0.1340403175354004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,32,power_law_1.01,0.14560256004333497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,32,power_law_1.01,0.17510143280029297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,32,power_law_1.01,0.23261056900024416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,32,power_law_1.01,0.3258982467651367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,32,power_law_1.01,0.43557376861572267
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,32,power_law_1.01,0.5365670394897462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,32,power_law_1.01,0.8174221038818359
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,1,32,balanced,0.01900032043457031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,32,power_law_1.01,0.9640447998046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,32,power_law_1.01,1.646387176513672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,1,32,balanced,0.038590719699859624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,1,32,balanced,0.031176960468292235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,32,power_law_1.01,0.03151360034942627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,32,power_law_1.01,0.03237760066986084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,32,power_law_1.01,0.038229761123657224
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,32,power_law_1.01,0.040168957710266115
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,32,power_law_1.01,1.9331455993652344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,32,power_law_1.01,0.03132927894592285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,1,32,balanced,0.030827519893646238
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,32,power_law_1.01,0.04542719841003418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,32,power_law_1.01,0.03993599891662598
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,32,power_law_1.01,0.041094398498535155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,32,power_law_1.01,0.042256641387939456
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,32,power_law_1.01,0.04523647785186767
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,32,power_law_1.01,0.04816512107849121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,32,power_law_1.01,0.05761407852172852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,32,power_law_1.01,0.05228288173675537
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,32,power_law_1.01,0.05670911788940429
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,1,32,balanced,0.047710719108581545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,32,power_law_1.01,0.05958271980285644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,32,power_law_1.01,0.07613056182861329
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,32,power_law_1.01,0.0849779224395752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,32,power_law_1.01,0.11936896324157716
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,32,power_law_1.01,0.1562816047668457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,32,power_law_1.01,0.04118783950805664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,32,power_law_1.01,0.2205311965942383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,32,power_law_1.01,0.2572287940979004
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,1,32,power_law_1.01,0.0187609601020813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,32,power_law_1.01,0.47075328826904295
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,1,32,power_law_1.01,0.019032319784164427
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,32,power_law_1.01,0.06530816078186034
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,1,32,power_law_1.01,0.018556159734725953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,1,32,power_law_1.01,0.01906559944152832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,1,32,power_law_1.01,0.018978559970855714
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,32,power_law_1.01,0.9134886169433594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,32,power_law_1.01,0.9880818939208984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,1,32,power_law_1.01,0.019023360013961793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,1,32,power_law_1.01,0.019055360555648805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,1,32,power_law_1.01,0.019456000328063966
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,1,32,power_law_1.01,0.019808000326156615
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,32,power_law_1.01,0.36407424926757814
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,1,32,power_law_1.01,0.019788800477981566
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,1,32,power_law_1.01,0.019875839948654175
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,1,32,power_law_1.01,0.0198579204082489
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,1,32,power_law_1.01,0.022731521129608155
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,1,32,power_law_1.01,0.01889407992362976
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,1,32,balanced,0.06214144229888916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,1,32,power_law_1.01,0.019055360555648805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,1,32,power_law_1.01,0.019452159404754636
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,1,32,balanced,0.07452799797058104
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,1,32,power_law_1.01,0.024518399238586425
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,1,32,power_law_1.01,0.046807041168212896
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,1,32,power_law_1.01,0.037557759284973145
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,1,32,power_law_1.01,0.029290239810943603
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,32,power_law_1.2,0.058283519744873044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,32,power_law_1.2,0.058145279884338374
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,1,32,power_law_1.01,0.032616961002349856
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,1,32,power_law_1.01,0.07287168025970459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,32,power_law_1.2,0.05730432033538818
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,32,power_law_1.2,0.08855936050415039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,32,power_law_1.2,0.09176704406738281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,32,power_law_1.2,0.0971660804748535
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,1,32,power_law_1.01,0.06107007980346679
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,32,power_law_1.2,0.09750656127929688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,32,power_law_1.2,0.10168319702148439
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,32,power_law_1.2,0.1017625617980957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,32,power_law_1.2,0.10527999877929688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,32,power_law_1.2,0.11411456108093261
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,32,power_law_1.2,0.11795712471008302
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,32,power_law_1.2,0.11364480018615723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,32,power_law_1.2,0.1199244785308838
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,32,power_law_1.2,0.1257919979095459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,32,power_law_1.2,0.13350144386291504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,32,power_law_1.2,0.17043327331542968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,32,power_law_1.2,0.20452480316162108
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,32,power_law_1.2,0.30568960189819333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,32,power_law_1.2,0.10847488403320313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,32,power_law_1.2,0.3558015823364258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,32,power_law_1.2,0.5706969451904297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,32,power_law_1.2,0.6820774078369141
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,1,32,power_law_1.01,0.10643584251403808
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,32,power_law_1.2,0.032325119972229005
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,32,power_law_1.2,0.03186048030853271
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,32,power_law_1.2,0.031998720169067386
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,32,power_law_1.2,0.03912064075469971
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,32,power_law_1.2,0.04077695846557618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,32,power_law_1.2,0.043207678794860836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,32,power_law_1.2,0.04494719982147217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,32,power_law_1.2,0.046311678886413576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,32,power_law_1.2,1.0957849884033204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,32,power_law_1.2,0.04191103935241699
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,32,power_law_1.2,0.04403711795806885
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,32,power_law_1.2,0.05033855915069581
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,32,power_law_1.2,0.05300096035003662
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,32,power_law_1.2,0.055493121147155766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,32,power_law_1.2,0.05355775833129883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,32,power_law_1.2,0.05844351768493652
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,32,power_law_1.2,0.06374527931213378
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,32,power_law_1.2,0.07178112030029296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,32,power_law_1.2,0.08747520446777343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,32,power_law_1.2,0.11062015533447264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,32,power_law_1.2,0.14864128112792968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,32,power_law_1.2,0.18644479751586912
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,32,power_law_1.2,0.2724352073669434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,32,power_law_1.2,0.3885977554321289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,32,power_law_1.2,2.107976989746094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,32,power_law_1.2,0.5392934417724609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,32,power_law_1.2,1.3866291809082032
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,1,32,power_law_1.2,0.018959360122680662
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,1,32,power_law_1.2,0.019052799940109253
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,1,32,power_law_1.2,0.018602240085601806
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,1,32,power_law_1.2,0.01860991954803467
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,1,32,power_law_1.2,0.01873152017593384
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,1,32,power_law_1.2,0.01861631989479065
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,1,32,power_law_1.2,0.01922943949699402
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,1,32,power_law_1.2,0.01904512047767639
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,1,32,power_law_1.2,0.019435520172119143
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,1,32,power_law_1.2,0.019443199634552003
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,1,32,power_law_1.2,0.019809279441833496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,32,power_law_1.2,0.7120166778564453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,1,32,power_law_1.2,0.019482879638671874
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,1,32,power_law_1.2,0.019799040555953978
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,1,32,power_law_1.2,0.02024960041046143
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,1,32,power_law_1.2,0.02249216079711914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,32,power_law_1.2,1.005660171508789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,1,32,power_law_1.01,0.12998784065246582
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,1,32,power_law_1.01,0.1804697608947754
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,1,32,power_law_1.01,0.21649408340454102
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,1,32,power_law_1.2,0.026758399009704593
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,1,32,power_law_1.2,0.022653439044952393
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,1,32,power_law_1.2,0.03631232023239136
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,1,32,power_law_1.2,0.05550464153289795
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,1,32,power_law_1.2,0.04658688068389892
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,1,32,power_law_1.2,0.032999680042266846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,32,power_law_1.2,1.3300889587402343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,16,balanced,0.027170560359954833
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,16,balanced,0.027221760749816894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,16,balanced,0.02682111978530884
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,16,balanced,0.043425278663635256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,16,balanced,0.055637760162353514
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,16,balanced,0.08962431907653809
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,16,balanced,0.09239935874938965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,16,balanced,0.0954099178314209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,16,balanced,0.0930355167388916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,16,balanced,0.09438464164733887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,16,balanced,0.09808128356933594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,16,balanced,0.10157183647155761
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,16,balanced,0.1040563201904297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,16,balanced,0.09805824279785155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,16,balanced,0.10216575622558595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,16,balanced,0.10596223831176758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,16,balanced,0.11153535842895508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,16,balanced,0.12191103935241698
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,16,balanced,0.12989695549011232
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,16,balanced,0.16088064193725587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,16,balanced,0.1817535972595215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,16,balanced,0.22990335464477538
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,16,balanced,0.2785164833068848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,16,balanced,0.413804817199707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,16,balanced,0.5200870513916016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,16,balanced,0.7441241455078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,16,balanced,0.9719692993164063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,16,balanced,0.028830718994140626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,16,balanced,0.028437759876251224
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,16,balanced,0.0288588809967041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,32,power_law_1.2,3.023747863769531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,16,balanced,0.03081984043121338
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,16,balanced,0.03357311964035034
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,16,balanced,0.040919041633605956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,16,balanced,0.04459392070770264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,16,balanced,0.04665855884552002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,16,balanced,0.04201087951660156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,16,balanced,0.04404352188110351
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,16,balanced,0.044051198959350585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,16,balanced,0.047297282218933104
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,16,balanced,0.050632958412170415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,16,balanced,0.04699007987976074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,16,balanced,0.050728960037231444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,16,balanced,0.05300864219665528
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,16,balanced,0.057762560844421384
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,16,balanced,0.06295680046081543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,16,balanced,0.06959871768951416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,16,balanced,0.08555520057678223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,16,balanced,0.09871359825134277
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,16,balanced,0.13456512451171876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,16,balanced,0.16266496658325197
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,16,balanced,0.2261529541015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,16,balanced,0.30153087615966795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,16,balanced,0.41832958221435546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,16,balanced,0.5497331237792968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,1,32,power_law_1.2,0.07490816116333007
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,1,32,power_law_1.2,0.11001728057861329
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,1,32,power_law_1.2,0.133253116607666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,1,32,power_law_1.2,0.21376768112182618
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,2,16,balanced,0.018224639892578127
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,2,16,balanced,0.018213119506835938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,2,16,balanced,0.020663039684295652
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,2,16,balanced,0.021116158962249755
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,2,16,balanced,0.021098239421844484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,2,16,balanced,0.021098239421844484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,2,16,balanced,0.02109951972961426
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,2,16,balanced,0.0215065598487854
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,2,16,balanced,0.02152575969696045
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,2,16,balanced,0.021777920722961426
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,2,16,balanced,0.021419520378112795
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,1,32,power_law_1.2,0.35249408721923825
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,2,16,balanced,0.02164736032485962
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,2,16,balanced,0.021663999557495116
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,2,16,balanced,0.021829121112823487
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,2,16,balanced,0.023192319869995117
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,2,16,balanced,0.02366080045700073
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,2,16,balanced,0.01794816017150879
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,2,16,balanced,0.019139839410781862
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,1,32,power_law_1.2,0.3147750473022461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,2,16,balanced,0.021026558876037597
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,2,16,balanced,0.020949759483337403
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,16,power_law_1.01,0.054453759193420416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,16,power_law_1.01,0.05477632045745849
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,16,power_law_1.01,0.05461887836456299
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,16,power_law_1.01,0.0545139217376709
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,16,power_law_1.01,0.08262528419494629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,16,power_law_1.01,0.08766207695007325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,16,power_law_1.01,0.09100543975830078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,16,power_law_1.01,0.09332736015319824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,16,power_law_1.01,0.09536383628845216
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,16,power_law_1.01,0.0981760025024414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,16,power_law_1.01,0.10033535957336426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,16,power_law_1.01,0.10724351882934571
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,16,power_law_1.01,0.10526847839355469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,16,power_law_1.01,0.1046399974822998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,16,power_law_1.01,0.10907008171081542
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,16,power_law_1.01,0.11102463722229003
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,16,power_law_1.01,0.12017919540405272
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,16,power_law_1.01,0.13720447540283204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,16,power_law_1.01,0.16584831237792969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,16,power_law_1.01,0.18679679870605467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,16,power_law_1.01,0.25290367126464847
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,16,power_law_1.01,0.3530495834350586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,16,power_law_1.01,0.47342720031738283
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,16,power_law_1.01,0.6933606719970703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,16,power_law_1.01,0.8872601318359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,16,power_law_1.01,1.1953587341308594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,16,power_law_1.01,1.6192037963867185
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,2,16,balanced,0.03581696033477783
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,2,16,balanced,0.021483519077301026
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,16,power_law_1.01,0.03117055892944336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,2,16,balanced,0.037971200942993166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,16,power_law_1.01,0.031746559143066407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,16,power_law_1.01,0.032971520423889164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,16,power_law_1.01,0.04239615917205811
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,16,power_law_1.01,0.033306879997253416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,16,power_law_1.01,0.04038015842437744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,16,power_law_1.01,0.044550399780273434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,16,power_law_1.01,0.04638207912445068
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,16,power_law_1.01,0.05485824108123779
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,16,power_law_1.01,0.050969600677490234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,16,power_law_1.01,0.05174911975860595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,16,power_law_1.01,0.053118720054626464
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,16,power_law_1.01,0.05553791999816895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,16,power_law_1.01,0.05958784103393555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,16,power_law_1.01,0.06270336151123047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,16,power_law_1.01,0.039631359577178955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,16,power_law_1.01,0.07275008201599122
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,16,power_law_1.01,0.09209856033325195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,16,power_law_1.01,0.04239615917205811
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,16,power_law_1.01,0.11894016265869141
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,2,16,balanced,0.048613119125366214
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,16,power_law_1.01,0.161343994140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,16,power_law_1.01,0.20577920913696288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,16,power_law_1.01,0.25896448135375977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,16,power_law_1.01,0.3577267074584961
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,2,16,power_law_1.01,0.020678400993347168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,16,power_law_1.01,0.48401023864746096
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,2,16,power_law_1.01,0.021125121116638185
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,2,16,power_law_1.01,0.02068864107131958
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,2,16,power_law_1.01,0.020616960525512696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,16,power_law_1.01,0.7163878631591797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,2,16,power_law_1.01,0.02025984048843384
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,2,16,power_law_1.01,0.020650238990783693
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,2,16,power_law_1.01,0.021269760131835937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,2,16,power_law_1.01,0.021335039138793945
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,2,16,power_law_1.01,0.02108288049697876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,2,16,power_law_1.01,0.021684479713439942
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,16,power_law_1.01,0.9617062377929688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,2,16,power_law_1.01,0.02112895965576172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,2,16,power_law_1.01,0.021181440353393553
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,2,16,power_law_1.01,0.02143615961074829
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,2,16,power_law_1.01,0.02144000053405762
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,2,16,power_law_1.01,0.021962239742279052
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,2,16,balanced,0.05762944221496582
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,2,16,balanced,0.07492095947265624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,2,16,balanced,0.09404671669006348
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,2,16,power_law_1.01,0.02354687929153442
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,2,16,power_law_1.01,0.029937920570373533
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,2,16,power_law_1.01,0.03712896108627319
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,2,16,power_law_1.01,0.026403839588165286
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,2,16,power_law_1.01,0.04280576229095459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,16,power_law_1.2,0.05395328044891358
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,16,power_law_1.2,0.054364161491394045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,16,power_law_1.2,0.054193921089172356
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,2,16,power_law_1.01,0.049413118362426754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,16,power_law_1.2,0.054015998840332036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,16,power_law_1.2,0.0838374423980713
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,2,16,power_law_1.01,0.06478464126586914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,16,power_law_1.2,0.09038208007812501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,16,power_law_1.2,0.0909119987487793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,16,power_law_1.2,0.09339903831481934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,16,power_law_1.2,0.09625984191894531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,16,power_law_1.2,0.10017663955688476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,16,power_law_1.2,0.10238080024719239
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,16,power_law_1.2,0.10735360145568848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,16,power_law_1.2,0.09923199653625489
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,16,power_law_1.2,0.1055577564239502
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,2,16,power_law_1.01,0.07634047985076904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,16,power_law_1.2,0.10930303573608398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,16,power_law_1.2,0.12846464157104492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,16,power_law_1.2,0.15631232261657715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,16,power_law_1.2,0.17663616180419922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,16,power_law_1.2,0.2223948860168457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,16,power_law_1.2,0.2993497657775879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,16,power_law_1.2,0.4478169631958008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,16,power_law_1.2,0.11666303634643556
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,16,power_law_1.2,1.0481241607666016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,16,power_law_1.2,0.5339456176757812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,16,power_law_1.2,0.031028480529785157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,16,power_law_1.2,0.034149119853973384
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,16,power_law_1.2,0.032016639709472654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,16,power_law_1.2,0.033729279041290285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,16,power_law_1.2,0.8400934600830079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,16,power_law_1.2,0.03790719985961914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,16,power_law_1.2,0.04174592018127442
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,16,power_law_1.2,0.04077824115753174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,16,power_law_1.2,0.044372482299804686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,16,power_law_1.2,0.04646912097930908
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,16,power_law_1.2,0.04799232006072998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,16,power_law_1.2,0.0525875186920166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,16,power_law_1.2,0.055351037979125974
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,16,power_law_1.2,0.0532044792175293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,16,power_law_1.2,0.055349760055542
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,16,power_law_1.2,0.058301439285278325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,16,power_law_1.2,0.0625548791885376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,16,power_law_1.2,0.06670207977294922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,16,power_law_1.2,0.08124544143676757
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,16,power_law_1.2,0.09914752006530761
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,16,power_law_1.2,0.12435711860656737
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,16,power_law_1.2,0.15353599548339844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,16,power_law_1.2,0.22720767974853517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,16,power_law_1.2,0.3195353507995605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,16,power_law_1.2,1.6540531921386719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,16,power_law_1.2,0.4468377685546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,16,power_law_1.2,0.5468697738647461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,2,16,power_law_1.2,0.020650238990783693
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,2,16,power_law_1.2,0.021072640419006347
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,2,16,power_law_1.2,0.02072959899902344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,2,16,power_law_1.2,0.020619521141052248
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,2,16,power_law_1.2,0.020227839946746824
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,2,16,power_law_1.2,0.02117759943008423
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,2,16,power_law_1.2,0.021089279651641847
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,2,16,power_law_1.2,0.021461760997772215
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,2,16,power_law_1.2,0.02116096019744873
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,2,16,power_law_1.2,0.021443839073181155
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,2,16,power_law_1.2,0.02143359899520874
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,2,16,power_law_1.2,0.021062400341033936
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,2,16,power_law_1.2,0.021745920181274414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,2,16,power_law_1.01,0.1272499179840088
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,2,16,power_law_1.01,0.10012288093566894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,16,power_law_1.2,0.8557952117919921
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,2,16,power_law_1.01,0.18233856201171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,2,16,power_law_1.2,0.025125119686126705
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,2,16,power_law_1.01,0.2829785537719727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,16,power_law_1.2,1.1319744110107421
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,2,16,power_law_1.2,0.022554879188537595
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,2,16,power_law_1.2,0.02456576108932495
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,2,16,power_law_1.2,0.02813184022903442
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,2,16,power_law_1.2,0.033381121158599855
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,2,16,power_law_1.2,0.039585280418396
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,2,16,power_law_1.2,0.045598721504211424
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,2,16,power_law_1.2,0.0555187177658081
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,8,balanced,0.02309760093688965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,8,balanced,0.025251839160919186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,8,balanced,0.026734080314636234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,8,balanced,0.04210303783416748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,8,balanced,0.05795328140258789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,8,balanced,0.08882559776306152
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,8,balanced,0.09242495536804199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,8,balanced,0.09539584159851075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,8,balanced,0.09954303741455078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,8,balanced,0.10443008422851563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,8,balanced,0.09560704231262207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,8,balanced,0.09945856094360352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,8,balanced,0.10214400291442871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,8,balanced,0.1086400032043457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,8,balanced,0.11171199798583983
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,8,balanced,0.11624320030212403
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,8,balanced,0.12327808380126952
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,8,balanced,0.14525312423706055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,8,balanced,0.16509056091308594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,8,balanced,0.20723072052001953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,8,balanced,0.24581247329711914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,8,balanced,0.3267161560058594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,8,balanced,0.4094259262084961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,8,balanced,0.6132339096069336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,8,balanced,0.7835020446777344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,16,power_law_1.2,2.08521728515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,8,balanced,1.1431168365478517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,8,balanced,0.027164158821105955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,8,balanced,0.028396799564361575
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,8,balanced,1.5002572631835938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,8,balanced,0.029633278846740725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,8,balanced,0.03299072027206421
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,8,balanced,0.03690880060195923
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,8,balanced,0.05111167907714844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,8,balanced,0.04616703987121582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,8,balanced,0.04803199768066406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,8,balanced,0.04933504104614258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,8,balanced,0.0528166389465332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,8,balanced,0.04854144096374512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,8,balanced,0.053600001335144046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,8,balanced,0.055080962181091306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,8,balanced,0.059827198982238775
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,8,balanced,0.06163839817047119
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,8,balanced,0.06359807968139648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,8,balanced,0.07002624034881591
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,8,balanced,0.08478591918945313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,8,balanced,0.0969983959197998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,8,balanced,0.12436736106872559
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,8,balanced,0.14780415534973143
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,8,balanced,0.2102118492126465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,8,balanced,0.25993600845336917
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,8,balanced,0.36708606719970704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,8,balanced,0.4856115341186523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,8,balanced,0.6958258819580079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,8,balanced,0.9198707580566406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,2,16,power_law_1.2,0.07534336090087891
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,2,16,power_law_1.2,0.10414591789245606
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,2,16,power_law_1.2,0.13190784454345703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,2,16,power_law_1.2,0.16872703552246093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,4,8,balanced,0.018654719591140748
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,4,8,balanced,0.026003201007843018
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,4,8,balanced,0.025963521003723143
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,4,8,balanced,0.01963263988494873
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,4,8,balanced,0.02601344108581543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,4,8,balanced,0.025022718906402584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,4,8,balanced,0.026241281032562257
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,4,8,balanced,0.024855039119720458
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,4,8,balanced,0.026068480014801027
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,4,8,balanced,0.025372159481048585
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,4,8,balanced,0.026179840564727785
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,4,8,balanced,0.025195519924163818
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,4,8,balanced,0.026027519702911377
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,4,8,balanced,0.027729918956756593
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,4,8,balanced,0.028759040832519532
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,4,8,balanced,0.03032320022583008
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,4,8,balanced,0.03271424055099487
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,4,8,balanced,0.020252161026000977
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,4,8,balanced,0.021497600078582764
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,4,8,balanced,0.024797439575195312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,2,16,power_law_1.2,0.28289535522460935
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,4,8,balanced,0.02519295930862427
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,8,power_law_1.01,0.030110719203948977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,8,power_law_1.01,0.04649087905883789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,8,power_law_1.01,0.04934656143188477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,8,power_law_1.01,0.053785600662231446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,8,power_law_1.01,0.054744319915771486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,8,power_law_1.01,0.0914675235748291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,8,power_law_1.01,0.08986880302429198
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,8,power_law_1.01,0.09384960174560547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,8,power_law_1.01,0.09545727729797363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,8,power_law_1.01,0.09796992301940918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,8,power_law_1.01,0.10436351776123047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,8,power_law_1.01,0.09833344459533691
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,8,power_law_1.01,0.09930496215820313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,8,power_law_1.01,0.10340864181518554
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,2,16,power_law_1.2,0.34576255798339844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,8,power_law_1.01,0.10701824188232423
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,8,power_law_1.01,0.11239168167114258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,8,power_law_1.01,0.1185574436187744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,8,power_law_1.01,0.14013567924499512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,8,power_law_1.01,0.15925503730773927
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,8,power_law_1.01,0.18905088424682617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,8,power_law_1.01,0.22623743057250975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,8,power_law_1.01,0.36148094177246093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,8,power_law_1.01,0.4491443252563476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,8,power_law_1.01,0.648384017944336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,8,power_law_1.01,0.8605465698242188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,8,power_law_1.01,1.207088623046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,8,power_law_1.01,0.02972287893295288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,8,power_law_1.01,0.03379456043243408
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,8,power_law_1.01,0.0371558403968811
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,8,power_law_1.01,1.6781593322753907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,8,power_law_1.01,0.03663615942001343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,8,power_law_1.01,0.04122496128082275
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,8,power_law_1.01,0.04389376163482666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,8,power_law_1.01,0.04782336235046387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,8,power_law_1.01,0.04872191905975341
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,8,power_law_1.01,0.05066239833831787
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,8,power_law_1.01,0.05133440017700195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,8,power_law_1.01,0.05793536186218262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,8,power_law_1.01,0.05462783813476563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,8,power_law_1.01,0.05425920009613037
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,8,power_law_1.01,0.058533120155334475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,8,power_law_1.01,0.05944960117340088
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,8,power_law_1.01,0.06234111785888672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,8,power_law_1.01,0.06850944042205812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,8,power_law_1.01,0.08203264236450195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,8,power_law_1.01,0.08889344215393066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,8,power_law_1.01,0.12148991584777832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,4,8,balanced,0.05207168102264405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,8,power_law_1.01,0.13759743690490722
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,4,8,balanced,0.048852481842041016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,8,power_law_1.01,0.1962124824523926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,8,power_law_1.01,0.25389055252075193
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,8,power_law_1.01,0.37474048614501954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,8,power_law_1.01,0.4725158309936523
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,4,8,power_law_1.01,0.021081600189208984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,4,8,power_law_1.01,0.02306175947189331
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,4,8,power_law_1.01,0.024354560375213625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,8,power_law_1.01,0.7242316436767579
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,4,8,power_law_1.01,0.02298111915588379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,8,power_law_1.01,0.8640512084960937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,4,8,power_law_1.01,0.025081601142883304
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,4,8,power_law_1.01,0.026394879817962645
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,4,8,power_law_1.01,0.025159680843353273
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,4,8,power_law_1.01,0.026435840129852294
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,4,8,power_law_1.01,0.025524480342864992
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,4,8,power_law_1.01,0.026001920700073244
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,4,8,power_law_1.01,0.02566015958786011
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,4,8,balanced,0.06624767780303956
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,4,8,power_law_1.01,0.027248640060424805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,4,8,power_law_1.01,0.0280294394493103
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,4,8,power_law_1.01,0.023292160034179686
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,4,8,power_law_1.01,0.03213824033737182
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,4,8,balanced,0.08042752265930175
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,4,8,balanced,0.10699520111083985
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,4,8,balanced,0.1386944007873535
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,4,8,power_law_1.01,0.02519808053970337
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,4,8,power_law_1.01,0.029402880668640136
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,4,8,power_law_1.01,0.036980481147766114
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,4,8,power_law_1.01,0.0441318416595459
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,4,8,power_law_1.01,0.05223423957824706
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,4,8,power_law_1.01,0.0640934419631958
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,8,power_law_1.2,0.03373055934906006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,8,power_law_1.2,0.04275712013244629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,8,power_law_1.2,0.04862207889556885
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,8,power_law_1.2,0.0508838415145874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,8,power_law_1.2,0.052469758987426764
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,8,power_law_1.2,0.08655488014221191
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,8,power_law_1.2,0.09004544258117676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,8,power_law_1.2,0.0935871982574463
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,8,power_law_1.2,0.0962611198425293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,8,power_law_1.2,0.10036607742309571
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,8,power_law_1.2,0.10565631866455079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,8,power_law_1.2,0.101911039352417
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,8,power_law_1.2,0.09936896324157715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,8,power_law_1.2,0.10352512359619141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,8,power_law_1.2,0.11141759872436523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,8,power_law_1.2,0.11469311714172363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,8,power_law_1.2,0.12320639610290526
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,8,power_law_1.2,0.13819135665893553
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,8,power_law_1.2,0.17111167907714844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,8,power_law_1.2,0.2116927909851074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,8,power_law_1.2,0.2656512069702148
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,4,8,power_law_1.01,0.0871731185913086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,8,power_law_1.2,0.5022848129272461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,8,power_law_1.2,0.3688550567626953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,8,power_law_1.2,0.9307123565673828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,4,8,power_law_1.01,0.10204544067382812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,8,power_law_1.2,0.032072958946228025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,8,power_law_1.2,0.6374054336547852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,8,power_law_1.2,0.030950400829315182
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,8,power_law_1.2,0.03567487955093383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,8,power_law_1.2,0.03623807907104492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,8,power_law_1.2,0.04443647861480713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,8,power_law_1.2,0.04451583862304688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,8,power_law_1.2,0.040206079483032224
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,8,power_law_1.2,0.04910719871520996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,8,power_law_1.2,0.04984960079193115
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,8,power_law_1.2,0.052294402122497564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,8,power_law_1.2,0.0560422420501709
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,8,power_law_1.2,0.052046079635620114
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,8,power_law_1.2,0.05513343811035156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,8,power_law_1.2,0.05931903839111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,8,power_law_1.2,0.062383360862731926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,8,power_law_1.2,0.06571392059326173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,8,power_law_1.2,0.0702400016784668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,8,power_law_1.2,0.08185215950012206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,8,power_law_1.2,0.0987609577178955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,8,power_law_1.2,0.15500415802001952
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,8,power_law_1.2,0.1272383975982666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,8,power_law_1.2,1.395851593017578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,8,power_law_1.2,0.22702207565307617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,8,power_law_1.2,0.28296575546264646
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,8,power_law_1.2,0.4214144134521485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,8,power_law_1.2,0.5480268859863282
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,4,8,power_law_1.01,0.13760128021240234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,4,8,power_law_1.2,0.02170624017715454
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,4,8,power_law_1.2,0.02397183895111084
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,4,8,power_law_1.2,0.023948800563812257
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,4,8,power_law_1.2,0.023481600284576416
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,4,8,power_law_1.2,0.025715200901031493
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,4,8,power_law_1.2,0.025521919727325437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,4,8,power_law_1.01,0.28733440399169924
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,4,8,power_law_1.2,0.025958399772644043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,4,8,power_law_1.2,0.025218560695648196
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,4,8,power_law_1.2,0.025945599079132083
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,4,8,power_law_1.2,0.02514559984207153
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,8,power_law_1.2,1.0286374664306641
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,4,8,power_law_1.2,0.025944321155548095
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,4,8,power_law_1.01,0.17549951553344728
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,8,power_law_1.2,0.7507955169677734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,4,8,power_law_1.2,0.03121920108795166
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,4,8,power_law_1.01,0.3466892623901367
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,4,8,power_law_1.2,0.026467840671539306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,8,power_law_1.2,1.7881625366210936
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,4,8,power_law_1.2,0.021096959114074706
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,4,8,power_law_1.2,0.028471040725708007
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,4,8,power_law_1.2,0.03049344062805176
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,4,8,power_law_1.2,0.04029183864593506
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,4,8,power_law_1.2,0.03572223901748657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,4,8,power_law_1.2,0.058873600959777825
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,4,8,power_law_1.2,0.05046527862548829
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,8,4,balanced,0.02353024005889893
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,8,4,balanced,0.02643455982208252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,8,4,balanced,0.02771712064743042
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,8,4,balanced,0.04361855983734131
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,8,4,balanced,0.05956992149353028
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,8,4,balanced,0.06294911861419678
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,8,4,balanced,0.09662848472595215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,8,4,balanced,0.09881343841552734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,8,4,balanced,0.10199808120727538
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,8,4,balanced,0.10609408378601075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,8,4,balanced,0.09774592399597168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,8,4,balanced,0.10042752265930174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,8,4,balanced,0.10497535705566405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,8,4,balanced,0.1076531219482422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,8,4,balanced,0.11026816368103028
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,8,4,balanced,0.11502976417541504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,8,4,balanced,0.12343168258666992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,8,4,balanced,0.14269951820373533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,8,4,balanced,0.1625382423400879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,8,4,balanced,0.20112512588500975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,8,4,balanced,0.23892736434936523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,8,4,balanced,0.3173798370361328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,8,4,balanced,0.3948441696166992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,8,4,balanced,0.5865011215209961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,8,4,balanced,0.7565516662597657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,8,4,balanced,1.1066802978515624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,8,4,balanced,1.4466636657714844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,8,4,balanced,0.028053760528564453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,8,4,balanced,0.029701120853424072
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,8,4,balanced,0.03344255924224854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,8,4,balanced,0.03757440090179444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,8,4,balanced,0.04550911903381348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,8,4,balanced,0.0500377607345581
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,8,4,balanced,0.05808896064758301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,8,4,balanced,0.0590169620513916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,8,4,balanced,0.06058368206024169
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,8,4,balanced,0.06285183906555177
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,8,4,balanced,0.05917439937591553
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,8,4,balanced,0.06149375915527344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,8,4,balanced,0.06382976055145265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,8,4,balanced,0.06794112205505372
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,8,4,balanced,0.06901887893676759
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,8,4,balanced,0.07218944072723389
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,8,4,balanced,0.07712384223937988
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,8,4,balanced,0.09167743682861328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,4,8,power_law_1.2,0.0782476806640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,8,4,balanced,0.10224639892578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,8,4,balanced,0.128154878616333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,8,4,balanced,0.15000448226928711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,8,4,balanced,0.21509376525878907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,8,4,balanced,0.260445442199707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,8,4,balanced,0.3831321716308594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,8,4,balanced,0.4921702575683594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,8,4,balanced,0.7191385650634766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,8,4,balanced,0.9466828918457031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,4,8,power_law_1.2,0.10355711936950684
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,4,8,power_law_1.2,0.14050944328308107
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,4,8,power_law_1.2,0.21403135299682616
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,4,8,power_law_1.2,0.28233472824096684
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,4,8,power_law_1.2,0.35025409698486326
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,4,8,power_law_1.2,0.5649817657470704
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,8,4,balanced,0.022355198860168457
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,8,4,balanced,0.03261440038681031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,8,4,balanced,0.03271039962768555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,8,4,balanced,0.032967679500579834
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,8,4,balanced,0.02226047992706299
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,8,4,balanced,0.03287807941436768
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,8,4,balanced,0.03331072092056274
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,8,4,balanced,0.0329638409614563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,8,4,balanced,0.034757120609283446
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,8,4,balanced,0.034849278926849366
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,8,4,balanced,0.03624576091766357
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,8,4,balanced,0.03889152050018311
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,8,4,balanced,0.04118783950805664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,8,4,balanced,0.04679296016693115
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,8,4,balanced,0.05114880084991456
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,8,4,balanced,0.023567359447479248
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,8,4,balanced,0.025940480232238772
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,8,4,balanced,0.0317196798324585
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,8,4,balanced,0.031777279376983644
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,8,4,balanced,0.03267839908599853
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,8,4,balanced,0.03176192045211792
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,8,4,power_law_1.01,0.028405759334564205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,8,4,power_law_1.01,0.04471551895141602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,8,4,power_law_1.01,0.033808639049530034
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,8,4,power_law_1.01,0.0493017578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,8,4,power_law_1.01,0.05441664218902588
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,8,4,power_law_1.01,0.06108160018920898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,8,4,power_law_1.01,0.0949824047088623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,8,4,power_law_1.01,0.09971455574035645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,8,4,power_law_1.01,0.10206975936889648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,8,4,power_law_1.01,0.10399104118347169
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,8,4,power_law_1.01,0.10954112052917479
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,8,4,power_law_1.01,0.09981184005737305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,8,4,power_law_1.01,0.10206975936889648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,8,4,power_law_1.01,0.10784895896911621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,8,4,power_law_1.01,0.11181568145751952
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,8,4,power_law_1.01,0.11554047584533692
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,8,4,power_law_1.01,0.12543744087219239
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,8,4,power_law_1.01,0.1457024002075195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,8,4,power_law_1.01,0.16640384674072267
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,8,4,power_law_1.01,0.20543743133544923
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,8,4,power_law_1.01,0.24080896377563477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,8,4,power_law_1.01,0.34273792266845704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,8,4,power_law_1.01,0.4454848098754883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,8,4,power_law_1.01,0.6041126251220703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,8,4,power_law_1.01,0.812569580078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,8,4,power_law_1.01,1.1717183685302734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,8,4,power_law_1.01,0.030099198818206784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,8,4,power_law_1.01,0.03414400100708008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,8,4,power_law_1.01,1.5152485656738282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,8,4,power_law_1.01,0.037171199321746826
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,8,4,power_law_1.01,0.04262784004211426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,8,4,power_law_1.01,0.03979903936386108
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,8,4,power_law_1.01,0.0563212776184082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,8,4,power_law_1.01,0.050623998641967774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,8,4,power_law_1.01,0.0592793607711792
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,8,4,power_law_1.01,0.06328192234039307
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,8,4,power_law_1.01,0.06550271987915039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,8,4,power_law_1.01,0.06262656211853027
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,8,4,power_law_1.01,0.06099199771881103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,8,4,power_law_1.01,0.06352767944335938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,8,4,power_law_1.01,0.06723072052001952
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,8,4,power_law_1.01,0.06932223796844482
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,8,4,power_law_1.01,0.07318912029266358
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,8,4,power_law_1.01,0.07894527912139893
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,8,4,power_law_1.01,0.09412863731384277
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,8,4,power_law_1.01,0.10818304061889647
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,8,4,power_law_1.01,0.13724800109863283
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,8,4,power_law_1.01,0.1641574478149414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,8,4,power_law_1.01,0.23976831436157225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,8,4,power_law_1.01,0.29788288116455075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,8,4,power_law_1.01,0.4023219299316406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,8,4,power_law_1.01,0.5373209762573242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,8,4,balanced,0.07140863895416259
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,8,4,balanced,0.07777664184570313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,8,4,power_law_1.01,0.7863091278076172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,8,4,power_law_1.01,0.024856319427490232
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,8,4,power_law_1.01,0.02727423906326294
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,8,4,power_law_1.01,0.027645440101623536
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,8,4,balanced,0.12902912139892578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,8,4,power_law_1.01,0.031709439754486084
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,8,4,power_law_1.01,0.029611520767211914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,8,4,power_law_1.01,1.038452453613281
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,8,4,balanced,0.10211711883544923
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,8,4,power_law_1.01,0.03147775888442993
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,8,4,power_law_1.01,0.03247872114181519
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,8,4,power_law_1.01,0.032880640029907225
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,8,4,power_law_1.01,0.03301759958267212
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,8,4,power_law_1.01,0.0337932801246643
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,8,4,power_law_1.01,0.03471231937408447
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,8,4,power_law_1.01,0.03572096109390259
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,8,4,power_law_1.01,0.02323456048965454
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,8,4,power_law_1.01,0.0377126407623291
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,8,4,balanced,0.17903104782104493
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,8,4,power_law_1.01,0.0371340799331665
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,8,4,balanced,0.2330291175842285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,8,4,power_law_1.01,0.048546562194824214
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,8,4,power_law_1.01,0.07524096012115479
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,8,4,power_law_1.01,0.042975997924804686
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,8,4,power_law_1.01,0.06736000061035156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,8,4,power_law_1.01,0.0990015983581543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,8,4,power_law_1.2,0.028491520881652833
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,8,4,power_law_1.2,0.0402675199508667
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,8,4,power_law_1.01,0.06000639915466308
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,8,4,power_law_1.2,0.043740158081054685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,8,4,power_law_1.2,0.0474073600769043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,8,4,power_law_1.2,0.05115007877349853
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,8,4,power_law_1.2,0.061090559959411615
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,8,4,power_law_1.2,0.09504256248474122
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,8,4,power_law_1.2,0.09921024322509767
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,8,4,power_law_1.2,0.10306559562683107
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,8,4,power_law_1.2,0.10455295562744141
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,8,4,power_law_1.01,0.13567999839782713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,8,4,power_law_1.2,0.1003711986541748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,8,4,power_law_1.2,0.10070272445678712
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,8,4,power_law_1.2,0.10417920112609864
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,8,4,power_law_1.2,0.10782079696655274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,8,4,power_law_1.2,0.11414015769958497
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,8,4,power_law_1.01,0.1509171199798584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,8,4,power_law_1.2,0.11743871688842773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,8,4,power_law_1.2,0.1269977569580078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,8,4,power_law_1.2,0.14606719970703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,8,4,power_law_1.2,0.16668159484863282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,8,4,power_law_1.2,0.20679807662963867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,8,4,power_law_1.2,0.24883968353271485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,8,4,power_law_1.2,0.34970752716064457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,8,4,power_law_1.2,0.4619801712036133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,8,4,power_law_1.2,0.6342438507080078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,8,4,power_law_1.2,0.030446081161499022
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,8,4,power_law_1.2,0.03256319999694825
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,8,4,power_law_1.2,0.03871360063552857
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,8,4,power_law_1.2,0.04253183841705323
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,8,4,power_law_1.2,0.043956480026245116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,8,4,power_law_1.2,0.8034086608886719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,8,4,power_law_1.2,0.05849984169006348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,8,4,power_law_1.2,0.06032767772674561
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,8,4,power_law_1.2,0.06241663932800293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,8,4,power_law_1.2,0.06246655941009521
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,8,4,power_law_1.2,0.06393727779388428
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,8,4,power_law_1.2,0.061423358917236326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,8,4,power_law_1.2,0.063438081741333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,8,4,power_law_1.2,0.0681113576889038
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,8,4,power_law_1.2,0.07161087989807129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,8,4,power_law_1.2,0.07394688129425049
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,8,4,power_law_1.2,1.2435379028320312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,8,4,power_law_1.2,0.08035712242126465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,8,4,power_law_1.2,0.09405823707580566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,8,4,power_law_1.2,0.10963071823120116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,8,4,power_law_1.2,0.1389798355102539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,8,4,power_law_1.2,0.1711167907714844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,8,4,power_law_1.2,0.0523199987411499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,8,4,power_law_1.2,0.23975040435791017
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,8,4,power_law_1.2,0.3015974426269531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,8,4,power_law_1.2,0.42466049194335936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,8,4,power_law_1.2,0.7805593872070313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,8,4,power_law_1.2,0.0235532808303833
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,8,4,power_law_1.2,0.02388351917266846
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,8,4,power_law_1.2,0.02594815969467163
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,8,4,power_law_1.2,0.026849279403686522
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,8,4,power_law_1.2,0.028487679958343504
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,8,4,power_law_1.2,0.031496961116790775
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,8,4,power_law_1.2,0.03205888032913208
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,8,4,power_law_1.01,0.21226623535156247
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,8,4,power_law_1.2,0.032636160850524905
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,8,4,power_law_1.2,0.03264512062072754
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,8,4,power_law_1.2,0.033786880970001223
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,8,4,power_law_1.2,0.034106879234313964
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,8,4,power_law_1.2,0.03543807983398438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,8,4,power_law_1.2,0.5577292633056641
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,8,4,power_law_1.2,0.03603327989578247
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,8,4,power_law_1.2,0.04358528137207031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,8,4,power_law_1.2,0.0467136001586914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,8,4,power_law_1.01,0.31049856185913083
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,8,4,power_law_1.2,0.0526259183883667
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,8,4,power_law_1.2,0.06250112056732178
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,8,4,power_law_1.01,0.4328192138671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,8,4,power_law_1.01,0.6459635162353516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,8,4,power_law_1.2,1.0700466918945313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,8,4,power_law_1.2,1.5723635864257812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,8,4,power_law_1.2,0.04081535816192627
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,8,4,power_law_1.2,0.07803264141082764
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,8,4,power_law_1.2,0.12528512001037598
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,16,2,balanced,0.026036479473114015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,16,2,balanced,0.026179840564727785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,16,2,balanced,0.029697279930114746
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,8,4,power_law_1.2,0.0992959976196289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,16,2,balanced,0.04821248054504394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,16,2,balanced,0.0646988821029663
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,16,2,balanced,0.06646912097930908
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,16,2,balanced,0.0673804807662964
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,16,2,balanced,0.06783999919891356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,16,2,balanced,0.10410880088806151
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,16,2,balanced,0.10496895790100098
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,16,2,balanced,0.10635904312133788
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,16,2,balanced,0.11111040115356445
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,8,4,power_law_1.2,0.1612224006652832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,16,2,balanced,0.1136627197265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,16,2,balanced,0.11960576057434083
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,16,2,balanced,0.12107263565063478
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,16,2,balanced,0.1261017608642578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,16,2,balanced,0.1339583969116211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,16,2,balanced,0.15361536026000977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,16,2,balanced,0.17211904525756835
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,8,4,power_law_1.2,0.21893247604370117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,16,2,balanced,0.21138431549072267
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,16,2,balanced,0.24778367996215822
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,16,2,balanced,0.3244057464599609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,16,2,balanced,0.4003519821166992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,16,2,balanced,0.5935308837890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,16,2,balanced,0.027566080093383787
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,16,2,balanced,0.7590835571289063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,16,2,balanced,0.02930304050445557
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,16,2,balanced,0.03551487922668457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,16,2,balanced,0.04462975978851318
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,16,2,balanced,0.05795968055725098
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,16,2,balanced,1.0989888000488282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,16,2,balanced,0.05652991771697998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,16,2,balanced,0.05634560108184814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,16,2,balanced,0.05793920040130616
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,16,2,balanced,0.07067264080047607
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,16,2,balanced,0.07327744007110595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,16,2,balanced,0.07286655902862549
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,16,2,balanced,0.07442815780639649
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,16,2,balanced,1.4365670776367188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,16,2,balanced,0.0790771198272705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,16,2,balanced,0.08229887962341309
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,16,2,balanced,0.08415616035461426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,16,2,balanced,0.08661503791809082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,16,2,balanced,0.09198719978332519
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,16,2,balanced,0.10310400009155274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,16,2,balanced,0.11271552085876464
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,16,2,balanced,0.13532671928405762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,16,2,balanced,0.15879679679870606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,16,2,balanced,0.2310041618347168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,16,2,balanced,0.27492351531982423
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,16,2,balanced,0.40336639404296876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,16,2,balanced,0.5188966369628906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,16,2,balanced,0.7587725067138672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,16,2,balanced,1.0022297668457032
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,8,4,power_law_1.2,0.32481918334960935
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,8,4,power_law_1.2,0.4046988677978516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,16,2,balanced,0.030059518814086916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,16,2,balanced,0.055239682197570805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,16,2,balanced,0.05549439907073974
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,16,2,balanced,0.05539968013763428
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,16,2,balanced,0.030502400398254394
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,16,2,balanced,0.05558271884918213
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,16,2,balanced,0.05672832012176514
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,16,2,balanced,0.056643838882446285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,16,2,balanced,0.059022078514099116
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,16,2,balanced,0.0593612813949585
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,16,2,balanced,0.03256448030471802
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,16,2,balanced,0.06034175872802734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,16,2,balanced,0.059422721862792974
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,16,2,balanced,0.06234752178192139
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,16,2,balanced,0.0374617600440979
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,16,2,balanced,0.06745728015899657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,16,2,balanced,0.07596543788909912
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,16,2,balanced,0.08336128234863281
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,16,2,balanced,0.06754816055297852
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,16,2,balanced,0.053958401679992676
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,16,2,balanced,0.05463551998138427
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,8,4,power_law_1.2,0.5868620681762695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,16,2,power_law_1.01,0.027246079444885253
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,16,2,power_law_1.01,0.036614398956298824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,16,2,power_law_1.01,0.04486144065856933
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,16,2,power_law_1.01,0.04994688034057617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,16,2,power_law_1.01,0.05352831840515136
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,16,2,power_law_1.01,0.06413440227508545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,16,2,power_law_1.01,0.06442368030548096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,16,2,power_law_1.01,0.06632575988769532
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,8,4,power_law_1.2,0.8704819488525392
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,16,2,power_law_1.01,0.1034342384338379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,16,2,power_law_1.01,0.10384127616882324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,16,2,power_law_1.01,0.10579839706420899
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,16,2,power_law_1.01,0.10968064308166503
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,16,2,power_law_1.01,0.11259136199951172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,16,2,power_law_1.01,0.11837823867797852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,16,2,power_law_1.01,0.12191871643066406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,16,2,power_law_1.01,0.1271168041229248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,16,2,power_law_1.01,0.1364352035522461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,16,2,power_law_1.01,0.15608320236206055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,16,2,power_law_1.01,0.17652351379394532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,16,2,power_law_1.01,0.2147135925292969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,16,2,power_law_1.01,0.252488956451416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,16,2,power_law_1.01,0.3371443176269531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,16,2,power_law_1.01,0.43215103149414064
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,16,2,balanced,0.12060031890869141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,16,2,power_law_1.01,0.610904312133789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,16,2,power_law_1.01,0.7903065490722656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,16,2,balanced,0.19045631408691405
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,16,2,balanced,0.13385343551635742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,16,2,power_law_1.01,0.028409600257873535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,16,2,power_law_1.01,0.03232255935668945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,16,2,power_law_1.01,1.1471936035156252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,16,2,power_law_1.01,0.03783040046691895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,16,2,power_law_1.01,0.04561535835266113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,16,2,power_law_1.01,0.04904704093933106
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,16,2,power_law_1.01,0.0563212776184082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,16,2,power_law_1.01,0.05482880115509033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,16,2,power_law_1.01,0.05665919780731201
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,16,2,power_law_1.01,0.06819456100463868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,16,2,power_law_1.01,1.5006297302246092
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,16,2,balanced,0.24049407958984376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,16,2,power_law_1.01,0.06836607933044433
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,16,2,power_law_1.01,0.07140992164611817
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,16,2,power_law_1.01,0.07681663990020751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,16,2,power_law_1.01,0.07763199806213379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,16,2,power_law_1.01,0.08230143547058105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,16,2,power_law_1.01,0.08361087799072266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,16,2,power_law_1.01,0.08847488403320312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,16,2,power_law_1.01,0.09464192390441895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,16,2,power_law_1.01,0.10798463821411133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,16,2,power_law_1.01,0.12369919776916505
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,16,2,power_law_1.01,0.15280896186828613
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,16,2,balanced,0.3414067077636719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,16,2,power_law_1.01,0.1846054458618164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,16,2,power_law_1.01,0.2601971244812012
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,16,2,power_law_1.01,0.31739904403686525
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,16,2,power_law_1.01,0.43627902984619144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,16,2,power_law_1.01,0.5713561630249023
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,16,2,power_law_1.01,0.03582207918167114
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,16,2,power_law_1.01,0.03924223899841309
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,16,2,power_law_1.01,0.04171008110046387
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,16,2,power_law_1.01,0.051330561637878414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,16,2,power_law_1.01,0.05355904102325439
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,16,2,power_law_1.01,1.0689651489257812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,16,2,power_law_1.01,0.054814720153808595
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,16,2,power_law_1.01,0.055413761138916016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,16,2,power_law_1.01,0.05646336078643799
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,16,2,power_law_1.01,0.05704576015472412
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,16,2,power_law_1.01,0.05751423835754395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,16,2,power_law_1.01,0.05941887855529785
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,16,2,power_law_1.01,0.06337535858154297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,16,2,power_law_1.01,0.06483712196350097
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,16,2,power_law_1.01,0.07745920181274414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,16,2,power_law_1.01,0.09333503723144532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,16,2,power_law_1.01,0.8227174377441406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,16,2,balanced,0.4393766403198242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,16,2,power_law_1.01,0.03128448009490967
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,16,2,power_law_1.01,0.03258752107620239
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,16,2,power_law_1.01,0.10920063972473144
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,16,2,power_law_1.01,0.07151487827301026
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,16,2,power_law_1.01,0.1594163227081299
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,16,2,power_law_1.01,0.24604543685913086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,16,2,power_law_1.2,0.025088000297546386
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,16,2,power_law_1.2,0.04028672218322754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,16,2,power_law_1.2,0.04413951873779297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,16,2,power_law_1.2,0.049455361366271974
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,16,2,power_law_1.2,0.05214079856872559
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,16,2,power_law_1.2,0.06743040084838867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,16,2,power_law_1.2,0.06428671836853027
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,16,2,power_law_1.2,0.06640128135681153
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,16,2,power_law_1.2,0.10303615570068358
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,16,2,power_law_1.01,0.13558143615722656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,16,2,power_law_1.2,0.10376832008361817
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,16,2,power_law_1.01,0.2756876754760742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,16,2,power_law_1.2,0.10553728103637697
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,16,2,power_law_1.2,0.11056768417358398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,16,2,power_law_1.2,0.11193087577819824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,16,2,power_law_1.2,0.11957119941711425
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,16,2,power_law_1.2,0.12212736129760741
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,16,2,power_law_1.2,0.1273203182220459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,16,2,power_law_1.2,0.13675264358520506
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,16,2,power_law_1.2,0.156757755279541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,16,2,power_law_1.2,0.17608064651489258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,16,2,power_law_1.2,0.21485567092895508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,16,2,power_law_1.2,0.2536294364929199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,16,2,power_law_1.2,0.3395212936401367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,16,2,power_law_1.2,0.4372915267944336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,16,2,power_law_1.2,0.7936473846435547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,16,2,power_law_1.2,0.029304320812225344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,16,2,power_law_1.2,0.6115315246582032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,16,2,power_law_1.2,0.032983040809631346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,16,2,power_law_1.2,0.03684607982635498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,16,2,power_law_1.2,0.04599552154541016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,16,2,power_law_1.2,0.048358402252197265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,16,2,power_law_1.2,0.059595518112182624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,16,2,power_law_1.2,0.057401599884033205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,16,2,power_law_1.2,0.05567615985870361
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,16,2,power_law_1.2,0.06975743770599366
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,16,2,power_law_1.2,0.07139455795288085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,16,2,power_law_1.2,0.074268159866333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,16,2,power_law_1.2,0.07514495849609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,16,2,power_law_1.2,0.07852799892425537
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,16,2,power_law_1.2,0.0833011245727539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,16,2,power_law_1.2,0.08661375999450684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,16,2,power_law_1.2,0.08946432113647461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,16,2,power_law_1.2,0.0958028793334961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,16,2,power_law_1.2,0.1105395221710205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,16,2,power_law_1.2,0.12567551612854005
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,16,2,power_law_1.2,0.18674560546875002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,16,2,power_law_1.2,0.15575807571411132
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,16,2,power_law_1.2,0.26013439178466796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,16,2,power_law_1.2,1.1611724853515626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,16,2,power_law_1.2,0.3188991928100586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,16,2,power_law_1.2,0.4369817733764648
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,16,2,power_law_1.01,0.3927872085571289
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,16,2,power_law_1.2,0.03127935886383056
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,16,2,power_law_1.2,0.032569599151611325
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,16,2,power_law_1.01,0.5729331207275391
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,16,2,power_law_1.2,0.03633536100387573
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,16,2,power_law_1.2,0.037557759284973145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,16,2,power_law_1.2,0.5775475311279297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,16,2,power_law_1.2,0.040447998046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,16,2,power_law_1.2,0.05175936222076416
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,16,2,power_law_1.2,0.05412479877471924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,16,2,power_law_1.2,1.0817600250244142
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,16,2,power_law_1.2,0.055082240104675294
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,16,2,power_law_1.2,0.05531007766723632
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,16,2,power_law_1.2,0.056954879760742184
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,16,2,power_law_1.2,0.05750016212463379
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,16,2,power_law_1.2,0.061178879737854006
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,16,2,power_law_1.2,0.06106368064880371
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,16,2,power_law_1.2,0.06868095874786377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,16,2,power_law_1.2,0.8052902221679688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,16,2,power_law_1.01,0.770561294555664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,16,2,power_law_1.01,1.0212210845947265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,16,2,power_law_1.2,1.5159091186523437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,16,2,power_law_1.2,0.07364863872528075
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,16,2,power_law_1.2,0.09046143531799317
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,16,2,power_law_1.2,0.1630131149291992
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,16,2,power_law_1.2,0.13509632110595704
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,16,2,power_law_1.2,0.11152640342712403
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,16,2,power_law_1.2,0.0786956787109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,32,1,balanced,0.03410559892654419
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,32,1,balanced,0.03449408054351807
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,32,1,balanced,0.04405248165130615
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,32,1,balanced,0.0612505578994751
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,16,2,power_law_1.2,0.21228288650512694
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,32,1,balanced,0.08609984397888185
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,32,1,balanced,0.08736319541931152
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,32,1,balanced,0.08904000282287597
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,32,1,balanced,0.09017919540405274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,32,1,balanced,0.0904748821258545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,32,1,balanced,0.09240320205688476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,32,1,balanced,0.09344639778137206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,32,1,balanced,0.14464127540588378
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,32,1,balanced,0.15218175888061522
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,32,1,balanced,0.15501376152038576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,32,1,balanced,0.15881855964660646
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,32,1,balanced,0.14793024063110352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,32,1,balanced,0.16724416732788086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,32,1,balanced,0.18590015411376953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,32,1,balanced,0.2032236862182617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,32,1,balanced,0.24551488876342775
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,32,1,balanced,0.2794745635986328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,32,1,balanced,0.35943809509277347
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,32,1,balanced,0.4507180786132812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,32,1,balanced,0.6408383941650391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,32,1,balanced,0.8111398315429688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,32,1,balanced,1.187109146118164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,32,1,balanced,0.03655296087265015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,16,2,power_law_1.2,0.3024716758728027
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,32,1,balanced,0.041461758613586426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,32,1,balanced,0.0643116807937622
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,32,1,balanced,1.5547795104980469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,32,1,balanced,0.09451135635375976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,32,1,balanced,0.09350527763366699
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,32,1,balanced,0.094136962890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,32,1,balanced,0.09304191589355468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,32,1,balanced,0.09476479530334472
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,32,1,balanced,0.09704383850097656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,32,1,balanced,0.09904191970825196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,32,1,balanced,0.048474879264831544
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,32,1,balanced,0.11605695724487304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,32,1,balanced,0.11889408111572267
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,32,1,balanced,0.12363648414611816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,32,1,balanced,0.12474752426147462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,32,1,balanced,0.12806655883789062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,32,1,balanced,0.13345215797424317
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,32,1,balanced,0.14356608390808107
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,32,1,balanced,0.15334272384643555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,32,1,balanced,0.17291967391967772
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,32,1,balanced,0.19565759658813478
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,32,1,balanced,0.3068556785583496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,32,1,balanced,0.34670654296875003
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,32,1,balanced,0.5019718551635742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,32,1,balanced,0.6520223999023438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,32,1,power_law_1.01,0.03327744007110596
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,32,1,balanced,0.9504844665527343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,32,1,power_law_1.01,0.03759104013442993
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,32,1,power_law_1.01,0.05399680137634277
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,32,1,power_law_1.01,0.05964992046356201
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,32,1,balanced,1.2374521636962892
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,32,1,power_law_1.01,0.06685503959655761
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,32,1,power_law_1.01,0.08116352081298828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,32,1,power_law_1.01,0.08315008163452149
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,32,1,power_law_1.01,0.08499456405639648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,32,1,power_law_1.01,0.0899404811859131
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,32,1,power_law_1.01,0.09135935783386231
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,32,1,power_law_1.01,0.09588224411010741
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,32,1,power_law_1.01,0.14457088470458984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,32,1,power_law_1.01,0.14661952018737795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,32,1,power_law_1.01,0.1538924789428711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,32,1,power_law_1.01,0.15633919715881348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,32,1,power_law_1.01,0.16208255767822266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,32,1,power_law_1.01,0.17086591720581054
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,32,1,power_law_1.01,0.19190208435058592
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,32,1,power_law_1.01,0.20965951919555664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,32,1,power_law_1.01,0.24971136093139648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,32,1,power_law_1.01,0.2918079948425293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,32,1,power_law_1.01,0.3834918212890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,32,1,power_law_1.01,0.49372798919677735
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,16,2,power_law_1.2,0.5088601684570312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,32,1,power_law_1.01,0.8487731170654296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,32,1,power_law_1.01,0.035925118923187255
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,16,2,power_law_1.2,0.406426887512207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,32,1,power_law_1.01,0.040628480911254886
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,32,1,power_law_1.01,1.2173760223388672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,32,1,power_law_1.01,0.04884031772613525
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,32,1,power_law_1.01,0.05908991813659668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,32,1,power_law_1.01,0.07907008171081543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,32,1,power_law_1.01,0.08962559700012207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,32,1,power_law_1.01,0.08982527732849122
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,32,1,power_law_1.01,0.09209792137145996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,32,1,power_law_1.01,0.09539263725280762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,32,1,power_law_1.01,0.6615570831298828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,32,1,power_law_1.01,0.09905856132507325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,32,1,power_law_1.01,0.11591296195983887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,32,1,power_law_1.01,0.11804351806640626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,32,1,power_law_1.01,0.1234489631652832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,32,1,power_law_1.01,0.12695039749145506
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,32,1,power_law_1.01,0.12954367637634276
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,32,1,power_law_1.01,0.14055551528930663
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,32,1,power_law_1.01,0.15541312217712402
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,32,1,power_law_1.01,0.09470144271850586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,32,1,power_law_1.01,0.1833216094970703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,32,1,power_law_1.01,0.21618560791015623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,32,1,power_law_1.01,0.25496511459350585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,32,1,power_law_1.01,0.3223596954345703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,32,1,power_law_1.01,0.40388542175292963
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,32,1,power_law_1.01,0.5557088088989258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,32,1,power_law_1.01,1.5837472534179686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,32,1,power_law_1.01,0.7001241302490234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,32,1,power_law_1.2,0.043055357933044436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,32,1,power_law_1.2,0.05425856113433838
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,32,1,power_law_1.01,0.9864723205566406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,32,1,power_law_1.2,0.05959807872772217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,32,1,power_law_1.2,0.06640128135681153
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,32,1,power_law_1.2,0.08510592460632324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,32,1,power_law_1.2,0.08743935585021972
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,32,1,power_law_1.01,1.282099151611328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,32,1,power_law_1.2,0.09145536422729492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,32,1,power_law_1.2,0.09235648155212403
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,32,1,power_law_1.2,0.0963321590423584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,32,1,power_law_1.2,0.03369983911514282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,32,1,power_law_1.2,0.1446668815612793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,32,1,power_law_1.2,0.14799360275268553
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,32,1,power_law_1.2,0.15389568328857423
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,32,1,power_law_1.2,0.15728960037231446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,32,1,power_law_1.2,0.1619705581665039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,32,1,power_law_1.2,0.1715763282775879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,32,1,power_law_1.2,0.19197120666503906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,32,1,power_law_1.2,0.21049663543701175
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,32,1,power_law_1.2,0.2518649673461914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,32,1,power_law_1.2,0.08094143867492676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,32,1,power_law_1.2,0.29244543075561524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,32,1,power_law_1.2,0.38438465118408205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,32,1,power_law_1.2,0.4961011123657227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,32,1,power_law_1.2,0.0363923192024231
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,32,1,power_law_1.2,0.04147840023040771
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,32,1,power_law_1.2,0.049375357627868655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,32,1,power_law_1.2,0.05952191829681397
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,16,2,power_law_1.2,0.7768307495117187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,32,1,power_law_1.2,0.0782969617843628
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,16,2,power_law_1.2,1.16023681640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,32,1,power_law_1.2,0.09068096160888672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,32,1,power_law_1.2,0.08928383827209473
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,32,1,power_law_1.2,0.09165375709533692
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,32,1,power_law_1.2,0.09477952003479004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,32,1,power_law_1.2,0.09416319847106933
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,32,1,power_law_1.2,0.09570367813110352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,32,1,power_law_1.2,0.11833024024963379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,32,1,power_law_1.2,0.11995712280273438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,32,1,power_law_1.2,0.12591423988342285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,32,1,power_law_1.2,0.6632575988769531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,32,1,power_law_1.2,0.12866496086120605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,32,1,power_law_1.2,0.13283328056335447
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,32,1,power_law_1.2,0.14214271545410156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,32,1,power_law_1.2,0.1658937644958496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,32,1,power_law_1.2,0.18707775115966796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,32,1,power_law_1.2,0.22402624130249021
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,32,1,power_law_1.2,0.25775871276855467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,32,1,power_law_1.2,0.3242335891723633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,32,1,power_law_1.2,0.40961471557617185
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,32,1,power_law_1.2,0.5561612701416015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,1,32,balanced,0.06556416034698487
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,1,32,balanced,0.06591360092163086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,1,32,balanced,0.06575615882873535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,1,32,balanced,0.07754240036010743
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,1,32,balanced,0.09384063720703126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,32,1,power_law_1.2,0.8510438537597655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,1,32,balanced,0.15734784126281737
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,1,32,balanced,0.15854975700378418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,1,32,balanced,0.1594278430938721
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,1,32,balanced,0.15929727554321288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,1,32,balanced,0.16099456787109373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,1,32,balanced,0.16780160903930663
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,1,32,balanced,0.16849151611328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,1,32,balanced,0.1721062469482422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,1,32,balanced,0.1757619285583496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,1,32,balanced,0.17824127197265624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,1,32,balanced,0.1823347282409668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,1,32,balanced,0.18131328582763673
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,1,32,balanced,0.19380735397338866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,1,32,balanced,0.20999040603637695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,1,32,balanced,0.27897344589233397
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,1,32,balanced,0.3112281608581543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,1,32,balanced,0.40375553131103514
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,1,32,balanced,0.46579967498779296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,1,32,balanced,0.598967056274414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,1,32,balanced,0.7625357055664063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,32,1,power_law_1.2,0.7025215911865235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,1,32,balanced,1.0707481384277344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,1,32,balanced,0.03822463989257813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,1,32,balanced,0.0375219202041626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,1,32,balanced,0.03827199935913086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,1,32,balanced,0.04768256187438965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,1,32,balanced,0.059953918457031255
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,1,32,balanced,1.3847001647949218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,1,32,balanced,0.07068927764892578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,1,32,balanced,0.07383808135986328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,1,32,balanced,0.07184256076812744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,1,32,balanced,0.08046208381652832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,1,32,balanced,0.07679743766784668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,1,32,balanced,0.0768998384475708
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,1,32,balanced,0.08478336334228516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,16,2,power_law_1.2,1.548071746826172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,1,32,balanced,0.086178560256958
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,1,32,balanced,0.09055232048034668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,1,32,balanced,0.07311488151550292
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,1,32,balanced,0.09388287544250488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,1,32,balanced,0.0980684757232666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,1,32,balanced,0.1108403205871582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,1,32,balanced,0.12516351699829104
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,1,32,balanced,0.07879168033599854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,1,32,balanced,0.17568767547607422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,1,32,balanced,0.20017919540405274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,1,32,balanced,0.2732671928405762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,1,32,balanced,0.3377139282226563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,1,32,balanced,0.08577024459838867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,1,32,balanced,0.4590131378173828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,1,32,balanced,0.5861183929443359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,32,1,power_law_1.2,1.2198035430908203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,32,1,power_law_1.2,0.9859552001953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,1,32,balanced,0.02145535945892334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,1,32,balanced,0.043147521018981935
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,1,32,balanced,0.02190592050552368
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,1,32,balanced,0.04340864181518554
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,1,32,balanced,0.022250239849090574
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,1,32,balanced,0.04358399868011474
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,1,32,balanced,0.043380479812622066
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,1,32,balanced,0.04365183830261231
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,1,32,balanced,0.044499201774597166
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,1,32,balanced,0.04427775859832764
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,1,32,balanced,0.04434559822082519
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,1,32,balanced,0.04443647861480713
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,1,32,balanced,0.04436607837677002
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,1,32,balanced,0.044605441093444824
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,1,32,balanced,0.045283842086791995
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,1,32,balanced,0.04511616230010986
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,1,32,balanced,0.04610176086425781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,1,32,balanced,0.0476364803314209
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,1,32,balanced,0.021532158851623535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,32,1,power_law_1.2,1.5879104614257813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,1,32,balanced,0.040943360328674315
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,1,32,balanced,0.04329343795776367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,1,32,power_law_1.01,0.10471551895141602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,1,32,power_law_1.01,0.10461695671081543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,1,32,power_law_1.01,0.10557824134826661
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,1,32,power_law_1.01,0.15574015617370604
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,1,32,power_law_1.01,0.15273088455200196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,1,32,power_law_1.01,0.16038528442382813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,32,1,power_law_1.2,1.2816812133789062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,1,32,power_law_1.01,0.16344192504882812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,1,32,power_law_1.01,0.16531328201293943
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,1,32,power_law_1.01,0.16854015350341797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,1,32,power_law_1.01,0.16637439727783204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,1,32,power_law_1.01,0.17177215576171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,1,32,power_law_1.01,0.1762726402282715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,1,32,power_law_1.01,0.1800934410095215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,1,32,power_law_1.01,0.1844211196899414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,1,32,power_law_1.01,0.19377792358398438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,1,32,power_law_1.01,0.19780096054077148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,1,32,power_law_1.01,0.21781503677368166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,1,32,power_law_1.01,0.304586238861084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,1,32,power_law_1.01,0.3542835235595703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,1,32,power_law_1.01,0.43704830169677733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,1,32,power_law_1.01,0.5625779342651367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,1,32,power_law_1.01,0.7887193298339843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,1,32,power_law_1.01,1.0124185943603516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,1,32,power_law_1.01,1.563910369873047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,1,32,power_law_1.01,2.010527954101563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,1,32,power_law_1.01,3.3476800537109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,1,32,power_law_1.01,0.06539008140563965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,1,32,power_law_1.01,0.06490496158599854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,1,32,power_law_1.01,0.06631296157836915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,1,32,power_law_1.01,0.06531968116760253
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,1,32,power_law_1.01,0.0637721586227417
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,1,32,power_law_1.01,4.3755929565429685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,1,32,power_law_1.01,0.07253248214721679
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,1,32,power_law_1.01,0.07481215953826904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,1,32,power_law_1.01,0.08071040153503418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,1,32,power_law_1.01,0.07662975788116455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,1,32,power_law_1.01,0.07793920040130616
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,1,32,power_law_1.01,0.08141951560974121
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,1,32,balanced,0.05913599967956543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,1,32,power_law_1.01,0.08598143577575683
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,1,32,power_law_1.01,0.08718463897705078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,1,32,power_law_1.01,0.09119999885559082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,1,32,power_law_1.01,0.09109248161315918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,1,32,power_law_1.01,0.09173888206481934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,1,32,power_law_1.01,0.09945728302001952
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,1,32,balanced,0.07748095989227295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,1,32,power_law_1.01,0.1179366397857666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,1,32,power_law_1.01,0.14964991569519043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,1,32,power_law_1.01,0.18613119125366212
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,1,32,power_law_1.01,0.22864383697509766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,1,32,power_law_1.01,0.3165145683288574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,1,32,power_law_1.01,0.39570945739746094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,1,32,balanced,0.09702015876770019
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,1,32,power_law_1.01,0.5458995056152344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,1,32,power_law_1.01,0.04284031867980957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,1,32,power_law_1.01,0.7449523162841797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,1,32,power_law_1.01,0.043297281265258794
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,1,32,power_law_1.01,0.04304384231567383
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,1,32,power_law_1.01,0.034131200313568116
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,1,32,power_law_1.01,0.04344575881958008
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,1,32,power_law_1.01,0.042789120674133305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,1,32,power_law_1.01,1.103562240600586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,1,32,power_law_1.01,0.043946242332458495
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,1,32,power_law_1.01,0.04371967792510986
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,1,32,power_law_1.01,0.04467199802398682
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,1,32,power_law_1.01,0.043915519714355464
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,1,32,power_law_1.01,1.6098355102539064
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,1,32,power_law_1.01,0.04418303966522217
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,1,32,power_law_1.01,0.04448128223419189
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,1,32,balanced,0.061669120788574214
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,1,32,power_law_1.01,0.04525568008422852
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,1,32,power_law_1.01,0.04483327865600586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,1,32,power_law_1.01,0.04937856197357178
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,1,32,power_law_1.01,0.05837183952331543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,1,32,power_law_1.01,0.043486719131469724
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,1,32,balanced,0.11377792358398438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,1,32,balanced,0.14846847534179688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,1,32,power_law_1.01,0.0436902379989624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,1,32,power_law_1.01,0.06499711990356445
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,1,32,power_law_1.01,0.08216447830200195
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,1,32,power_law_1.01,0.1257151985168457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,1,32,power_law_1.2,0.105099515914917
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,1,32,power_law_1.01,0.047322878837585444
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,1,32,power_law_1.01,0.10377599716186522
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,1,32,power_law_1.2,0.10434816360473634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,1,32,power_law_1.2,0.10503552436828614
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,1,32,power_law_1.2,0.14616959571838378
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,1,32,power_law_1.2,0.1498969554901123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,1,32,power_law_1.2,0.15467647552490232
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,1,32,power_law_1.2,0.15912192344665527
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,1,32,power_law_1.2,0.16280576705932617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,1,32,power_law_1.2,0.167509765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,1,32,power_law_1.01,0.16802175521850587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,1,32,power_law_1.2,0.17575551986694335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,1,32,power_law_1.2,0.17887487411499023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,1,32,power_law_1.2,0.1849407958984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,1,32,power_law_1.2,0.18969728469848632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,1,32,power_law_1.2,0.2047641563415527
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,1,32,power_law_1.2,0.21659776687622073
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,1,32,power_law_1.2,0.23687168121337893
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,1,32,power_law_1.2,0.34470401763916014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,1,32,power_law_1.2,0.15508223533630372
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,1,32,power_law_1.2,0.4137484741210938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,1,32,power_law_1.2,0.5671142578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,1,32,power_law_1.2,0.662767333984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,1,32,power_law_1.2,1.1724390411376953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,1,32,power_law_1.01,0.2632383918762207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,1,32,power_law_1.2,1.3035276794433595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,1,32,power_law_1.2,0.06461184024810791
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,1,32,power_law_1.2,0.06503680229187012
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,1,32,power_law_1.2,0.0647052812576294
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,1,32,power_law_1.2,0.06468607902526854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,1,32,power_law_1.2,0.0622489595413208
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,1,32,power_law_1.2,0.07001984119415283
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,1,32,power_law_1.2,2.244340515136719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,1,32,power_law_1.2,0.07375103950500488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,1,32,power_law_1.2,0.07568640232086181
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,1,32,power_law_1.2,0.0737011194229126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,1,32,power_law_1.2,0.07745408058166504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,1,32,power_law_1.2,0.0813145637512207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,1,32,power_law_1.2,0.0861900806427002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,1,32,power_law_1.2,0.0904038429260254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,1,32,power_law_1.2,0.0903104019165039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,1,32,power_law_1.2,0.0920691204071045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,1,32,power_law_1.2,0.09760512351989746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,1,32,power_law_1.2,0.10823552131652832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,1,32,power_law_1.2,0.13920512199401855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,1,32,power_law_1.2,0.15860608100891113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,1,32,power_law_1.2,0.22366975784301757
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,1,32,power_law_1.2,0.25456640243530276
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,1,32,power_law_1.2,0.3898534393310547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,1,32,power_law_1.2,0.49327617645263666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,1,32,power_law_1.2,2.685321044921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,1,32,power_law_1.2,0.7573414611816406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,1,32,power_law_1.2,0.04314623832702637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,1,32,power_law_1.2,0.9064345550537108
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,1,32,power_law_1.2,0.04328320026397705
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,1,32,power_law_1.2,0.04294400215148926
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,1,32,power_law_1.2,0.03622015953063965
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,1,32,power_law_1.2,0.031946239471435545
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,1,32,power_law_1.2,0.04347008228302002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,1,32,power_law_1.2,4.312926635742188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,1,32,power_law_1.2,0.04360576152801514
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,1,32,power_law_1.2,0.04365183830261231
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,1,32,power_law_1.2,0.04237696170806885
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,1,32,power_law_1.2,0.043692798614501954
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,1,32,power_law_1.2,0.04430975914001465
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,1,32,power_law_1.2,0.04457727909088135
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,1,32,power_law_1.2,0.04444928169250488
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,1,32,power_law_1.2,0.04526080131530762
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,1,32,power_law_1.2,0.04867712020874023
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,1,32,power_law_1.2,0.04432384014129639
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,1,32,power_law_1.2,1.5582476806640624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,1,32,power_law_1.2,6.334730224609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,1,32,power_law_1.2,2.066603546142578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,1,32,power_law_1.01,0.35494400024414063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,1,32,power_law_1.01,0.48801536560058595
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,1,32,power_law_1.2,0.0467251205444336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,1,32,power_law_1.2,0.056655359268188474
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,1,32,power_law_1.2,0.06500991821289062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,1,32,power_law_1.2,0.08113408088684082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,2,16,balanced,0.035018239021301266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,2,16,balanced,0.03463295936584473
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,1,32,power_law_1.2,0.09546112060546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,2,16,balanced,0.06051455974578858
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,2,16,balanced,0.07100416183471679
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,2,16,balanced,0.09436032295227051
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,2,16,balanced,0.15592063903808592
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,2,16,balanced,0.16429183959960938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,2,16,balanced,0.1610483169555664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,2,16,balanced,0.16131839752197266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,2,16,balanced,0.15976063728332518
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,2,16,balanced,0.16335615158081054
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,1,32,power_law_1.2,0.12968576431274415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,2,16,balanced,0.1722368049621582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,2,16,balanced,0.17663488388061524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,2,16,balanced,0.17053951263427733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,2,16,balanced,0.1765158462524414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,2,16,balanced,0.1774707221984863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,2,16,balanced,0.18529792785644533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,2,16,balanced,0.20557056427001955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,2,16,balanced,0.22632192611694335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,2,16,balanced,0.27796735763549807
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,2,16,balanced,0.31197696685791015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,2,16,balanced,0.4524364852905274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,2,16,balanced,0.5391769790649414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,2,16,balanced,0.7192166137695313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,2,16,balanced,0.923135986328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,2,16,balanced,0.035403521060943605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,2,16,balanced,0.03543936014175415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,2,16,balanced,1.3200166320800781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,2,16,balanced,0.043576321601867675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,2,16,balanced,0.06306687831878663
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,2,16,balanced,1.7219354248046876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,2,16,balanced,0.07468416213989257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,2,16,balanced,0.07558784008026123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,2,16,balanced,0.07801727771759033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,2,16,balanced,0.0780518388748169
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,2,16,balanced,0.08010496139526367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,2,16,balanced,0.035526399612426755
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,2,16,balanced,0.08544128417968751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,2,16,balanced,0.0862169647216797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,2,16,balanced,0.08775296211242675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,2,16,balanced,0.08319616317749023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,2,16,balanced,0.08514176368713379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,2,16,balanced,0.08851712226867677
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,2,16,balanced,0.0926144027709961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,2,16,balanced,0.1007091236114502
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,2,16,balanced,0.1170368003845215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,2,16,balanced,0.1348966407775879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,2,16,balanced,0.1565068817138672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,2,16,balanced,0.2242483139038086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,2,16,balanced,0.26576255798339843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,2,16,balanced,0.3730636978149414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,2,16,balanced,0.46376575469970704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,2,16,balanced,0.6417305755615235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,2,16,balanced,0.8407462310791015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,1,32,power_law_1.2,0.1763327980041504
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,1,32,power_law_1.2,0.2553036880493164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,1,32,power_law_1.2,0.30968063354492187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,1,32,power_law_1.2,0.4809408187866211
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,2,16,balanced,0.020752639770507814
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,2,16,balanced,0.04489855766296387
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,2,16,balanced,0.02103935956954956
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,2,16,balanced,0.020286719799041748
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,2,16,balanced,0.0446131181716919
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,2,16,balanced,0.04469632148742676
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,1,32,power_law_1.2,0.7672064208984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,2,16,balanced,0.044800000190734865
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,2,16,balanced,0.04486527919769287
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,2,16,balanced,0.04476287841796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,2,16,balanced,0.045009918212890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,2,16,balanced,0.045227518081665044
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,2,16,balanced,0.04583424091339111
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,2,16,balanced,0.046440958976745605
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,2,16,balanced,0.047618560791015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,2,16,balanced,0.04905087947845459
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,2,16,balanced,0.050599679946899415
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,2,16,balanced,0.0408025598526001
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,2,16,balanced,0.023819520473480224
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,2,16,balanced,0.04454016208648682
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,2,16,balanced,0.044280319213867186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,2,16,power_law_1.01,0.0892966365814209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,2,16,power_law_1.01,0.10617728233337402
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,2,16,power_law_1.01,0.09909119606018066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,2,16,power_law_1.01,0.10619135856628419
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,2,16,power_law_1.01,0.135098876953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,2,16,power_law_1.01,0.14696576118469237
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,2,16,power_law_1.01,0.15683072090148925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,2,16,power_law_1.01,0.1557209587097168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,2,16,power_law_1.01,0.15460351943969725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,2,16,power_law_1.01,0.15597824096679686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,2,16,power_law_1.01,0.16196352005004883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,2,16,power_law_1.01,0.17382400512695312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,2,16,power_law_1.01,0.1705254364013672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,2,16,power_law_1.01,0.17723136901855469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,2,16,power_law_1.01,0.1841996765136719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,2,16,power_law_1.01,0.18609151840209961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,2,16,power_law_1.01,0.1997043228149414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,2,16,power_law_1.01,0.24554496765136719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,2,16,power_law_1.01,0.29090944290161136
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,2,16,balanced,0.06427264213562012
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,2,16,power_law_1.01,0.392492790222168
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,2,16,balanced,0.04489471912384033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,2,16,power_law_1.01,0.44357246398925787
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,2,16,balanced,0.06820223808288575
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,2,16,power_law_1.01,0.6491033935546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,2,16,power_law_1.01,0.7837529754638671
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,2,16,balanced,0.08789504051208497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,2,16,power_law_1.01,0.06133120059967041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,2,16,power_law_1.01,1.180253448486328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,2,16,power_law_1.01,0.06687744140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,2,16,power_law_1.01,0.06433023929595946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,2,16,power_law_1.01,1.4608934020996094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,2,16,power_law_1.01,0.06730751991271973
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,2,16,power_law_1.01,0.0622540807723999
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,2,16,power_law_1.01,0.07249152183532714
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,2,16,power_law_1.01,0.07445504188537597
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,2,16,power_law_1.01,0.07785855770111084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,2,16,power_law_1.01,0.07795328140258789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,2,16,power_law_1.01,0.0818892765045166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,2,16,power_law_1.01,0.0854041576385498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,2,16,power_law_1.01,2.389187774658203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,2,16,power_law_1.01,0.08944767951965332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,2,16,power_law_1.01,0.08379391670227052
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,2,16,power_law_1.01,0.08979711532592774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,2,16,power_law_1.01,0.08970111846923828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,2,16,power_law_1.01,0.09193216323852539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,2,16,power_law_1.01,0.09851263999938965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,2,16,power_law_1.01,0.12569600105285644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,2,16,power_law_1.01,0.13960448265075684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,2,16,power_law_1.01,0.1863859176635742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,2,16,power_law_1.01,0.20579967498779297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,2,16,power_law_1.01,0.29319551467895505
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,2,16,power_law_1.01,0.35794433593749997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,2,16,power_law_1.01,3.5100531005859374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,2,16,power_law_1.01,0.5185472106933593
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,2,16,power_law_1.01,0.041311998367309574
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,2,16,power_law_1.01,0.044026880264282225
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,2,16,power_law_1.01,0.04041600227355957
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,2,16,power_law_1.01,0.04227456092834473
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,2,16,power_law_1.01,0.032927999496459956
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,2,16,power_law_1.01,0.040826878547668456
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,2,16,power_law_1.01,1.0670706939697265
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,2,16,power_law_1.01,0.04312448024749756
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,2,16,power_law_1.01,0.04294784069061279
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,2,16,power_law_1.01,0.04352511882781983
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,2,16,power_law_1.01,0.04369664192199707
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,2,16,power_law_1.01,0.04428287982940674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,2,16,power_law_1.01,1.4212031555175781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,2,16,power_law_1.01,0.04457600116729736
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,2,16,power_law_1.01,0.04442751884460449
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,2,16,power_law_1.01,0.6831667327880859
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,2,16,power_law_1.01,0.04563583850860596
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,2,16,power_law_1.01,0.046337280273437496
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,2,16,power_law_1.01,0.05516799926757813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,2,16,balanced,0.10896639823913574
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,2,16,balanced,0.13364480018615724
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,2,16,balanced,0.17505535125732422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,2,16,power_law_1.01,0.07238783836364746
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,2,16,power_law_1.01,0.08148608207702637
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,2,16,power_law_1.01,0.04844416141510009
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,2,16,power_law_1.01,0.06017920017242432
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,2,16,power_law_1.01,0.04952703952789307
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,2,16,power_law_1.2,0.0916262435913086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,2,16,power_law_1.01,0.10005248069763184
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,2,16,power_law_1.2,0.09555071830749512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,2,16,power_law_1.2,0.10487168312072755
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,2,16,power_law_1.2,0.10700287818908691
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,2,16,power_law_1.2,0.14626303672790528
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,2,16,power_law_1.2,0.14608768463134764
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,2,16,power_law_1.2,0.15631872177124023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,2,16,power_law_1.2,0.15500288009643554
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,2,16,power_law_1.01,0.130513916015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,2,16,power_law_1.2,0.1544166374206543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,2,16,power_law_1.2,0.1548633575439453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,2,16,power_law_1.2,0.17790592193603516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,2,16,power_law_1.2,0.17304319381713867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,2,16,power_law_1.2,0.18182144165039063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,2,16,power_law_1.2,0.18648832321166992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,2,16,power_law_1.2,0.19303680419921876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,2,16,power_law_1.2,0.20402559280395507
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,2,16,power_law_1.2,0.27066879272460936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,2,16,power_law_1.2,0.31900159835815434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,2,16,power_law_1.2,0.42925567626953126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,2,16,power_law_1.2,0.17250688552856447
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,2,16,power_law_1.2,0.5470937728881836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,2,16,power_law_1.2,0.8119193267822264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,2,16,power_law_1.2,1.0507558441162108
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,2,16,power_law_1.2,1.5666598510742187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,2,16,power_law_1.2,0.0625497579574585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,2,16,power_law_1.2,2.0257472229003906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,2,16,power_law_1.2,0.0644595193862915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,2,16,power_law_1.2,0.06675327777862548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,2,16,power_law_1.2,0.06915840148925781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,2,16,power_law_1.2,0.06459136009216308
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,2,16,power_law_1.2,0.07728640079498292
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,2,16,power_law_1.2,0.08354432106018066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,2,16,power_law_1.2,0.07889408111572266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,2,16,power_law_1.2,0.0800767993927002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,2,16,power_law_1.2,0.08258815765380859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,2,16,power_law_1.2,0.08408320426940918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,2,16,power_law_1.2,0.09110527992248535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,2,16,power_law_1.2,0.08552960395812989
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,2,16,power_law_1.2,0.08624383926391602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,2,16,power_law_1.2,0.0919052791595459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,2,16,power_law_1.2,0.09513983726501465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,2,16,power_law_1.2,0.10221568107604981
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,2,16,power_law_1.2,0.12272128105163574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,2,16,power_law_1.2,0.16604671478271485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,2,16,power_law_1.2,0.20742399215698243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,2,16,power_law_1.2,0.2669683265686035
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,2,16,power_law_1.01,0.1926848030090332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,2,16,power_law_1.2,0.3577791976928711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,2,16,power_law_1.2,0.4582515335083007
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,2,16,power_law_1.2,0.6293427276611328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,2,16,power_law_1.2,0.881844482421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,2,16,power_law_1.2,0.04153855800628662
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,2,16,power_law_1.2,0.04257791996002197
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,2,16,power_law_1.2,0.04284287929534912
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,2,16,power_law_1.2,0.044094719886779786
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,2,16,power_law_1.2,0.03429888010025024
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,2,16,power_law_1.2,0.04360320091247559
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,2,16,power_law_1.2,1.212895965576172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,2,16,power_law_1.2,0.04306303977966309
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,2,16,power_law_1.2,0.04253824234008789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,2,16,power_law_1.2,0.04451839923858643
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,2,16,power_law_1.2,0.04371071815490722
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,2,16,power_law_1.2,0.04434815883636475
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,2,16,power_law_1.2,0.04492288112640381
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,2,16,power_law_1.2,0.04559360027313232
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,2,16,power_law_1.2,0.04564095973968506
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,2,16,power_law_1.2,0.04738048076629638
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,2,16,power_law_1.2,4.364080505371094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,2,16,power_law_1.2,0.05697279930114746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,2,16,power_law_1.2,3.0672357177734377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,2,16,power_law_1.2,1.6004978942871095
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,2,16,power_law_1.01,0.2362598419189453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,2,16,power_law_1.01,0.387720947265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,2,16,power_law_1.01,0.4347494506835938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,2,16,power_law_1.2,0.049200639724731446
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,2,16,power_law_1.2,0.04986495971679687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,2,16,power_law_1.2,0.06628352165222168
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,2,16,power_law_1.2,0.1049574375152588
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,4,8,balanced,0.033989119529724124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,4,8,balanced,0.03334399938583374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,4,8,balanced,0.0564031982421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,2,16,power_law_1.2,0.07902463912963867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,4,8,balanced,0.07214975833892823
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,4,8,balanced,0.09454591751098633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,4,8,balanced,0.149432315826416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,4,8,balanced,0.15293312072753906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,4,8,balanced,0.15684351921081544
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,4,8,balanced,0.16154624938964843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,4,8,balanced,0.1649279975891113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,4,8,balanced,0.1587763214111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,4,8,balanced,0.16261375427246094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,4,8,balanced,0.16703872680664061
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,4,8,balanced,0.17399679183959962
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,4,8,balanced,0.17764863967895508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,4,8,balanced,0.18561023712158203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,4,8,balanced,0.20178943634033203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,4,8,balanced,0.22802816390991212
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,4,8,balanced,0.2583897590637207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,4,8,balanced,0.3297292709350586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,4,8,balanced,0.3844416046142578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,4,8,balanced,0.5839731216430664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,4,8,balanced,0.7094131469726562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,4,8,balanced,0.9820198059082031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,4,8,balanced,1.285100860595703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,4,8,balanced,1.8640832519531252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,4,8,balanced,0.03298943996429443
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,4,8,balanced,0.03441663980484009
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,4,8,balanced,0.03824896097183227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,4,8,balanced,2.454106903076172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,4,8,balanced,0.04654848098754883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,4,8,balanced,0.06829951763153076
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,4,8,balanced,0.08757375717163086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,4,8,balanced,0.08905344009399414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,4,8,balanced,0.08956031799316407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,4,8,balanced,0.09336064338684083
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,4,8,balanced,0.09604864120483399
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,4,8,balanced,0.08769280433654784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,4,8,balanced,0.09020671844482422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,4,8,balanced,0.09355135917663573
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,4,8,balanced,0.09865983963012695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,4,8,balanced,0.09947903633117676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,4,8,balanced,0.10329983711242677
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,4,8,balanced,0.11598976135253906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,2,16,power_law_1.2,0.13229567527770997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,4,8,balanced,0.1321728038787842
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,4,8,balanced,0.15215231895446776
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,4,8,balanced,0.18673919677734374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,4,8,balanced,0.22363008499145506
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,2,16,power_law_1.2,0.18629375457763672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,4,8,balanced,0.32541694641113283
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,4,8,balanced,0.3976947021484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,4,8,balanced,0.5668928146362304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,4,8,balanced,0.7194329833984374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,4,8,balanced,1.029864959716797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,4,8,balanced,1.349951934814453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,2,16,power_law_1.2,0.24546815872192385
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,2,16,power_law_1.2,0.36018177032470705
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,2,16,power_law_1.2,0.696126708984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,2,16,power_law_1.2,0.5522867202758789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,4,8,balanced,0.021451520919799804
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,4,8,balanced,0.021968638896942137
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,4,8,balanced,0.049838080406188964
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,4,8,balanced,0.05022208213806152
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,4,8,balanced,0.050195198059082034
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,4,8,balanced,0.05051519870758057
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,4,8,balanced,0.050425601005554196
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,4,8,balanced,0.05087103843688965
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,4,8,balanced,0.050933761596679686
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,4,8,balanced,0.051270399093627926
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,4,8,balanced,0.05296256065368652
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,4,8,balanced,0.053387517929077144
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,4,8,balanced,0.02667135953903198
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,4,8,balanced,0.05493247985839844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,4,8,balanced,0.057145600318908696
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,4,8,balanced,0.0613708782196045
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,4,8,balanced,0.02355072021484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,4,8,balanced,0.045203199386596685
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,4,8,balanced,0.049176321029663086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,4,8,balanced,0.04925055980682373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,4,8,power_law_1.01,0.06941311836242677
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,4,8,power_law_1.01,0.08520575523376464
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,4,8,power_law_1.01,0.07862912178039551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,4,8,power_law_1.01,0.08931584358215332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,4,8,power_law_1.01,0.08739456176757812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,4,8,power_law_1.01,0.13753600120544435
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,4,8,power_law_1.01,0.140829439163208
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,4,8,power_law_1.01,0.14622591972351073
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,4,8,power_law_1.01,0.15275775909423828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,4,8,power_law_1.01,0.15426560401916506
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,4,8,power_law_1.01,0.1627814483642578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,4,8,power_law_1.01,0.15963520050048827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,4,8,power_law_1.01,0.15828991889953614
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,4,8,power_law_1.01,0.16582015991210936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,4,8,power_law_1.01,0.16926975250244142
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,4,8,power_law_1.01,0.17640703201293945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,4,8,power_law_1.01,0.19681407928466796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,4,8,power_law_1.01,0.22280576705932614
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,4,8,balanced,0.049498882293701175
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,4,8,power_law_1.01,0.27773824691772464
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,4,8,power_law_1.01,0.3375500869750977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,4,8,power_law_1.01,0.4591052627563476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,4,8,power_law_1.01,0.6325273513793945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,4,8,power_law_1.01,0.8065382385253906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,4,8,power_law_1.01,1.1315814208984374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,4,8,power_law_1.01,1.482266845703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,4,8,balanced,0.081015043258667
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,4,8,balanced,0.08877311706542969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,4,8,power_law_1.01,0.044062719345092774
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,4,8,balanced,0.11978495597839356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,4,8,power_law_1.01,0.06061823844909668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,4,8,power_law_1.01,0.05332863807678223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,4,8,power_law_1.01,0.059448318481445314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,4,8,power_law_1.01,2.2148582458496096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,4,8,power_law_1.01,0.064300799369812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,4,8,power_law_1.01,0.07836031913757324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,4,8,power_law_1.01,0.08262016296386718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,4,8,power_law_1.01,0.08140416145324707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,4,8,power_law_1.01,0.0843289566040039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,4,8,power_law_1.01,0.08856320381164551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,4,8,power_law_1.01,0.09381248474121093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,4,8,power_law_1.01,0.09126655578613281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,4,8,power_law_1.01,0.08958720207214356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,4,8,power_law_1.01,0.09071616172790528
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,4,8,power_law_1.01,0.0906214427947998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,4,8,power_law_1.01,0.10455936431884764
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,4,8,balanced,0.14673919677734376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,4,8,power_law_1.01,0.12613887786865235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,4,8,power_law_1.01,0.14427904129028318
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,4,8,power_law_1.01,0.19797376632690428
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,4,8,power_law_1.01,2.885711364746094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,4,8,power_law_1.01,0.23159679412841797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,4,8,power_law_1.01,0.2990208053588867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,4,8,power_law_1.01,0.3500864028930664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,4,8,balanced,0.18978048324584962
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,4,8,power_law_1.01,0.09373056411743164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,4,8,power_law_1.01,0.5731046295166016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,4,8,power_law_1.01,0.026433279514312746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,4,8,power_law_1.01,0.6381299209594726
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,4,8,power_law_1.01,0.04257919788360596
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,4,8,power_law_1.01,0.031166720390319824
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,4,8,power_law_1.01,0.03951103925704956
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,4,8,power_law_1.01,0.03947776079177857
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,4,8,power_law_1.01,0.04656511783599854
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,4,8,power_law_1.01,0.04621568202972412
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,4,8,power_law_1.01,0.04737664222717285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,4,8,power_law_1.01,1.007557144165039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,4,8,power_law_1.01,0.047736320495605465
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,4,8,power_law_1.01,0.04951807975769043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,4,8,power_law_1.01,0.050058240890502925
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,4,8,power_law_1.01,0.05025919914245606
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,4,8,power_law_1.01,0.05254655838012695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,4,8,power_law_1.01,1.1919129943847655
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,4,8,power_law_1.01,0.051814398765563964
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,4,8,power_law_1.01,0.053614082336425783
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,4,8,power_law_1.01,0.04765183925628662
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,4,8,balanced,0.2519782447814941
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,4,8,power_law_1.01,0.12243455886840819
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,4,8,power_law_1.01,0.10252415657043457
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,4,8,power_law_1.01,0.06325503826141357
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,4,8,power_law_1.01,0.05670271873474121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,4,8,power_law_1.2,0.07100287914276124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,4,8,power_law_1.2,0.07405951976776123
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,4,8,power_law_1.01,0.08666239738464356
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,4,8,power_law_1.01,0.06970111846923828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,4,8,power_law_1.2,0.08088959693908691
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,4,8,power_law_1.2,0.09191424369812011
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,4,8,power_law_1.2,0.09402751922607422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,4,8,power_law_1.01,0.17265024185180664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,4,8,power_law_1.2,0.14144895553588868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,4,8,power_law_1.2,0.14840703964233398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,4,8,power_law_1.2,0.15364224433898926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,4,8,power_law_1.2,0.1566323184967041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,4,8,power_law_1.2,0.16600704193115234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,4,8,power_law_1.2,0.1608755111694336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,4,8,power_law_1.2,0.16277376174926758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,4,8,power_law_1.2,0.16993024826049805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,4,8,power_law_1.2,0.1757030487060547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,4,8,power_law_1.2,0.18222463607788086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,4,8,power_law_1.2,0.20117631912231446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,4,8,power_law_1.2,0.23711360931396483
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,4,8,power_law_1.2,0.14302720069885252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,4,8,power_law_1.2,0.3052467155456543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,4,8,power_law_1.2,0.3964019012451172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,4,8,power_law_1.2,0.48865921020507813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,4,8,power_law_1.2,0.6317030334472656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,4,8,power_law_1.2,0.931188507080078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,4,8,power_law_1.2,1.6815731811523438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,4,8,power_law_1.2,0.04486656188964844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,4,8,power_law_1.2,1.2328089904785156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,4,8,power_law_1.2,0.04967040061950684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,4,8,power_law_1.2,0.05691391944885253
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,4,8,power_law_1.2,0.061053438186645506
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,4,8,power_law_1.2,0.06263040065765381
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,4,8,power_law_1.2,0.07887487888336182
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,4,8,power_law_1.2,0.0798412799835205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,4,8,power_law_1.2,0.08422528266906738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,4,8,power_law_1.2,0.08645248413085938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,4,8,power_law_1.2,0.08787839889526368
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,4,8,power_law_1.2,0.09365119934082031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,4,8,power_law_1.2,0.09309056282043457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,4,8,power_law_1.2,0.08709247589111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,4,8,power_law_1.2,0.09440768241882325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,4,8,power_law_1.2,0.09376383781433105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,4,8,power_law_1.2,0.09759872436523438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,4,8,power_law_1.2,0.11440383911132812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,4,8,power_law_1.2,0.13385727882385254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,4,8,power_law_1.2,0.15521280288696288
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,4,8,power_law_1.01,0.2560204887390137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,4,8,power_law_1.2,0.20353279113769532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,4,8,power_law_1.2,0.24848384857177735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,4,8,power_law_1.2,0.349296646118164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,4,8,power_law_1.2,0.41706497192382813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,4,8,power_law_1.2,0.5821887969970703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,4,8,power_law_1.2,2.3550746154785154
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,4,8,power_law_1.2,0.026071040630340575
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,4,8,power_law_1.2,0.03073535919189453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,4,8,power_law_1.2,1.1325631713867188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,4,8,power_law_1.2,0.0345472002029419
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,4,8,power_law_1.2,0.0370252799987793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,4,8,power_law_1.2,0.04247551918029785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,4,8,power_law_1.2,1.4785382080078127
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,4,8,power_law_1.2,0.04502528190612793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,4,8,power_law_1.2,0.04715007781982422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,4,8,power_law_1.2,0.0464844799041748
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,4,8,power_law_1.2,0.04834432125091553
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,4,8,power_law_1.2,0.04983039855957031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,4,8,power_law_1.2,3.2811700439453126
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,4,8,power_law_1.2,0.05010176181793213
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,4,8,power_law_1.2,0.05108736038208008
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,4,8,power_law_1.2,0.0532748794555664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,4,8,power_law_1.2,0.052309761047363276
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,4,8,power_law_1.2,0.055641598701477044
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,4,8,power_law_1.2,0.04657663822174073
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,4,8,power_law_1.2,0.7539801788330078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,4,8,power_law_1.01,0.27177215576171876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,4,8,power_law_1.01,0.4773337554931641
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,4,8,power_law_1.01,0.6594073486328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,4,8,power_law_1.2,0.10451711654663084
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,4,8,power_law_1.2,0.06961023807525635
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,4,8,power_law_1.2,0.11943679809570314
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,4,8,power_law_1.2,0.07718527793884278
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,4,8,power_law_1.2,0.06255871772766113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,8,4,balanced,0.03273344039916992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,8,4,balanced,0.03339904069900513
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,8,4,balanced,0.060428800582885744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,8,4,balanced,0.07293312072753906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,8,4,balanced,0.09886207580566406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,8,4,balanced,0.1129139232635498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,8,4,balanced,0.15481599807739257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,8,4,balanced,0.15806207656860352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,8,4,balanced,0.16136831283569336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,8,4,balanced,0.16479488372802736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,8,4,balanced,0.15923968315124512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,8,4,balanced,0.16130304336547852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,8,4,balanced,0.16613632202148437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,8,4,balanced,0.17072128295898437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,8,4,balanced,0.17349248886108398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,8,4,balanced,0.18020736694335937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,8,4,balanced,0.1941427230834961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,8,4,balanced,0.2212940788269043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,8,4,balanced,0.25010175704956056
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,8,4,balanced,0.31112447738647464
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,8,4,balanced,0.36428031921386717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,8,4,balanced,0.5550361633300781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,8,4,balanced,0.6761215972900391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,8,4,balanced,0.9293913269042969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,8,4,balanced,1.221313247680664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,8,4,balanced,1.775330505371094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,8,4,balanced,0.03658240079879761
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,8,4,balanced,0.03704960107803344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,8,4,balanced,0.043564801216125486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,8,4,balanced,0.05705984115600586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,8,4,balanced,0.07882751941680907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,8,4,balanced,2.3303321838378905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,8,4,balanced,0.0912499237060547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,8,4,balanced,0.1028287982940674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,8,4,balanced,0.10797823905944823
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,8,4,balanced,0.11001728057861329
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,8,4,balanced,0.11125375747680663
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,8,4,balanced,0.10353407859802247
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,8,4,balanced,0.10715519905090332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,8,4,balanced,0.10973695755004882
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,8,4,balanced,0.11435775756835938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,8,4,balanced,0.11887359619140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,8,4,balanced,0.12752767562866213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,8,4,balanced,0.14551039695739748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,8,4,balanced,0.16004352569580077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,8,4,balanced,0.19260927200317385
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,8,4,balanced,0.11368448257446288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,8,4,balanced,0.2280985641479492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,8,4,balanced,0.33854591369628906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,8,4,balanced,0.4081280136108399
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,4,8,power_law_1.2,0.1670579147338867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,8,4,balanced,0.5831782531738281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,8,4,balanced,0.7608064270019531
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,4,8,power_law_1.2,0.24582271575927733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,8,4,balanced,1.070992660522461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,8,4,balanced,1.3985101318359374
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,4,8,power_law_1.2,0.33756542205810547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,4,8,power_law_1.2,0.4102016067504882
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,4,8,power_law_1.2,0.7274854278564453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,4,8,power_law_1.2,0.6862246704101562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,8,4,balanced,0.026443519592285157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,8,4,balanced,0.06081664085388183
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,8,4,balanced,0.02801919937133789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,8,4,balanced,0.0610214376449585
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,8,4,balanced,0.06129280090332031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,8,4,balanced,0.061564159393310544
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,8,4,balanced,0.06154623985290527
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,8,4,balanced,0.062707200050354
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,8,4,balanced,0.025181438922882084
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,8,4,balanced,0.06319359779357911
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,8,4,balanced,0.06427775859832764
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,8,4,balanced,0.06496895790100098
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,8,4,balanced,0.06836351871490479
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,8,4,balanced,0.07136256217956542
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,8,4,balanced,0.07651840209960938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,8,4,balanced,0.08182144165039062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,8,4,balanced,0.03420799970626831
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,8,4,balanced,0.05548031806945801
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,8,4,balanced,0.060753917694091795
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,8,4,balanced,0.0605951976776123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,8,4,power_law_1.01,0.05673088073730469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,8,4,power_law_1.01,0.07001599788665772
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,8,4,power_law_1.01,0.07024127960205077
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,8,4,balanced,0.06060800075531005
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,8,4,power_law_1.01,0.08230655670166016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,8,4,power_law_1.01,0.09054976463317871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,8,4,power_law_1.01,0.11198975563049315
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,8,4,power_law_1.01,0.1454425621032715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,8,4,power_law_1.01,0.14989312171936037
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,8,4,power_law_1.01,0.15393280029296874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,8,4,power_law_1.01,0.1578700828552246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,8,4,power_law_1.01,0.1582323169708252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,8,4,power_law_1.01,0.15667584419250488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,8,4,power_law_1.01,0.1590873622894287
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,8,4,power_law_1.01,0.16565120697021485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,8,4,power_law_1.01,0.17093120574951173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,8,4,power_law_1.01,0.17746944427490235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,8,4,power_law_1.01,0.19537792205810547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,8,4,power_law_1.01,0.22198144912719728
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,8,4,power_law_1.01,0.2551168060302734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,8,4,power_law_1.01,0.3416691207885742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,8,4,power_law_1.01,0.42369918823242186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,8,4,power_law_1.01,0.5649868774414062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,8,4,power_law_1.01,0.7560038757324219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,8,4,power_law_1.01,1.02887939453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,8,4,power_law_1.01,1.3319821166992187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,8,4,power_law_1.01,1.9490917968750001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,8,4,power_law_1.01,0.03914752006530762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,8,4,power_law_1.01,0.04979584217071533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,8,4,power_law_1.01,0.053344001770019536
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,8,4,power_law_1.01,2.512193298339844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,8,4,power_law_1.01,0.06519296169281005
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,8,4,power_law_1.01,0.06984320163726807
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,8,4,power_law_1.01,0.09622783660888672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,8,4,power_law_1.01,0.10085632324218749
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,8,4,power_law_1.01,0.10210816383361816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,8,4,power_law_1.01,0.1033625602722168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,8,4,power_law_1.01,0.11224448204040527
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,8,4,balanced,0.11529088020324707
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,8,4,balanced,0.12744959831237793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,8,4,power_law_1.01,0.1012390422821045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,8,4,power_law_1.01,0.10282496452331542
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,8,4,power_law_1.01,0.10929792404174805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,8,4,power_law_1.01,0.10806912422180175
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,8,4,power_law_1.01,0.1108505630493164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,8,4,power_law_1.01,0.12284159660339355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,8,4,power_law_1.01,0.14624640464782715
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,8,4,balanced,0.1790630340576172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,8,4,power_law_1.01,0.17006208419799804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,8,4,power_law_1.01,0.22126976013183594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,8,4,power_law_1.01,0.25796735763549805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,8,4,power_law_1.01,0.33713153839111326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,8,4,power_law_1.01,0.44623489379882814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,8,4,power_law_1.01,0.6143616104125977
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,8,4,balanced,0.22632320404052733
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,8,4,power_law_1.01,0.02759167909622192
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,8,4,power_law_1.01,0.7820851135253906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,8,4,power_law_1.01,0.09045120239257812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,8,4,power_law_1.01,0.035092480182647705
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,8,4,power_law_1.01,0.03570175886154175
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,8,4,power_law_1.01,0.040252161026000974
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,8,4,power_law_1.01,0.04198912143707275
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,8,4,power_law_1.01,0.05596672058105469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,8,4,power_law_1.01,0.05893887996673584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,8,4,power_law_1.01,1.1578892517089845
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,8,4,power_law_1.01,0.05883776187896729
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,8,4,power_law_1.01,0.05883008003234863
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,8,4,power_law_1.01,0.060949759483337404
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,8,4,power_law_1.01,0.06076672077178955
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,8,4,power_law_1.01,0.06160128116607666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,8,4,power_law_1.01,0.06368639945983887
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,8,4,power_law_1.01,0.06491392135620118
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,8,4,power_law_1.01,0.06710400104522704
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,8,4,power_law_1.01,0.06888192176818848
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,8,4,power_law_1.01,0.11318528175354003
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,8,4,balanced,0.29788415908813476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,8,4,power_law_1.01,1.472515869140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,8,4,balanced,0.39863296508789064
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,8,4,power_law_1.01,0.1410905647277832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,8,4,power_law_1.01,0.09568256378173828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,8,4,power_law_1.01,0.07552000045776366
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,8,4,power_law_1.01,0.08444031715393066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,8,4,power_law_1.2,0.05650303840637207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,8,4,power_law_1.2,0.06352255821228028
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,8,4,power_law_1.01,0.20386432647705077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,8,4,power_law_1.2,0.07064832210540771
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,8,4,power_law_1.2,0.0791654396057129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,8,4,power_law_1.2,0.08628095626831055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,8,4,power_law_1.2,0.1072550392150879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,8,4,power_law_1.2,0.14566143989562988
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,8,4,power_law_1.2,0.15356032371520995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,8,4,power_law_1.2,0.15734272003173827
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,8,4,power_law_1.01,0.3722995376586914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,8,4,power_law_1.2,0.15842432022094727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,8,4,power_law_1.2,0.1625075149536133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,8,4,power_law_1.2,0.15990400314331055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,8,4,power_law_1.2,0.1696460723876953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,8,4,power_law_1.2,0.16352256774902343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,8,4,power_law_1.2,0.17430400848388672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,8,4,power_law_1.2,0.18167680740356446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,8,4,power_law_1.2,0.19625728607177734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,8,4,power_law_1.01,0.24286592483520505
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,8,4,power_law_1.2,0.22334207534790038
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,8,4,power_law_1.2,0.2614323234558106
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,8,4,power_law_1.2,0.3527897644042969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,8,4,power_law_1.2,0.44695552825927737
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,8,4,power_law_1.2,0.5872524642944337
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,8,4,power_law_1.2,0.780794906616211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,8,4,power_law_1.2,1.4011494445800783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,8,4,power_law_1.2,0.03916287899017334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,8,4,power_law_1.2,0.04616447925567627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,8,4,power_law_1.2,0.05369344234466553
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,8,4,power_law_1.2,1.0632921600341798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,8,4,power_law_1.2,0.06206079959869385
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,8,4,power_law_1.2,0.06888192176818848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,8,4,power_law_1.2,0.09478015899658203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,8,4,power_law_1.2,0.08311936378479004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,8,4,power_law_1.2,0.09914496421813965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,8,4,power_law_1.2,0.10217215538024901
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,8,4,power_law_1.2,0.10789759635925293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,8,4,power_law_1.2,0.10148096084594727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,8,4,power_law_1.2,0.1045580768585205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,8,4,power_law_1.2,0.10945664405822755
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,8,4,power_law_1.2,0.11065855979919434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,8,4,power_law_1.2,0.11398271560668945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,8,4,power_law_1.2,0.13116415977478027
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,8,4,power_law_1.2,0.1483123207092285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,8,4,power_law_1.2,0.10254847526550293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,8,4,power_law_1.2,0.1809561538696289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,8,4,power_law_1.2,0.22956031799316406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,8,4,power_law_1.2,0.26994176864624025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,8,4,power_law_1.2,0.34722431182861324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,8,4,power_law_1.2,0.4412416076660156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,8,4,power_law_1.2,0.7869414520263672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,8,4,power_law_1.2,2.036115264892578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,8,4,power_law_1.2,0.027653119564056396
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,8,4,power_law_1.2,0.03326848030090332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,8,4,power_law_1.2,0.03550848007202149
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,8,4,power_law_1.2,0.0404915189743042
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,8,4,power_law_1.2,0.6382694244384766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,8,4,power_law_1.2,0.0466649580001831
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,8,4,power_law_1.2,0.054248957633972174
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,8,4,power_law_1.01,0.5028646469116211
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,8,4,power_law_1.2,0.05841152191162109
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,8,4,power_law_1.2,0.058005762100219724
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,8,4,power_law_1.2,0.05845119953155518
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,8,4,power_law_1.2,0.06040832042694092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,8,4,power_law_1.2,2.7195559692382814
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,8,4,power_law_1.2,0.06204671859741211
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,8,4,power_law_1.2,0.06303872108459473
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,8,4,power_law_1.2,0.06481535911560059
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,8,4,power_law_1.2,0.06654719829559326
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,8,4,power_law_1.2,0.06940159797668458
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,8,4,power_law_1.2,0.0813862419128418
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,8,4,power_law_1.2,0.0937446403503418
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,8,4,power_law_1.2,0.10770815849304198
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,8,4,power_law_1.2,1.5779315185546874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,8,4,power_law_1.2,1.2329036712646484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,8,4,power_law_1.01,0.8260173034667968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,8,4,power_law_1.01,0.8531916809082031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,8,4,power_law_1.2,0.07345920085906983
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,8,4,power_law_1.2,0.15518976211547852
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,8,4,power_law_1.2,0.199683837890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,16,2,balanced,0.031750400066375736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,16,2,balanced,0.05222655773162842
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,16,2,balanced,0.059599361419677734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,8,4,power_law_1.2,0.2787276840209961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,16,2,balanced,0.07786623954772949
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,16,2,balanced,0.10709247589111329
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,16,2,balanced,0.12149760246276856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,16,2,balanced,0.12240256309509276
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,16,2,balanced,0.12271231651306151
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,16,2,balanced,0.12422143936157226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,16,2,balanced,0.17274368286132813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,16,2,balanced,0.17540096282958983
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,16,2,balanced,0.180947208404541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,16,2,balanced,0.1832012748718262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,16,2,balanced,0.19036031723022462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,16,2,balanced,0.1939481544494629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,16,2,balanced,0.19877119064331056
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,16,2,balanced,0.21196928024291992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,16,2,balanced,0.2366464042663574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,16,2,balanced,0.26308095932006836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,16,2,balanced,0.32743808746337894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,16,2,balanced,0.37861248016357424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,16,2,balanced,0.5690047836303711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,16,2,balanced,0.6868096160888671
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,16,2,balanced,0.9299916839599609
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,8,4,power_law_1.2,0.339628791809082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,16,2,balanced,1.2213299560546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,16,2,balanced,0.03784832000732422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,16,2,balanced,0.03385727882385254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,16,2,balanced,0.04665728092193604
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,16,2,balanced,0.07159552097320557
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,16,2,balanced,0.09987711906433105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,16,2,balanced,0.12043264389038086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,16,2,balanced,1.7640179443359376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,16,2,balanced,0.11459584236145019
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,16,2,balanced,0.11586048126220702
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,16,2,balanced,0.11673343658447266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,16,2,balanced,0.15455743789672852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,16,2,balanced,0.15615872383117674
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,8,4,power_law_1.2,0.5637670516967773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,16,2,balanced,0.15861503601074217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,16,2,balanced,2.3188543701171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,16,2,balanced,0.16070144653320312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,16,2,balanced,0.16478208541870115
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,16,2,balanced,0.1639295959472656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,16,2,balanced,0.16646272659301758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,16,2,balanced,0.1747417640686035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,16,2,balanced,0.1934553527832031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,16,2,balanced,0.20686336517333986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,16,2,balanced,0.23624191284179688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,16,2,balanced,0.26960895538330076
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,16,2,balanced,0.40598400115966793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,16,2,balanced,0.4671628952026367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,16,2,balanced,0.6653247833251953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,16,2,balanced,0.8648780822753906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,16,2,balanced,1.185623016357422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,16,2,balanced,1.5753074645996095
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,8,4,power_law_1.2,0.8198963165283203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,8,4,power_law_1.2,1.256533737182617
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,8,4,power_law_1.2,1.4848562622070314
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,16,2,balanced,0.03455487966537475
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,16,2,balanced,0.04075263977050782
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,16,2,balanced,0.053647360801696777
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,16,2,balanced,0.08831744194030762
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,16,2,balanced,0.09042176246643066
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,16,2,balanced,0.0906214427947998
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,16,2,balanced,0.0906048011779785
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,16,2,balanced,0.09373056411743164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,16,2,balanced,0.09487744331359863
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,16,2,balanced,0.09504768371582031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,16,2,balanced,0.03617408037185669
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,16,2,balanced,0.09737088203430175
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,16,2,balanced,0.10006015777587891
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,16,2,balanced,0.10814720153808595
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,16,2,balanced,0.10986495971679688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,16,2,balanced,0.11714431762695313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,16,2,balanced,0.12754816055297852
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,16,2,balanced,0.07765376091003418
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,16,2,balanced,0.08973183631896972
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,16,2,balanced,0.08782719612121583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,16,2,power_law_1.01,0.0356275200843811
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,16,2,power_law_1.01,0.05877247810363769
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,16,2,power_law_1.01,0.06605696201324462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,16,2,power_law_1.01,0.07826176166534424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,16,2,power_law_1.01,0.09209600448608399
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,16,2,power_law_1.01,0.11483008384704589
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,16,2,power_law_1.01,0.10904319763183594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,16,2,power_law_1.01,0.11033984184265137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,16,2,power_law_1.01,0.11385855674743653
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,16,2,power_law_1.01,0.16753023147583007
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,16,2,power_law_1.01,0.17190528869628907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,16,2,power_law_1.01,0.17592063903808594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,16,2,power_law_1.01,0.1805299186706543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,16,2,power_law_1.01,0.18782976150512695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,16,2,power_law_1.01,0.1914944076538086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,16,2,power_law_1.01,0.19797887802124023
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,16,2,balanced,0.08988800048828124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,16,2,power_law_1.01,0.21475584030151368
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,16,2,power_law_1.01,0.24240640640258787
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,16,2,power_law_1.01,0.27143680572509765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,16,2,power_law_1.01,0.3331123352050781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,16,2,power_law_1.01,0.4080678558349609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,16,2,power_law_1.01,0.5761356735229493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,16,2,power_law_1.01,0.7032371520996094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,16,2,power_law_1.01,1.0043596649169921
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,16,2,power_law_1.01,1.2628096008300782
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,16,2,balanced,0.38771968841552734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,16,2,balanced,0.19511423110961915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,16,2,power_law_1.01,1.8144895935058591
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,16,2,balanced,0.30102527618408204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,16,2,power_law_1.01,0.03700992107391358
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,16,2,power_law_1.01,0.04283775806427002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,16,2,power_law_1.01,0.05340672016143799
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,16,2,power_law_1.01,0.08161791801452636
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,16,2,power_law_1.01,0.07181951999664307
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,16,2,balanced,0.5186329650878906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,16,2,power_law_1.01,0.10143744468688964
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,16,2,balanced,0.21704832077026368
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,16,2,power_law_1.01,0.10618751525878907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,16,2,power_law_1.01,0.10790016174316405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,16,2,power_law_1.01,0.11046400070190429
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,16,2,power_law_1.01,0.12376576423645021
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,16,2,power_law_1.01,0.13675135612487793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,16,2,power_law_1.01,0.1386342430114746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,16,2,power_law_1.01,0.15201919555664062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,16,2,power_law_1.01,0.14715392112731934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,16,2,power_law_1.01,0.15074560165405274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,16,2,power_law_1.01,2.3777804565429688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,16,2,power_law_1.01,0.14965760231018066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,16,2,power_law_1.01,0.164005126953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,16,2,power_law_1.01,0.1956070327758789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,16,2,power_law_1.01,0.21330047607421876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,16,2,power_law_1.01,0.25881343841552734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,16,2,power_law_1.01,0.30420095443725587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,16,2,power_law_1.01,0.40431743621826166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,16,2,power_law_1.01,0.49490303039550787
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,16,2,power_law_1.01,0.037438719272613524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,16,2,power_law_1.01,0.6797491455078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,16,2,power_law_1.01,0.04207104206085205
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,16,2,power_law_1.01,0.045966081619262696
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,16,2,power_law_1.01,0.05224575996398926
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,16,2,power_law_1.01,0.06003200054168702
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,16,2,power_law_1.01,0.07826560020446778
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,16,2,power_law_1.01,0.08236800193786621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,16,2,power_law_1.01,1.2570777893066407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,16,2,power_law_1.01,0.08544639587402345
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,16,2,power_law_1.01,0.08742655754089355
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,16,2,power_law_1.01,0.09389568328857421
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,16,2,power_law_1.01,0.08908032417297364
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,16,2,power_law_1.01,0.09451775550842285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,16,2,power_law_1.01,1.6654284667968748
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,16,2,power_law_1.01,0.09831808090209962
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,16,2,power_law_1.01,0.10135680198669435
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,16,2,power_law_1.01,0.10869248390197754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,16,2,power_law_1.01,0.8826150512695312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,16,2,power_law_1.01,0.1589260768890381
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,16,2,power_law_1.01,0.08324992179870605
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,16,2,balanced,0.6937945556640626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,16,2,power_law_1.01,0.13905407905578612
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,16,2,power_law_1.01,0.11739263534545899
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,16,2,power_law_1.01,0.25717119216918943
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,16,2,power_law_1.01,0.3544460678100586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,16,2,power_law_1.01,0.1988684844970703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,16,2,power_law_1.2,0.057786879539489744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,16,2,power_law_1.2,0.06695168018341065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,16,2,power_law_1.2,0.07485439777374267
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,16,2,power_law_1.2,0.08514176368713379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,16,2,power_law_1.2,0.10521599769592285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,16,2,power_law_1.2,0.10773887634277343
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,16,2,power_law_1.01,0.7165888214111329
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,16,2,power_law_1.2,0.11657471656799316
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,16,2,power_law_1.01,0.45813888549804693
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,16,2,power_law_1.2,0.1712883186340332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,16,2,power_law_1.2,0.17259008407592774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,16,2,power_law_1.2,0.17832319259643553
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,16,2,power_law_1.2,0.18247552871704104
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,16,2,power_law_1.2,0.03373440027236939
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,16,2,power_law_1.2,0.1891302490234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,16,2,power_law_1.2,0.19320831298828126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,16,2,power_law_1.2,0.20107391357421872
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,16,2,power_law_1.2,0.2157683181762695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,16,2,power_law_1.2,0.2432793617248535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,16,2,power_law_1.2,0.2736832046508789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,16,2,power_law_1.2,0.340682258605957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,16,2,power_law_1.2,0.41694591522216795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,16,2,power_law_1.2,0.11295359611511231
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,16,2,power_law_1.2,0.5766182327270508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,16,2,power_law_1.2,0.7071628570556641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,16,2,power_law_1.2,0.040133118629455566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,16,2,power_law_1.2,1.0048729705810548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,16,2,power_law_1.2,0.04353536128997802
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,16,2,power_law_1.2,0.05548927783966064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,16,2,power_law_1.2,1.2779315185546873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,16,2,power_law_1.2,0.06817920207977295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,16,2,power_law_1.2,0.07998464107513428
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,16,2,power_law_1.2,0.09978495597839356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,16,2,power_law_1.2,0.10388863563537598
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,16,2,power_law_1.2,0.10709888458251952
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,16,2,power_law_1.2,0.11039999961853027
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,16,2,power_law_1.2,0.12921855926513673
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,16,2,power_law_1.2,0.1374233627319336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,16,2,power_law_1.01,0.924505615234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,16,2,power_law_1.2,0.15026559829711914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,16,2,power_law_1.2,0.14929408073425293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,16,2,power_law_1.2,0.14947456359863281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,16,2,power_law_1.2,0.16077823638916017
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,16,2,power_law_1.2,0.15820927619934083
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,16,2,power_law_1.2,0.18176256179809572
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,16,2,power_law_1.2,0.1999283218383789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,16,2,power_law_1.2,0.22151424407958986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,16,2,power_law_1.2,0.268984317779541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,16,2,power_law_1.2,0.31627391815185546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,16,2,power_law_1.2,0.41176193237304687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,16,2,power_law_1.2,0.5010470581054688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,16,2,power_law_1.2,0.8970726776123048
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,16,2,power_law_1.2,0.037067520618438723
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,16,2,power_law_1.2,0.04147071838378906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,16,2,power_law_1.2,0.04533504009246826
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,16,2,power_law_1.2,0.6889958190917969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,16,2,power_law_1.2,0.05162752151489258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,16,2,power_law_1.2,0.05559552192687989
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,16,2,power_law_1.2,1.842487030029297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,16,2,power_law_1.2,0.07840640068054198
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,16,2,power_law_1.2,0.08054656028747559
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,16,2,power_law_1.2,0.08527104377746582
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,16,2,power_law_1.2,0.08477567672729493
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,16,2,power_law_1.2,0.08700544357299804
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,16,2,power_law_1.2,0.09076607704162598
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,16,2,power_law_1.2,0.0940339183807373
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,16,2,power_law_1.2,0.09604736328124999
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,16,2,power_law_1.2,0.10324992179870604
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,16,2,power_law_1.2,0.11256447792053223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,16,2,power_law_1.2,2.399982147216797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,16,2,power_law_1.2,1.690855712890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,16,2,power_law_1.2,0.2369740867614746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,16,2,power_law_1.2,1.2818623352050782
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,16,2,power_law_1.01,1.6627801513671876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,16,2,power_law_1.2,0.12067584037780761
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,16,2,power_law_1.2,0.13790719985961913
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,16,2,power_law_1.2,0.15171711921691894
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,16,2,power_law_1.2,0.3251571273803711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,16,2,power_law_1.2,0.19188095092773438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,32,1,balanced,0.04843200206756591
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,32,1,balanced,0.05488383769989014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,32,1,balanced,0.07733248233795166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,32,1,balanced,0.09873536109924316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,32,1,balanced,0.1361900806427002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,32,1,balanced,0.1581510353088379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,32,1,balanced,0.1588326358795166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,32,1,balanced,0.16229696273803712
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,32,1,balanced,0.16401023864746095
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,16,2,power_law_1.2,0.4358284759521484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,32,1,balanced,0.16585855484008788
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,32,1,balanced,0.1682918357849121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,32,1,balanced,0.17316799163818358
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,32,1,balanced,0.24358848571777342
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,32,1,balanced,0.25133951187133785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,32,1,balanced,0.2532550430297852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,32,1,balanced,0.26049856185913084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,32,1,balanced,0.27252735137939454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,32,1,balanced,0.2994291114807129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,32,1,balanced,0.32507392883300784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,32,1,balanced,0.4150886535644531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,32,1,balanced,0.4669510269165039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,32,1,balanced,0.6596364593505859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,32,1,balanced,0.8083987426757812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,32,1,balanced,1.0991321563720704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,32,1,balanced,1.4182470703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,32,1,balanced,2.0310815429687503
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,32,1,balanced,0.049689598083496093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,32,1,balanced,2.6377529907226562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,32,1,balanced,0.06617087841033935
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,32,1,balanced,0.05329792022705078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,32,1,balanced,0.09547264099121093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,32,1,balanced,0.1487724781036377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,32,1,balanced,0.1979520034790039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,32,1,balanced,0.19327167510986326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,32,1,balanced,0.19363519668579102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,32,1,balanced,0.1958937644958496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,32,1,balanced,0.19664127349853516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,32,1,balanced,0.20059072494506838
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,32,1,balanced,0.20284160614013672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,32,1,balanced,0.23072639465332032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,32,1,balanced,0.2343276786804199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,32,1,balanced,0.23464576721191407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,32,1,balanced,0.23734016418457032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,32,1,balanced,0.24540544509887696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,32,1,balanced,0.2593324851989746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,32,1,balanced,0.2754886436462402
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,32,1,balanced,0.30855871200561524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,32,1,balanced,0.34051841735839844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,32,1,balanced,0.52801025390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,32,1,balanced,0.5894099044799804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,32,1,balanced,0.8385497283935546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,32,1,balanced,1.0866675567626953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,32,1,balanced,1.4776736450195311
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,32,1,power_law_1.01,0.043875842094421386
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,16,2,power_law_1.2,0.9261363220214843
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,16,2,power_law_1.2,1.1583897399902345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,32,1,power_law_1.01,0.05792895793914795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,32,1,balanced,1.9783474731445314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,32,1,power_law_1.01,0.07818175792694092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,32,1,power_law_1.01,0.11424575805664064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,32,1,power_law_1.01,0.0912076759338379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,32,1,power_law_1.01,0.13816831588745118
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,32,1,power_law_1.01,0.14869440078735352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,32,1,power_law_1.01,0.151015043258667
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,32,1,power_law_1.01,0.1568876838684082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,32,1,power_law_1.01,0.16825920104980469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,32,1,power_law_1.01,0.17877567291259766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,32,1,power_law_1.01,0.24048383712768553
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,32,1,power_law_1.01,0.2497651290893555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,32,1,power_law_1.01,0.25472703933715823
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,32,1,power_law_1.01,0.1435264015197754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,32,1,power_law_1.01,0.25978431701660154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,32,1,power_law_1.01,0.2769555282592774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,32,1,power_law_1.01,0.3072467231750488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,32,1,power_law_1.01,0.33737663269042967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,32,1,power_law_1.01,0.41456127166748047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,32,1,power_law_1.01,0.5009427261352539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,32,1,power_law_1.01,0.6635673522949219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,32,1,power_law_1.01,0.8117574310302734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,32,1,power_law_1.01,1.1120947265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,32,1,power_law_1.01,0.049730558395385746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,32,1,power_law_1.01,1.4212037658691405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,32,1,power_law_1.01,0.05294079780578613
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,32,1,power_law_1.01,0.07349120140075684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,32,1,power_law_1.01,0.08912063598632812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,32,1,power_law_1.01,0.1214419174194336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,32,1,power_law_1.01,2.034460754394531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,32,1,power_law_1.01,0.15623616218566894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,32,1,power_law_1.01,0.16080575942993164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,32,1,power_law_1.01,0.16486591339111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,32,1,power_law_1.01,0.1678175926208496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,32,1,power_law_1.01,0.17376064300537108
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,32,1,power_law_1.01,2.6569726562500002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,32,1,power_law_1.01,0.17833984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,32,1,power_law_1.01,0.18705408096313475
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,16,2,power_law_1.2,2.4971469116210936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,32,1,power_law_1.01,0.21013568878173827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,32,1,power_law_1.01,0.2258412742614746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,32,1,power_law_1.01,0.22609535217285157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,32,1,power_law_1.01,0.2457049560546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,32,1,power_law_1.01,0.23213888168334962
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,32,1,power_law_1.01,0.269683837890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,32,1,power_law_1.01,0.29162431716918946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,32,1,power_law_1.01,0.3412140655517578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,32,1,power_law_1.01,0.39884864807128906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,32,1,power_law_1.01,0.5203923034667969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,32,1,power_law_1.01,0.6211590576171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,32,1,power_law_1.2,0.04314432144165039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,32,1,power_law_1.01,0.8508608245849609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,32,1,power_law_1.2,0.0675116777420044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,32,1,power_law_1.01,1.0709324645996094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,32,1,power_law_1.2,0.07844480037689208
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,32,1,power_law_1.2,0.08738176345825195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,32,1,power_law_1.2,0.10713983535766602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,32,1,power_law_1.2,0.14307519912719727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,32,1,power_law_1.2,0.14178175926208497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,32,1,power_law_1.01,1.5009138488769531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,32,1,power_law_1.2,0.1498624038696289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,32,1,power_law_1.2,0.15499008178710938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,32,1,power_law_1.2,0.1602681541442871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,32,1,power_law_1.2,0.17043071746826172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,32,1,power_law_1.2,0.18441343307495117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,32,1,power_law_1.2,0.2429305648803711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,32,1,power_law_1.01,1.9541836547851563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,32,1,power_law_1.2,0.25016063690185547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,32,1,power_law_1.2,0.2535840034484863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,32,1,power_law_1.2,0.2628985595703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,32,1,power_law_1.2,0.2777337646484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,32,1,power_law_1.2,0.3109939193725586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,32,1,power_law_1.2,0.3396448135375977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,32,1,power_law_1.2,0.42073089599609376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,32,1,power_law_1.2,0.5030732727050781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,32,1,power_law_1.2,0.6651564788818359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,32,1,power_law_1.2,0.8183340454101563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,32,1,power_law_1.2,0.0462611198425293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,32,1,power_law_1.2,1.1185727691650391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,32,1,power_law_1.2,0.052937598228454584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,32,1,power_law_1.2,0.06931583881378174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,32,1,power_law_1.2,0.112675199508667
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,32,1,power_law_1.2,1.428656005859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,32,1,power_law_1.2,0.0852678394317627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,32,1,power_law_1.2,0.16808319091796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,32,1,power_law_1.2,0.15504639625549316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,32,1,power_law_1.2,0.15957695960998536
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,32,1,power_law_1.2,0.17492799758911132
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,32,1,power_law_1.2,0.1896396827697754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,32,1,power_law_1.2,0.17207231521606445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,32,1,power_law_1.2,0.1936524772644043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,32,1,power_law_1.2,0.22623424530029296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,32,1,power_law_1.2,0.23371776580810547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,32,1,power_law_1.2,0.24060672760009766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,32,1,power_law_1.2,0.25386943817138674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,32,1,power_law_1.2,0.27312383651733396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,32,1,power_law_1.2,0.2308736038208008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,32,1,power_law_1.2,0.2978937530517578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,32,1,power_law_1.2,0.35011199951171873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,32,1,power_law_1.2,2.040123596191406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,32,1,power_law_1.2,0.40302974700927735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,32,1,power_law_1.2,0.5281107330322266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,32,1,power_law_1.2,0.6274854278564452
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,32,1,power_law_1.2,0.8571532440185546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,32,balanced,0.07146880149841309
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,32,balanced,0.07130623817443847
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,32,balanced,0.07155839920043945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,32,balanced,0.07888256072998047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,32,balanced,0.09527680397033692
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,32,1,power_law_1.2,1.0742700958251954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,32,balanced,0.14535552024841308
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,32,balanced,0.21439104080200194
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,32,balanced,0.2400614356994629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,32,balanced,0.2397311973571777
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,32,balanced,0.2400511932373047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,32,balanced,0.2440358352661133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,32,balanced,0.24948863983154296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,32,balanced,0.24972288131713868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,32,balanced,0.2558412742614746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,32,balanced,0.2608563232421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,32,balanced,0.2633087921142578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,32,balanced,0.2663385581970215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,32,balanced,0.2781491279602051
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,32,balanced,0.3027315139770508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,32,balanced,0.3068992042541504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,32,balanced,0.33815937042236327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,32,balanced,0.38828033447265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,32,balanced,0.5364748764038085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,32,balanced,0.6044172668457032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,32,balanced,0.8005747222900391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,32,balanced,1.0338047790527345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,32,balanced,0.038979840278625486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,32,1,power_law_1.2,1.5075257873535155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,32,balanced,0.039051520824432376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,32,balanced,1.4375526428222656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,32,balanced,0.039161601066589356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,32,balanced,0.04494080066680908
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,32,balanced,0.061246719360351566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,32,balanced,0.0868620777130127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,32,balanced,0.1278873634338379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,32,balanced,0.1375372791290283
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,32,balanced,0.13870592117309571
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,32,balanced,0.14117759704589844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,32,balanced,0.1444979190826416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,32,balanced,0.1437875175476074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,32,balanced,0.1483955192565918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,32,balanced,0.15080960273742677
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,32,balanced,0.1522982406616211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,32,1,power_law_1.2,2.6565191650390623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,32,balanced,0.14548224449157715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,32,balanced,0.15082880020141604
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,32,balanced,0.13782912254333496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,32,balanced,0.15346303939819336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,32,balanced,0.16580608367919922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,32,balanced,0.18442752838134766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,32,balanced,0.21362943649291993
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,32,balanced,0.23419647216796874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,32,balanced,0.28510719299316406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,32,balanced,0.3724774551391602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,32,balanced,0.46453887939453126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,32,balanced,0.6114457702636719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,32,1,power_law_1.2,1.9610450744628907
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,1,32,balanced,0.02229759931564331
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,1,32,balanced,0.02228224039077759
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,1,32,balanced,0.05410816192626953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,1,32,balanced,0.023848960399627687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,1,32,balanced,0.07140992164611817
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,1,32,balanced,0.0714739179611206
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,1,32,balanced,0.07197184085845947
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,1,32,balanced,0.022717440128326417
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,1,32,balanced,0.07196928024291992
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,1,32,balanced,0.07350527763366699
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,1,32,balanced,0.07279871940612793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,1,32,balanced,0.07220608234405518
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,1,32,balanced,0.07361663818359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,1,32,balanced,0.0732313585281372
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,1,32,balanced,0.0735155200958252
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,1,32,balanced,0.07345920085906983
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,1,32,balanced,0.07358719825744628
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,1,32,balanced,0.07551231861114502
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,1,32,balanced,0.07435776233673096
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,1,32,balanced,0.0782809591293335
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,1,32,balanced,0.08013567924499512
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,1,32,balanced,0.08400511741638184
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,32,power_law_1.01,0.14361087799072264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,32,power_law_1.01,0.20272512435913087
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,1,32,balanced,0.07623167991638183
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,32,power_law_1.01,0.20821760177612303
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,32,power_law_1.01,0.2141312026977539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,32,power_law_1.01,0.21507455825805666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,32,power_law_1.01,0.2268864059448242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,32,power_law_1.01,0.22781183242797853
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,32,power_law_1.01,0.22804735183715819
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,1,32,balanced,0.0342195200920105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,32,power_law_1.01,0.2381260871887207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,32,power_law_1.01,0.23454336166381834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,32,power_law_1.01,0.23557247161865233
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,32,power_law_1.01,0.2322431945800781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,32,power_law_1.01,0.24035200119018554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,32,power_law_1.01,0.2468070411682129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,32,power_law_1.01,0.24695552825927733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,32,power_law_1.01,0.25453311920166016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,32,power_law_1.01,0.26761215209960937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,32,power_law_1.01,0.3168793678283691
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,32,power_law_1.01,0.37792129516601564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,32,power_law_1.01,0.4252044677734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,32,power_law_1.01,0.47712001800537107
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,32,power_law_1.01,0.6478668975830078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,32,power_law_1.01,0.768647689819336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,32,power_law_1.01,1.0351078033447265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,32,power_law_1.01,1.4300953674316408
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,32,power_law_1.01,0.08575360298156738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,32,power_law_1.01,0.1260915184020996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,32,power_law_1.01,2.0543807983398437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,32,power_law_1.01,0.1265779209136963
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,32,power_law_1.01,0.12752384185791016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,32,power_law_1.01,0.12553600311279295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,32,power_law_1.01,0.1070297622680664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,32,power_law_1.01,0.11274239540100098
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,32,power_law_1.01,0.11465215682983398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,32,power_law_1.01,2.9968408203125003
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,32,power_law_1.01,0.12765567779541015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,32,power_law_1.01,0.1302847957611084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,32,power_law_1.01,0.13319040298461915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,32,power_law_1.01,0.134455041885376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,32,power_law_1.01,0.1405183982849121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,32,power_law_1.01,0.1462886428833008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,32,power_law_1.01,0.14505984306335448
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,32,power_law_1.01,0.14136704444885254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,32,power_law_1.01,0.154398717880249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,32,power_law_1.01,0.1399385643005371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,32,power_law_1.01,0.1784998321533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,32,power_law_1.01,0.20350080490112304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,32,power_law_1.01,0.22484479904174806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,32,power_law_1.01,0.30641151428222657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,32,power_law_1.01,0.36958206176757813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,32,power_law_1.01,0.5135590362548828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,32,power_law_1.01,0.6119180679321289
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,1,32,power_law_1.01,0.05498623847961426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,32,power_law_1.01,0.8861631774902344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,1,32,power_law_1.01,0.07154431819915771
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,1,32,power_law_1.01,0.07121151924133301
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,1,32,power_law_1.01,0.07179520130157471
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,32,power_law_1.01,1.245681915283203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,1,32,power_law_1.01,0.07146624088287354
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,1,32,power_law_1.01,0.05859327793121338
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,1,32,power_law_1.01,0.06245632171630859
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,1,32,power_law_1.01,0.06345727920532226
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,1,32,power_law_1.01,0.06728960037231445
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,1,32,power_law_1.01,0.06910848140716552
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,1,32,power_law_1.01,0.06773248195648193
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,1,32,power_law_1.01,0.06950655937194825
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,1,32,power_law_1.01,0.06908671855926514
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,1,32,power_law_1.01,0.06998015880584717
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,1,32,power_law_1.01,0.07156735897064209
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,1,32,power_law_1.01,0.07443583965301513
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,1,32,power_law_1.01,0.07419519901275635
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,1,32,power_law_1.01,0.06827392101287841
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,1,32,balanced,0.11339648246765137
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,1,32,balanced,0.12131839752197267
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,1,32,balanced,0.15582847595214844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,1,32,power_law_1.01,0.07605760097503662
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,1,32,power_law_1.01,0.0951296043395996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,1,32,power_law_1.01,0.08710271835327148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,32,power_law_1.2,0.14468735694885254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,32,power_law_1.2,0.20316287994384769
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,32,power_law_1.2,0.20665855407714845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,32,power_law_1.2,0.2120614433288574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,32,power_law_1.2,0.21246976852416993
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,32,power_law_1.2,0.19939327239990234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,32,power_law_1.2,0.1943756866455078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,32,power_law_1.2,0.21680767059326173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,32,power_law_1.2,0.22613759994506838
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,32,power_law_1.2,0.22043264389038084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,32,power_law_1.2,0.2293657684326172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,32,power_law_1.2,0.2358937644958496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,32,power_law_1.2,0.244267520904541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,32,power_law_1.2,0.24293760299682615
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,32,power_law_1.2,0.2595673561096191
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,32,power_law_1.2,0.27184127807617187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,32,power_law_1.2,0.33187198638916016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,32,power_law_1.2,0.408031997680664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,32,power_law_1.2,0.4690214538574219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,32,power_law_1.2,0.5793587112426758
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,1,32,power_law_1.01,0.11877632141113281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,32,power_law_1.2,0.8195571136474609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,32,power_law_1.2,1.0932236480712891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,32,power_law_1.2,1.4414425659179688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,1,32,power_law_1.01,0.12363903999328614
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,32,power_law_1.2,1.9423207092285157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,32,power_law_1.2,0.08463487625122071
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,32,power_law_1.2,0.12562944412231444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,32,power_law_1.2,0.12629247665405274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,32,power_law_1.2,2.8876724243164062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,32,power_law_1.2,0.12715647697448731
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,32,power_law_1.2,0.2493836784362793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,32,power_law_1.2,0.12862208366394043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,32,power_law_1.2,0.1091801643371582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,32,power_law_1.2,0.12555647850036622
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,32,power_law_1.2,0.1289395236968994
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,32,power_law_1.2,0.13255552291870115
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,32,power_law_1.2,0.13605631828308104
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,32,power_law_1.2,0.11949952125549317
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,32,power_law_1.2,4.071037292480469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,1,32,power_law_1.01,0.17910272598266602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,32,power_law_1.2,0.14302207946777343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,32,power_law_1.2,0.13826175689697268
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,32,power_law_1.2,0.14224512100219727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,32,power_law_1.2,0.16087167739868163
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,32,power_law_1.2,0.1982579231262207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,32,power_law_1.2,0.2296268844604492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,32,power_law_1.2,0.27347200393676757
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,32,power_law_1.2,0.36579711914062496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,32,power_law_1.2,0.1425766372680664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,32,power_law_1.2,0.39768062591552733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,32,power_law_1.2,0.599662094116211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,32,power_law_1.2,0.7815449523925782
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,1,32,power_law_1.2,0.05479551792144775
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,32,power_law_1.2,1.1375475311279297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,1,32,power_law_1.2,0.07104256153106689
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,1,32,power_law_1.2,0.0712281608581543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,1,32,power_law_1.2,0.07133952140808106
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,1,32,power_law_1.2,0.07137919902801514
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,32,power_law_1.2,1.5675021362304686
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,1,32,power_law_1.2,0.058512639999389646
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,32,power_law_1.2,0.13827456474304198
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,1,32,power_law_1.2,0.06476160049438476
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,1,32,power_law_1.2,0.06597760200500488
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,1,32,power_law_1.2,0.06782847881317139
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,1,32,power_law_1.2,0.06784255981445313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,1,32,power_law_1.2,0.06811647891998292
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,1,32,power_law_1.2,0.05447296142578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,1,32,power_law_1.2,0.06927616119384765
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,1,32,power_law_1.2,0.07048831939697266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,1,32,power_law_1.2,0.0731059217453003
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,1,32,power_law_1.2,0.0706611204147339
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,1,32,power_law_1.2,0.07530879974365234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,1,32,power_law_1.2,0.08286975860595704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,32,power_law_1.2,0.10292863845825195
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,1,32,power_law_1.01,0.22629375457763673
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,1,32,power_law_1.01,0.31487232208251953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,1,32,power_law_1.01,0.3787763214111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,1,32,power_law_1.2,0.06701183795928956
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,1,32,power_law_1.2,0.09645567893981934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,16,balanced,0.03865983963012695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,16,balanced,0.03905407905578613
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,16,balanced,0.061900801658630364
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,16,balanced,0.0751692819595337
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,16,balanced,0.09905920028686524
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,1,32,power_law_1.2,0.16510080337524413
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,16,balanced,0.14462847709655763
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,1,32,power_law_1.2,0.10961152076721192
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,16,balanced,0.23524991989135743
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,16,balanced,0.23704959869384767
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,16,balanced,0.2372991943359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,16,balanced,0.244323844909668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,16,balanced,0.24728960037231446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,16,balanced,0.2529702377319336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,16,balanced,0.24505983352661134
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,16,balanced,0.25342464447021484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,16,balanced,0.25550336837768556
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,16,balanced,0.26306432723999024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,16,balanced,0.2820902442932129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,16,balanced,0.3009600067138672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,16,balanced,0.33635070800781247
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,16,balanced,0.3845747375488281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,16,balanced,0.4364646530151367
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,1,32,power_law_1.2,0.13585408210754396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,16,balanced,0.6190758514404296
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,1,32,power_law_1.2,0.24250751495361328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,16,balanced,0.7514367675781249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,16,balanced,0.03658495903015137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,16,balanced,0.03626879930496216
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,16,balanced,0.9956147003173829
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,1,32,power_law_1.2,0.2834598350524902
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,16,balanced,0.03827455997467041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,16,balanced,0.05060991764068603
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,16,balanced,1.3401394653320313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,16,balanced,0.06655871868133545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,16,balanced,0.09288831710815429
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,16,balanced,0.14560895919799804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,16,balanced,0.14680319786071777
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,1,32,power_law_1.2,0.46653694152832037
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,16,balanced,0.147457275390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,16,balanced,0.1526655960083008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,16,balanced,0.1558579158782959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,16,balanced,0.14945664405822753
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,16,balanced,0.15769856452941894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,16,balanced,0.1516723155975342
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,16,balanced,0.15357312202453613
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,16,balanced,1.8248524475097656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,16,balanced,0.155413761138916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,16,balanced,0.16705535888671874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,16,balanced,0.18406400680541993
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,16,balanced,0.20740991592407226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,16,balanced,0.22714624404907227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,16,balanced,0.2670028877258301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,16,balanced,0.3116902351379395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,16,balanced,0.39608959197998045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,16,balanced,0.1487168025970459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,16,balanced,0.5224319839477539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,16,balanced,0.6738969421386718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,16,balanced,0.8986521911621093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,16,balanced,0.20714879989624024
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,1,32,power_law_1.2,0.5724940872192382
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,2,16,balanced,0.0227891206741333
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,2,16,balanced,0.0753331184387207
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,2,16,balanced,0.07589632034301758
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,2,16,balanced,0.022785279750823974
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,2,16,balanced,0.07608704090118408
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,2,16,balanced,0.023187201023101806
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,2,16,balanced,0.07697408199310303
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,2,16,balanced,0.036239359378814694
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,2,16,balanced,0.07703296184539796
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,2,16,balanced,0.07690112113952637
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,2,16,balanced,0.07734399795532226
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,2,16,balanced,0.07832575798034667
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,2,16,balanced,0.0791654396057129
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,2,16,balanced,0.05626368045806884
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,2,16,balanced,0.08042112350463868
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,2,16,balanced,0.08270079612731933
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,2,16,balanced,0.08644479751586914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,2,16,balanced,0.08985600471496583
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,2,16,balanced,0.09631232261657716
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,2,16,balanced,0.0769868803024292
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,2,16,balanced,0.07740543842315674
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,2,16,balanced,0.026879999637603756
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,16,power_law_1.01,0.13836671829223632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,16,power_law_1.01,0.1949452781677246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,16,power_law_1.01,0.09637632369995117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,16,power_law_1.01,0.188799991607666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,16,power_law_1.01,0.2118515205383301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,16,power_law_1.01,0.22960895538330078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,16,power_law_1.01,0.2105241584777832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,16,power_law_1.01,0.21600511550903317
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,16,power_law_1.01,0.22099199295043945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,16,power_law_1.01,0.2221708869934082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,16,power_law_1.01,0.2273843193054199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,16,power_law_1.01,0.23238399505615231
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,16,power_law_1.01,0.23317760467529297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,16,power_law_1.01,0.2258380889892578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,16,power_law_1.01,0.2353228759765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,16,power_law_1.01,0.24746879577636718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,16,power_law_1.01,0.24997375488281248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,16,power_law_1.01,0.27679359436035156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,16,power_law_1.01,0.32492671966552733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,16,power_law_1.01,0.4300083160400391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,16,power_law_1.01,0.4844927978515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,16,power_law_1.01,0.6198348617553711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,2,16,balanced,0.07531775951385497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,16,power_law_1.01,0.782840347290039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,16,power_law_1.01,1.0799565124511719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,16,power_law_1.01,0.0644927978515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,16,power_law_1.01,1.3205862426757813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,16,power_law_1.01,0.0892249584197998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,16,power_law_1.01,0.1307161617279053
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,16,power_law_1.01,0.12517503738403318
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,16,power_law_1.01,0.13961088180541992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,16,power_law_1.01,2.2203993225097656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,16,power_law_1.01,0.13304960250854492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,16,power_law_1.01,2.360112609863281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,16,power_law_1.01,0.10913536071777344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,16,power_law_1.01,0.13387392044067384
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,16,power_law_1.01,0.13293567657470703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,16,power_law_1.01,0.13844608306884765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,16,power_law_1.01,0.14961024284362795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,16,power_law_1.01,0.14380800247192382
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,16,power_law_1.01,0.1438643169403076
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,16,power_law_1.01,0.13933823585510255
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,16,power_law_1.01,0.1494054412841797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,16,power_law_1.01,0.14616576194763184
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,16,power_law_1.01,0.1674470329284668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,16,power_law_1.01,0.14380672454833984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,16,power_law_1.01,0.18818944931030274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,16,power_law_1.01,0.21564800262451173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,16,power_law_1.01,0.24811519622802733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,16,power_law_1.01,0.31885055541992186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,16,power_law_1.01,0.3977395248413086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,16,power_law_1.01,0.5443648147583008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,16,power_law_1.01,0.6352537536621095
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,2,16,power_law_1.01,0.03749759912490845
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,2,16,power_law_1.01,0.05697535991668701
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,16,power_law_1.01,0.917834243774414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,2,16,power_law_1.01,0.0752281618118286
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,2,16,power_law_1.01,0.07409791946411133
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,2,16,power_law_1.01,0.07531136035919189
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,16,power_law_1.01,1.231604461669922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,2,16,power_law_1.01,0.07039999961853027
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,2,16,power_law_1.01,0.0623795223236084
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,2,16,power_law_1.01,0.06584959983825683
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,2,16,power_law_1.01,0.06817791938781739
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,2,16,power_law_1.01,0.068504319190979
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,2,16,power_law_1.01,0.07154943943023681
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,2,16,power_law_1.01,0.07099391937255858
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,2,16,power_law_1.01,0.07441023826599122
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,2,16,power_law_1.01,0.0721779203414917
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,2,16,power_law_1.01,0.07196671962738037
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,2,16,power_law_1.01,0.07229440212249756
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,2,16,power_law_1.01,0.07800960063934326
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,2,16,power_law_1.01,0.08094464302062988
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,2,16,balanced,0.07518720149993896
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,2,16,balanced,0.13153535842895508
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,2,16,balanced,0.1455564785003662
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,2,16,balanced,0.18986751556396483
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,2,16,power_law_1.01,0.09245311737060546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,2,16,power_law_1.01,0.08325504302978516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,2,16,power_law_1.01,0.1020083236694336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,16,power_law_1.2,0.09801471710205079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,16,power_law_1.2,0.13910783767700194
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,16,power_law_1.2,0.18214271545410157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,16,power_law_1.2,0.18710399627685548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,16,power_law_1.2,0.2065382385253906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,16,power_law_1.2,0.23142784118652343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,16,power_law_1.2,0.19256959915161134
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,16,power_law_1.2,0.20721920013427733
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,2,16,power_law_1.01,0.12033920288085938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,16,power_law_1.2,0.21352832794189452
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,16,power_law_1.2,0.21574399948120115
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,16,power_law_1.2,0.2120742416381836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,16,power_law_1.2,0.23545984268188475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,16,power_law_1.2,0.23247615814208986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,16,power_law_1.2,0.21925888061523438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,16,power_law_1.2,0.2317132759094238
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,16,power_law_1.2,0.241712646484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,16,power_law_1.2,0.2609638404846192
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,16,power_law_1.2,0.2990924835205078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,16,power_law_1.2,0.37645313262939456
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,16,power_law_1.2,0.4461888122558594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,16,power_law_1.2,0.5222911834716797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,16,power_law_1.2,0.7106124877929687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,16,power_law_1.2,0.8250265502929688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,16,power_law_1.2,1.4013427734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,16,power_law_1.2,1.639764404296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,2,16,power_law_1.01,0.15049599647521972
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,16,power_law_1.2,0.06464640140533448
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,16,power_law_1.2,0.11723008155822753
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,16,power_law_1.2,0.12558208465576173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,16,power_law_1.2,0.13141759872436523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,16,power_law_1.2,2.3051135253906248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,16,power_law_1.2,0.11592320442199706
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,16,power_law_1.2,0.10314240455627441
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,16,power_law_1.2,0.11419008255004884
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,16,power_law_1.2,0.11961983680725097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,16,power_law_1.2,0.13009663581848147
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,16,power_law_1.2,0.12773504257202148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,16,power_law_1.2,3.1540005493164065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,16,power_law_1.2,0.13730048179626464
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,16,power_law_1.2,0.14542336463928224
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,16,power_law_1.2,0.13004672050476074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,16,power_law_1.2,0.14210176467895508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,16,power_law_1.2,0.1486975955963135
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,16,power_law_1.2,0.1489267158508301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,16,power_law_1.2,0.1777676773071289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,16,power_law_1.2,0.1938047981262207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,16,power_law_1.2,0.22735231399536132
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,16,power_law_1.2,0.2599155235290527
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,16,power_law_1.2,0.35794174194335937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,16,power_law_1.2,0.43387390136718745
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,16,power_law_1.2,0.08869376182556152
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,16,power_law_1.2,0.6084249496459961
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,2,16,power_law_1.2,0.03845887899398804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,16,power_law_1.2,0.7450611114501953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,2,16,power_law_1.2,0.056696319580078126
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,2,16,power_law_1.2,0.06924416065216064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,16,power_law_1.2,1.1018879699707032
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,2,16,power_law_1.2,0.07407487869262695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,2,16,power_law_1.2,0.0746611213684082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,2,16,power_law_1.2,0.06385280132293701
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,16,power_law_1.2,1.398254089355469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,2,16,power_law_1.2,0.06615295886993408
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,2,16,power_law_1.2,0.061584639549255374
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,2,16,power_law_1.2,0.06629504203796387
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,2,16,power_law_1.2,0.07044735908508301
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,2,16,power_law_1.2,0.0720140790939331
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,2,16,power_law_1.2,0.07381504058837891
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,2,16,power_law_1.2,0.06864895820617675
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,2,16,power_law_1.2,0.07435647964477539
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,2,16,power_law_1.2,0.07560832023620605
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,2,16,power_law_1.2,0.07555327892303468
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,2,16,power_law_1.2,0.09058431625366212
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,2,16,power_law_1.01,0.19053823471069337
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,2,16,power_law_1.01,0.23480703353881838
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,2,16,power_law_1.2,0.06668928146362305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,2,16,power_law_1.01,0.3612275314331055
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,2,16,power_law_1.01,0.44484096527099604
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,2,16,power_law_1.2,0.11896703720092774
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,2,16,power_law_1.2,0.08380672454833984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,2,16,power_law_1.2,0.19964544296264647
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,4,8,balanced,0.0378278398513794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,8,balanced,0.03836159944534302
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,8,balanced,0.06381440162658691
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,8,balanced,0.07790847778320312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,8,balanced,0.09880703926086426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,8,balanced,0.14664319992065428
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,8,balanced,0.2098547172546387
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,2,16,power_law_1.2,0.15547264099121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,8,balanced,0.23488639831542968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,8,balanced,0.23848192214965822
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,2,16,power_law_1.2,0.10372480392456054
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,8,balanced,0.24243328094482425
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,8,balanced,0.23704063415527346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,8,balanced,0.24348543167114256
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,2,16,power_law_1.2,0.2536140823364258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,8,balanced,0.24817407608032224
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,8,balanced,0.2573580741882324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,8,balanced,0.2624460792541504
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,2,16,power_law_1.2,0.37025665283203124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,8,balanced,0.27217023849487304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,8,balanced,0.31804288864135744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,8,balanced,0.28539520263671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,8,balanced,0.34908161163330076
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,8,balanced,0.41372032165527345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,8,balanced,0.4762636947631836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,8,balanced,0.6087257766723633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,8,balanced,0.03621887922286987
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,8,balanced,0.8217830657958984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,8,balanced,0.03913216114044189
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,8,balanced,0.04610432147979736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,8,balanced,1.0767065429687501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,8,balanced,0.0606169605255127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,8,balanced,0.0774515199661255
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,8,balanced,1.4102578735351563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,8,balanced,0.1509171199798584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,8,balanced,0.1680281639099121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,8,balanced,0.1701913642883301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,8,balanced,0.17250816345214842
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,8,balanced,0.16539775848388671
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,8,balanced,0.16800512313842772
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,8,balanced,0.17194623947143556
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,8,balanced,0.175281925201416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,8,balanced,1.9986329650878907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,8,balanced,0.17838848114013672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,8,balanced,0.18485504150390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,8,balanced,0.196942081451416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,8,balanced,0.10967424392700195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,8,balanced,0.21519615173339846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,8,balanced,0.2374553680419922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,8,balanced,0.27527040481567383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,8,balanced,0.31016447067260744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,8,balanced,2.7871002197265624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,8,balanced,0.38543231964111324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,8,balanced,0.470513916015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,8,balanced,0.612250862121582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,8,balanced,0.8143551635742188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,8,balanced,1.1148492431640624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,8,balanced,1.4712089538574218
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,2,16,power_law_1.2,0.4933081436157226
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,2,16,power_law_1.2,0.6506521606445312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,4,8,balanced,0.028126718997955324
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,4,8,balanced,0.026900479793548582
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,4,8,balanced,0.0864038372039795
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,4,8,balanced,0.08656384468078612
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,4,8,balanced,0.08731391906738281
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,4,8,balanced,0.08751744270324707
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,4,8,balanced,0.026419200897216798
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,4,8,balanced,0.087457275390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,4,8,balanced,0.03339263916015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,4,8,balanced,0.08758912086486817
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,4,8,balanced,0.08892800331115722
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,4,8,balanced,0.08988544464111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,4,8,balanced,0.09223423957824707
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,4,8,balanced,0.06620416164398193
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,4,8,balanced,0.0932915210723877
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,4,8,balanced,0.09700480461120606
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,4,8,balanced,0.09886719703674315
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,4,8,balanced,0.04495488166809082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,4,8,balanced,0.1053990364074707
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,4,8,balanced,0.11086848258972168
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,4,8,balanced,0.12361344337463379
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,4,8,balanced,0.08704128265380859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,8,power_law_1.01,0.09596672058105468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,8,power_law_1.01,0.13802623748779297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,8,power_law_1.01,0.13377920150756836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,8,power_law_1.01,0.15342464447021484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,4,8,power_law_1.01,0.07600512027740478
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,8,power_law_1.01,0.18396415710449218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,8,power_law_1.01,0.1601535987854004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,8,power_law_1.01,0.2147212791442871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,8,power_law_1.01,0.21336320877075193
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,8,power_law_1.01,0.22242303848266604
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,8,power_law_1.01,0.2234752082824707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,8,power_law_1.01,0.21795328140258788
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,8,power_law_1.01,0.21025663375854492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,8,power_law_1.01,0.22672000885009766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,8,power_law_1.01,0.2314240074157715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,8,power_law_1.01,0.23631744384765624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,8,power_law_1.01,0.24650495529174804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,8,power_law_1.01,0.27911039352416994
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,8,power_law_1.01,0.3196697616577148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,8,power_law_1.01,0.4057011032104493
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,4,8,balanced,0.08581376075744629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,8,power_law_1.01,0.5031577682495116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,8,power_law_1.01,0.6522035217285156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,8,power_law_1.01,0.75093505859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,8,power_law_1.01,1.011383056640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,8,power_law_1.01,0.04585343837738037
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,8,power_law_1.01,1.3793215942382813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,8,power_law_1.01,0.06914303779602052
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,8,power_law_1.01,0.09781248092651366
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,8,power_law_1.01,0.09638655662536622
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,8,power_law_1.01,1.9991807556152346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,8,power_law_1.01,0.11254528045654297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,8,power_law_1.01,0.12544384002685546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,8,power_law_1.01,0.11731712341308594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,8,power_law_1.01,0.1397721576690674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,8,power_law_1.01,0.1326809597015381
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,8,power_law_1.01,0.13850751876831055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,8,power_law_1.01,2.4969984436035153
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,8,power_law_1.01,0.13328384399414062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,8,power_law_1.01,0.13617152214050293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,8,power_law_1.01,0.13550975799560547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,8,power_law_1.01,0.16150783538818358
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,8,power_law_1.01,0.15801983833312988
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,8,power_law_1.01,0.1590169620513916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,8,power_law_1.01,0.17633024215698243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,8,power_law_1.01,0.19982336044311524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,8,power_law_1.01,0.20934656143188474
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,8,power_law_1.01,0.24699647903442382
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,8,power_law_1.01,0.28451200485229494
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,8,power_law_1.01,0.33414657592773434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,8,power_law_1.01,0.4268236923217773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,8,power_law_1.01,0.5918835067749024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,8,power_law_1.01,0.7470207977294921
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,4,8,power_law_1.01,0.03136127948760986
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,4,8,power_law_1.01,0.046302719116210936
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,4,8,power_law_1.01,0.06698624134063721
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,4,8,power_law_1.01,0.06296319961547851
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,4,8,power_law_1.01,0.0708902406692505
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,8,power_law_1.01,1.3568141174316406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,4,8,power_law_1.01,0.07800191879272461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,4,8,power_law_1.01,0.08210687637329102
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,4,8,power_law_1.01,0.08291711807250976
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,4,8,power_law_1.01,0.0784332799911499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,8,power_law_1.01,1.0107622528076172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,4,8,power_law_1.01,0.07556352138519287
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,4,8,power_law_1.01,0.07640831947326661
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,4,8,power_law_1.01,0.0726643180847168
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,4,8,power_law_1.01,0.07913983821868896
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,4,8,power_law_1.01,0.0832089614868164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,4,8,power_law_1.01,0.08500224113464355
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,4,8,power_law_1.01,0.08544768333435059
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,4,8,power_law_1.01,0.08569600105285644
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,4,8,power_law_1.01,0.0939417552947998
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,4,8,balanced,0.0870143985748291
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,4,8,balanced,0.26037376403808593
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,4,8,balanced,0.19908735275268555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,4,8,balanced,0.17169023513793946
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,4,8,power_law_1.01,0.11250304222106933
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,4,8,power_law_1.01,0.12417663574218749
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,4,8,power_law_1.2,0.07631360054016113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,8,power_law_1.2,0.09579520225524903
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,8,power_law_1.2,0.11475711822509767
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,4,8,power_law_1.01,0.1514431953430176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,8,power_law_1.2,0.13015551567077638
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,4,8,power_law_1.01,0.10047103881835937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,8,power_law_1.2,0.16298879623413084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,8,power_law_1.2,0.1687884712219238
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,8,power_law_1.2,0.1869491195678711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,8,power_law_1.2,0.20255615234375002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,8,power_law_1.2,0.21520639419555665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,8,power_law_1.2,0.21687295913696286
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,8,power_law_1.2,0.2177356719970703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,8,power_law_1.2,0.2102783966064453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,8,power_law_1.2,0.22409088134765626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,8,power_law_1.2,0.22915199279785153
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,8,power_law_1.2,0.212936954498291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,8,power_law_1.2,0.2342540740966797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,8,power_law_1.2,0.2546995162963867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,8,power_law_1.2,0.299685115814209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,8,power_law_1.2,0.3376268768310547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,8,power_law_1.2,0.47669376373291017
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,8,power_law_1.2,0.5290854263305664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,8,power_law_1.2,0.6802239990234374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,8,power_law_1.2,0.8246182250976564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,8,power_law_1.2,1.1680384063720703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,8,power_law_1.2,1.4731814575195312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,8,power_law_1.2,0.04933631896972656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,8,power_law_1.2,0.06771200180053712
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,8,power_law_1.2,2.2535423278808593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,8,power_law_1.2,0.09472384452819824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,8,power_law_1.2,0.1175705623626709
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,8,power_law_1.2,2.8347943115234377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,8,power_law_1.2,0.08170880317687988
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,4,8,power_law_1.01,0.18465024948120118
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,8,power_law_1.2,0.12404224395751953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,8,power_law_1.2,0.13463680267333983
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,8,power_law_1.2,0.13523200035095215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,8,power_law_1.2,0.14163071632385255
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,8,power_law_1.2,0.13596799850463867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,8,power_law_1.2,0.14160767555236817
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,8,power_law_1.2,0.14017919540405271
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,8,power_law_1.2,0.15165184020996095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,8,power_law_1.2,0.15703935623168946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,8,power_law_1.2,0.1568819236755371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,8,power_law_1.2,0.17611007690429686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,8,power_law_1.2,0.19361536026000975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,8,power_law_1.2,0.21025535583496097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,8,power_law_1.2,0.25258239746093747
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,8,power_law_1.2,0.28612224578857426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,8,power_law_1.2,0.13149056434631348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,8,power_law_1.2,0.3729280090332031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,8,power_law_1.2,0.4754982376098633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,8,power_law_1.2,0.6081792068481445
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,4,8,power_law_1.2,0.031152639389038082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,4,8,power_law_1.2,0.04552192211151123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,8,power_law_1.2,1.1661286163330078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,4,8,power_law_1.2,0.05955584049224853
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,8,power_law_1.2,1.3468199157714844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,4,8,power_law_1.2,0.06058879852294922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,4,8,power_law_1.2,0.0695462417602539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,8,power_law_1.2,0.7869785308837891
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,4,8,power_law_1.2,0.07035903930664063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,4,8,power_law_1.2,0.0787340784072876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,4,8,power_law_1.2,0.08449024200439453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,4,8,power_law_1.2,0.07476736068725585
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,4,8,power_law_1.2,0.07138559818267823
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,4,8,power_law_1.2,0.07961855888366699
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,4,8,power_law_1.2,0.07905663967132567
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,4,8,power_law_1.2,0.08348159790039063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,4,8,power_law_1.2,0.08676480293273926
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,4,8,power_law_1.2,0.08840319633483887
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,4,8,power_law_1.2,0.09058943748474121
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,4,8,power_law_1.2,0.11269248008728028
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,4,8,power_law_1.2,0.07622144222259522
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,4,8,power_law_1.01,0.2893888092041016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,4,8,power_law_1.01,0.32612865447998046
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,4,8,power_law_1.01,0.44933120727539066
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,4,8,power_law_1.01,0.6182976150512696
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,4,8,power_law_1.2,0.10083200454711913
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,4,8,power_law_1.2,0.12918784141540526
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,4,8,power_law_1.2,0.15425663948059082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,8,4,balanced,0.03267967939376831
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,8,4,balanced,0.03663232088088989
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,4,8,power_law_1.2,0.20762752532958983
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,8,4,balanced,0.06221439838409424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,8,4,balanced,0.08016511917114258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,4,8,power_law_1.2,0.2741939163208008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,8,4,balanced,0.10172160148620606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,8,4,balanced,0.15744383811950685
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,4,8,power_law_1.2,0.40344577789306635
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,8,4,balanced,0.21044864654541015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,8,4,balanced,0.21402751922607424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,8,4,balanced,0.216943359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,8,4,balanced,0.21573247909545898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,8,4,balanced,0.2536409568786621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,8,4,balanced,0.2581222343444824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,8,4,balanced,0.26402687072753905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,8,4,balanced,0.26633983612060547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,8,4,balanced,0.271201286315918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,8,4,balanced,0.28219520568847656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,8,4,balanced,0.2953126335144043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,8,4,balanced,0.32515327453613285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,8,4,balanced,0.3539161682128906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,8,4,balanced,0.4117849731445313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,8,4,balanced,0.4741171264648438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,8,4,balanced,0.596545295715332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,8,4,balanced,0.815308837890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,8,4,balanced,0.03515007972717285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,8,4,balanced,1.0601817321777343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,8,4,balanced,0.037914879322052
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,8,4,balanced,0.046197757720947266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,8,4,balanced,1.3847103881835938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,8,4,balanced,0.06679679870605468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,8,4,balanced,0.0960166358947754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,8,4,balanced,0.1316979217529297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,8,4,balanced,0.17959552764892578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,8,4,balanced,0.1814988708496094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,8,4,balanced,0.18257152557373046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,8,4,balanced,0.18449920654296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,8,4,balanced,1.9559129333496095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,8,4,balanced,0.20770687103271485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,8,4,balanced,0.21152000427246093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,8,4,balanced,0.21389951705932617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,8,4,balanced,0.21922048568725586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,8,4,balanced,0.2254105567932129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,8,4,balanced,0.23358720779418946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,8,4,balanced,0.2526963233947754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,8,4,balanced,0.2703398323059082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,8,4,balanced,0.3089049530029297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,8,4,balanced,2.7128448486328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,8,4,balanced,0.3443084716796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,4,8,power_law_1.2,0.5119193649291992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,8,4,balanced,0.41326080322265624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,8,4,balanced,0.48751361846923824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,8,4,balanced,0.21999616622924806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,8,4,balanced,0.6310259246826172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,8,4,balanced,0.8614911651611328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,8,4,balanced,1.1469964599609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,8,4,balanced,1.5172479248046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,4,8,power_law_1.2,0.7266585540771484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,4,8,power_law_1.2,1.1138726043701173
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,8,4,balanced,0.03746815919876099
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,8,4,balanced,0.11410047531127929
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,8,4,balanced,0.03591808080673218
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,8,4,balanced,0.114520320892334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,8,4,balanced,0.0395417594909668
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,8,4,balanced,0.11484543800354004
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,8,4,balanced,0.11491328239440919
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,8,4,balanced,0.11454463958740235
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,8,4,balanced,0.11626496315002441
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,8,4,balanced,0.11695232391357421
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,8,4,balanced,0.11681535720825195
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,8,4,balanced,0.12057600021362305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,8,4,balanced,0.12163455963134766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,8,4,balanced,0.12584447860717773
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,8,4,balanced,0.13061887741088868
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,8,4,balanced,0.1362880039215088
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,8,4,balanced,0.08793472290039063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,8,4,balanced,0.1461184024810791
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,8,4,balanced,0.1581158447265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,8,4,balanced,0.18103424072265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,8,4,balanced,0.04630015850067139
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,8,4,power_law_1.01,0.05842175960540772
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,8,4,balanced,0.0615065622329712
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,8,4,power_law_1.01,0.07513984203338622
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,8,4,power_law_1.01,0.0967244815826416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,8,4,power_law_1.01,0.09741056442260743
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,8,4,power_law_1.01,0.14880767822265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,8,4,power_law_1.01,0.13978495597839355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,8,4,power_law_1.01,0.17551231384277344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,8,4,power_law_1.01,0.17862911224365235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,8,4,power_law_1.01,0.1802572822570801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,8,4,power_law_1.01,0.2275187110900879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,8,4,power_law_1.01,0.12121855735778808
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,8,4,power_law_1.01,0.2338175964355469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,8,4,power_law_1.01,0.23713663101196286
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,8,4,power_law_1.01,0.24628223419189452
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,8,4,power_law_1.01,0.2504767990112305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,8,4,power_law_1.01,0.25931007385253907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,8,4,power_law_1.01,0.27639423370361327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,8,4,power_law_1.01,0.3056755256652832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,8,4,balanced,0.11346688270568847
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,8,4,power_law_1.01,0.34099967956542965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,8,4,power_law_1.01,0.41435520172119145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,8,4,power_law_1.01,0.5296409606933594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,8,4,power_law_1.01,0.6577420806884765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,8,4,power_law_1.01,0.8228659057617188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,8,4,power_law_1.01,1.146882553100586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,8,4,power_law_1.01,1.4606425476074218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,8,4,power_law_1.01,0.039934720993041996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,8,4,power_law_1.01,0.0493017578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,8,4,power_law_1.01,2.004267578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,8,4,power_law_1.01,0.07665023803710938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,8,4,power_law_1.01,0.08487296104431152
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,8,4,power_law_1.01,0.11107328414916992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,8,4,power_law_1.01,0.12340864181518554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,8,4,power_law_1.01,0.11989888191223144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,8,4,power_law_1.01,0.14596735954284668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,8,4,power_law_1.01,0.14899200439453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,8,4,power_law_1.01,0.15063551902770994
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,8,4,power_law_1.01,2.6395828247070314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,8,4,power_law_1.01,0.16971008300781248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,8,4,power_law_1.01,0.1736934471130371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,8,4,power_law_1.01,0.19305728912353515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,8,4,power_law_1.01,0.1905574417114258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,8,4,power_law_1.01,0.20117887496948242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,8,4,power_law_1.01,0.22091903686523437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,8,4,power_law_1.01,0.22957183837890627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,8,4,power_law_1.01,0.25417984008789063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,8,4,power_law_1.01,0.2943603134155274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,8,4,power_law_1.01,0.1684172821044922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,8,4,power_law_1.01,0.33407871246337895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,8,4,power_law_1.01,0.4195212936401367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,8,4,power_law_1.01,0.49313793182373045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,8,4,power_law_1.01,0.6816563415527344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,8,4,power_law_1.01,0.03992831945419312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,8,4,power_law_1.01,0.8448563385009766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,8,4,power_law_1.01,0.046487040519714355
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,8,4,power_law_1.01,0.06324992179870606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,8,4,power_law_1.01,1.200680923461914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,8,4,power_law_1.01,0.06200319766998291
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,8,4,power_law_1.01,0.07689216136932372
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,8,4,power_law_1.01,0.08071552276611328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,8,4,power_law_1.01,0.08451711654663085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,8,4,power_law_1.01,1.535020751953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,8,4,power_law_1.01,0.0967244815826416
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,8,4,power_law_1.01,0.09566080093383789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,8,4,power_law_1.01,0.09518464088439942
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,8,4,power_law_1.01,0.09924736022949218
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,8,4,power_law_1.01,0.10342016220092773
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,8,4,power_law_1.01,0.1056499195098877
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,8,4,power_law_1.01,0.10543999671936036
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,8,4,power_law_1.01,0.10735615730285644
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,8,4,power_law_1.01,0.10870271682739259
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,8,4,power_law_1.01,0.11947263717651366
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,8,4,power_law_1.01,0.13154560089111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,8,4,balanced,0.11428352355957032
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,8,4,balanced,0.261342716217041
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,8,4,balanced,0.41020542144775385
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,8,4,balanced,0.31087488174438477
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,8,4,power_law_1.01,0.14342656135559081
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,8,4,power_law_1.01,0.16852991104125975
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,8,4,power_law_1.01,0.1973619270324707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,8,4,power_law_1.2,0.059422721862792974
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,8,4,power_law_1.2,0.07454976081848144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,8,4,power_law_1.2,0.08898816108703614
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,8,4,power_law_1.2,0.09746432304382324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,8,4,power_law_1.2,0.12139136314392089
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,8,4,power_law_1.2,0.13645183563232421
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,8,4,power_law_1.2,0.14540672302246094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,8,4,power_law_1.01,0.2538803291320801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,8,4,power_law_1.2,0.15727744102478028
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,8,4,power_law_1.2,0.1705638313293457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,8,4,power_law_1.2,0.18284799575805663
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,8,4,power_law_1.2,0.22772607803344727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,8,4,power_law_1.2,0.23339775085449216
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,8,4,power_law_1.2,0.2391001510620117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,8,4,power_law_1.2,0.24618495941162108
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,8,4,power_law_1.2,0.25112319946289063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,8,4,power_law_1.2,0.26307584762573244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,8,4,power_law_1.2,0.28152832031249997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,8,4,power_law_1.2,0.31083648681640624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,8,4,power_law_1.2,0.35474815368652346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,8,4,power_law_1.2,0.44841342926025385
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,8,4,power_law_1.2,0.5616332626342773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,8,4,power_law_1.2,0.6843942260742187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,8,4,power_law_1.2,0.813960952758789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,8,4,power_law_1.2,1.1590796661376952
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,8,4,power_law_1.2,1.4730189514160155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,8,4,power_law_1.2,0.04361728191375732
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,8,4,power_law_1.01,0.28673023223876953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,8,4,power_law_1.2,0.049754881858825685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,8,4,power_law_1.2,2.0186329650878907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,8,4,power_law_1.2,0.06703104019165039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,8,4,power_law_1.2,0.08184320449829101
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,8,4,power_law_1.2,0.10901760101318358
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,8,4,power_law_1.2,0.11935104370117186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,8,4,power_law_1.2,0.12218111991882323
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,8,4,power_law_1.2,0.13852031707763673
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,8,4,power_law_1.2,0.1489907169342041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,8,4,power_law_1.2,0.14526847839355467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,8,4,power_law_1.2,2.785309448242187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,8,4,power_law_1.2,0.1711667251586914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,8,4,power_law_1.2,0.17631616592407226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,8,4,power_law_1.2,0.184532470703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,8,4,power_law_1.2,0.18344831466674805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,8,4,power_law_1.2,0.18310911178588868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,8,4,power_law_1.2,0.202295036315918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,8,4,power_law_1.2,0.21976064682006835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,8,4,power_law_1.2,0.23957887649536133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,8,4,power_law_1.2,0.2626982307434082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,8,4,power_law_1.2,0.30352767944335934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,8,4,power_law_1.2,0.33845249176025394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,8,4,power_law_1.2,0.4243251037597656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,8,4,power_law_1.2,0.522929916381836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,8,4,power_law_1.2,0.6916915130615234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,8,4,power_law_1.2,0.8560076904296874
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,8,4,power_law_1.2,0.039891200065612795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,8,4,power_law_1.2,1.2572045135498047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,8,4,power_law_1.2,0.04673791885375976
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,8,4,power_law_1.2,0.056139521598815925
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,8,4,power_law_1.2,0.0706112003326416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,8,4,power_law_1.2,1.599764404296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,8,4,power_law_1.2,0.07901696205139161
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,8,4,power_law_1.2,0.08277631759643554
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,8,4,power_law_1.2,0.09116031646728515
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,8,4,power_law_1.2,0.09642496109008789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,8,4,power_law_1.2,0.09707008361816406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,8,4,power_law_1.2,0.09835136413574219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,8,4,power_law_1.2,0.1032038402557373
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,8,4,power_law_1.2,0.06007936000823975
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,8,4,power_law_1.2,0.09808768272399902
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,8,4,power_law_1.2,0.10291199684143067
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,8,4,power_law_1.01,0.4667033767700195
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,8,4,power_law_1.2,0.10841216087341307
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,8,4,power_law_1.2,0.11098624229431153
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,8,4,power_law_1.2,0.11958271980285644
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,8,4,power_law_1.2,0.16248319625854493
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,8,4,power_law_1.01,0.545516815185547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,8,4,power_law_1.01,0.804221420288086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,8,4,power_law_1.01,1.0475981140136719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,8,4,power_law_1.2,0.20356224060058592
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,8,4,power_law_1.2,0.34205184936523436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,16,2,balanced,0.03296895980834961
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,8,4,power_law_1.2,0.24582143783569338
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,16,2,balanced,0.03624320030212402
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,16,2,balanced,0.06398848056793213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,16,2,balanced,0.08085760116577148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,16,2,balanced,0.10844799995422363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,16,2,balanced,0.16809984207153322
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,16,2,balanced,0.22687231063842775
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,8,4,power_law_1.2,0.13706624031066894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,16,2,balanced,0.22566144943237304
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,8,4,power_law_1.2,0.6929036712646484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,16,2,balanced,0.22646272659301755
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,16,2,balanced,0.22883840560913088
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,16,2,balanced,0.2378291130065918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,16,2,balanced,0.24611583709716797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,16,2,balanced,0.2411520004272461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,8,4,power_law_1.2,0.4085196685791016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,16,2,balanced,0.3293094253540039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,16,2,balanced,0.3324019241333008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,16,2,balanced,0.3409126281738281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,16,2,balanced,0.35419521331787107
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,16,2,balanced,0.3851340866088867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,16,2,balanced,0.4113715362548828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,8,4,power_law_1.2,0.8517068481445312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,16,2,balanced,0.4666815948486328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,16,2,balanced,0.5257126235961914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,16,2,balanced,0.6440460968017578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,16,2,balanced,0.03546112060546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,16,2,balanced,0.8709977722167969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,16,2,balanced,0.03998080015182495
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,16,2,balanced,0.04931327819824219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,16,2,balanced,1.1088038635253905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,16,2,balanced,0.07469696044921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,16,2,balanced,0.11835136413574218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,16,2,balanced,0.1853670310974121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,16,2,balanced,1.4318553161621093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,16,2,balanced,0.2471692848205566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,16,2,balanced,0.24962175369262693
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,16,2,balanced,0.2512665557861328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,16,2,balanced,0.25264896392822267
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,16,2,balanced,0.2552511978149414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,16,2,balanced,2.014799346923828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,16,2,balanced,0.26014335632324215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,16,2,balanced,0.2635289573669434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,16,2,balanced,0.2986534309387207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,16,2,balanced,0.29834239959716796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,16,2,balanced,0.3043481636047363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,16,2,balanced,0.3127987289428711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,16,2,balanced,0.3279296112060547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,16,2,balanced,0.3452851104736328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,16,2,balanced,0.37916801452636717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,16,2,balanced,0.41413505554199215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,16,2,balanced,2.77842041015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,16,2,balanced,0.48564094543457037
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,16,2,balanced,0.5516352081298829
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,16,2,balanced,0.6870566558837891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,16,2,balanced,0.9739020538330078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,16,2,balanced,1.2448806762695312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,16,2,balanced,1.673374786376953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,8,4,power_law_1.2,1.2739699554443358
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,8,4,power_law_1.2,1.7808908081054688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,16,2,balanced,0.0567846393585205
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,16,2,balanced,0.17306367874145506
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,16,2,balanced,0.17258623123168945
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,16,2,balanced,0.06274816036224365
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,16,2,balanced,0.17415552139282225
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,16,2,balanced,0.07343488216400147
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,16,2,balanced,0.17415807723999024
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,16,2,balanced,0.17744640350341795
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,16,2,balanced,0.0583897590637207
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,16,2,balanced,0.17772415161132812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,16,2,balanced,0.17916032791137695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,16,2,balanced,0.18247552871704104
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,16,2,balanced,0.18846208572387696
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,16,2,balanced,0.1948556709289551
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,16,2,balanced,0.09206015586853027
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,16,2,balanced,0.20466688156127927
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,16,2,balanced,0.2162892723083496
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,16,2,balanced,0.23958911895751953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,16,2,balanced,0.2641164779663086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,16,2,balanced,0.3130278396606445
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,16,2,balanced,0.17543296813964843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,16,2,power_law_1.01,0.04191999912261963
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,16,2,power_law_1.01,0.05798272132873535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,16,2,power_law_1.01,0.07836671829223632
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,16,2,balanced,0.13182080268859864
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,16,2,power_law_1.01,0.08890239715576172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,16,2,power_law_1.01,0.11189375877380373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,16,2,power_law_1.01,0.1509823989868164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,16,2,power_law_1.01,0.14581503868103027
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,16,2,power_law_1.01,0.17922048568725585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,16,2,power_law_1.01,0.17990400314331054
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,16,2,power_law_1.01,0.19003263473510743
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,16,2,power_law_1.01,0.1932032012939453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,16,2,power_law_1.01,0.19830400466918946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,16,2,power_law_1.01,0.20970239639282226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,16,2,power_law_1.01,0.310327033996582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,16,2,power_law_1.01,0.31703296661376956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,16,2,power_law_1.01,0.3227711868286133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,16,2,power_law_1.01,0.33564033508300783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,16,2,power_law_1.01,0.3758118438720703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,16,2,power_law_1.01,0.4031872177124024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,16,2,power_law_1.01,0.4752908706665039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,16,2,power_law_1.01,0.5611328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,16,2,power_law_1.01,0.7369062042236327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,16,2,power_law_1.01,0.8734822082519532
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,16,2,balanced,0.1716979217529297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,16,2,power_law_1.01,1.1907635498046876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,16,2,power_law_1.01,0.039287040233612056
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,16,2,power_law_1.01,1.4966937255859374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,16,2,power_law_1.01,0.04606336116790771
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,16,2,power_law_1.01,0.06338047981262207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,16,2,power_law_1.01,0.08291071891784668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,16,2,power_law_1.01,0.1269209575653076
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,16,2,power_law_1.01,0.16009471893310548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,16,2,power_law_1.01,2.081470794677734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,16,2,power_law_1.01,0.1673075294494629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,16,2,power_law_1.01,0.19332351684570312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,16,2,power_law_1.01,0.2055129623413086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,16,2,power_law_1.01,0.20741376876831055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,16,2,power_law_1.01,0.2249420738220215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,16,2,power_law_1.01,0.23024511337280273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,16,2,power_law_1.01,2.6829388427734377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,16,2,power_law_1.01,0.22452608108520508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,16,2,power_law_1.01,0.2622003173828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,16,2,power_law_1.01,0.264456958770752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,16,2,power_law_1.01,0.27040000915527346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,16,2,power_law_1.01,0.28909183502197267
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,16,2,power_law_1.01,0.30520191192626955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,16,2,power_law_1.01,0.32529151916503907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,16,2,power_law_1.01,0.37632511138916014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,16,2,power_law_1.01,0.4222016143798828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,16,2,power_law_1.01,0.5102143859863281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,16,2,power_law_1.01,0.5966527938842774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,16,2,power_law_1.01,0.8082688140869141
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,16,2,power_law_1.01,0.05915520191192627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,16,2,power_law_1.01,0.9764339447021484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,16,2,power_law_1.01,0.06324992179870606
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,16,2,power_law_1.01,0.07687808036804199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,16,2,power_law_1.01,1.3661170959472657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,16,2,power_law_1.01,0.08027008056640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,16,2,power_law_1.01,0.09463168144226074
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,16,2,power_law_1.01,0.11129728317260743
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,16,2,power_law_1.01,0.11994367599487306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,16,2,power_law_1.01,1.7269247436523436
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,16,2,power_law_1.01,0.13758720397949217
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,16,2,power_law_1.01,0.13820032119750975
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,16,2,power_law_1.01,0.14142463684082032
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,16,2,power_law_1.01,0.15270400047302246
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,16,2,power_law_1.01,0.1492518424987793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,16,2,power_law_1.01,0.16171775817871095
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,16,2,power_law_1.01,0.16776704788208008
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,16,2,power_law_1.01,0.1745382308959961
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,16,2,power_law_1.01,0.18478208541870117
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,16,2,power_law_1.01,0.19861631393432616
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,16,2,power_law_1.01,0.1556057643890381
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,16,2,balanced,0.17189504623413085
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,16,2,balanced,0.5618547058105469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,16,2,balanced,0.4595955276489258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,16,2,balanced,0.7518937683105469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,16,2,power_law_1.01,0.2613657569885254
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,16,2,power_law_1.01,0.22321407318115236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,16,2,power_law_1.2,0.03585024118423462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,16,2,power_law_1.2,0.061212158203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,16,2,power_law_1.2,0.07602176189422608
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,16,2,power_law_1.01,0.3083942413330078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,16,2,power_law_1.2,0.09085056304931641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,16,2,power_law_1.2,0.10840064048767091
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,16,2,power_law_1.2,0.13935104370117188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,16,2,power_law_1.2,0.14519935607910156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,16,2,power_law_1.2,0.1698700714111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,16,2,power_law_1.2,0.17799808502197265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,16,2,power_law_1.2,0.18271488189697266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,16,2,power_law_1.2,0.19273727416992187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,16,2,power_law_1.2,0.19864959716796876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,16,2,power_law_1.2,0.21342208862304685
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,16,2,power_law_1.01,0.3918374252319336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,16,2,power_law_1.2,0.3104857635498047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,16,2,power_law_1.2,0.31514495849609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,16,2,power_law_1.2,0.32476928710937497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,16,2,power_law_1.2,0.33907711029052734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,16,2,power_law_1.2,0.3717772674560547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,16,2,power_law_1.2,0.41491840362548826
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,16,2,power_law_1.2,0.4983539199829101
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,16,2,power_law_1.2,0.5825689697265626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,16,2,power_law_1.2,0.742840347290039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,16,2,power_law_1.2,0.8864141082763671
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,16,2,power_law_1.2,1.2079078674316406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,16,2,power_law_1.01,0.5182118225097656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,16,2,power_law_1.2,0.04032383918762207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,16,2,power_law_1.2,1.5346022033691407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,16,2,power_law_1.2,0.06155903816223145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,16,2,power_law_1.2,0.0824563217163086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,16,2,power_law_1.2,2.106591949462891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,16,2,power_law_1.2,0.046069760322570805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,16,2,power_law_1.2,0.12056960105895995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,16,2,power_law_1.2,0.16027008056640626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,16,2,power_law_1.2,0.1648921585083008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,16,2,power_law_1.2,0.1907891273498535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,16,2,power_law_1.2,0.20083967208862305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,16,2,power_law_1.2,2.7603378295898438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,16,2,power_law_1.2,0.20854400634765624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,16,2,power_law_1.2,0.23090303421020505
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,16,2,power_law_1.2,0.23303808212280272
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,16,2,power_law_1.2,0.2595264053344727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,16,2,power_law_1.2,0.26249727249145505
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,16,2,power_law_1.2,0.2745331192016601
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,16,2,power_law_1.2,0.2938572883605957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,16,2,power_law_1.2,0.33360126495361325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,16,2,power_law_1.2,0.31334144592285157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,16,2,power_law_1.2,0.21608959197998048
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,16,2,power_law_1.2,0.382795524597168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,16,2,power_law_1.2,0.4283660888671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,16,2,power_law_1.2,0.5270374298095704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,16,2,power_law_1.2,0.6143462371826172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,16,2,power_law_1.2,0.8227507019042969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,16,2,power_law_1.2,0.05958015918731689
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,16,2,power_law_1.2,1.0000806427001954
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,16,2,power_law_1.2,0.06345727920532226
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,16,2,power_law_1.2,0.07218815803527831
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,16,2,power_law_1.2,1.3904396057128907
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,16,2,power_law_1.2,0.0800767993927002
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,16,2,power_law_1.2,0.09108096122741699
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,16,2,power_law_1.2,0.10710528373718262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,16,2,power_law_1.2,1.755230712890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,16,2,power_law_1.01,0.724632339477539
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,16,2,power_law_1.2,0.11125503540039064
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,16,2,power_law_1.2,0.13148544311523438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,16,2,power_law_1.2,0.1366207981109619
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,16,2,power_law_1.2,0.14573439598083496
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,16,2,power_law_1.2,0.14965375900268554
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,16,2,power_law_1.2,0.15088512420654296
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,16,2,power_law_1.2,0.16320127487182617
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,16,2,power_law_1.2,0.17389312744140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,16,2,power_law_1.2,0.18039424896240236
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,16,2,power_law_1.2,0.1896780776977539
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,16,2,power_law_1.2,0.21836544036865235
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,16,2,power_law_1.01,0.9201522827148437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,16,2,power_law_1.01,1.4456716918945314
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,16,2,power_law_1.2,0.15492480278015136
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,16,2,power_law_1.2,1.0338700866699218
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,16,2,power_law_1.01,1.7619955444335935
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,16,2,power_law_1.2,0.30516351699829103
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,16,2,power_law_1.2,0.378260498046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,32,1,balanced,0.04554815769195557
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,16,2,power_law_1.2,0.2495142364501953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,32,1,balanced,0.07062208175659179
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,32,1,balanced,0.08033663749694823
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,32,1,balanced,0.1008403205871582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,32,1,balanced,0.13824064254760743
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,32,1,balanced,0.21237823486328128
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,32,1,balanced,0.29742591857910156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,32,1,balanced,0.29659072875976566
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,16,2,power_law_1.2,0.6920601654052734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,16,2,power_law_1.2,0.5844390487670899
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,32,1,balanced,0.2991276741027832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,32,1,balanced,0.3059891128540039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,32,1,balanced,0.31044864654541016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,32,1,balanced,0.3133452796936035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,32,1,balanced,0.3194278335571289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,32,1,balanced,0.3282342529296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,32,1,balanced,0.3321273422241211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,32,1,balanced,0.3394675064086914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,32,1,balanced,0.4965216064453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,32,1,balanced,0.5273177719116211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,32,1,balanced,0.5535763168334962
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,32,1,balanced,0.6147654342651367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,32,1,balanced,0.6677484893798828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,16,2,power_law_1.2,1.47053955078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,32,1,balanced,0.7979225921630859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,32,1,balanced,0.04977215766906738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,32,1,balanced,1.0427884674072265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,32,1,balanced,0.05719168186187744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,32,1,balanced,0.07048448085784911
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,16,2,power_law_1.2,2.135458526611328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,32,1,balanced,0.10123328208923339
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,32,1,balanced,1.2865414428710937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,32,1,balanced,0.16321344375610353
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,32,1,balanced,0.4047027206420898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,32,1,balanced,0.27642112731933594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,32,1,balanced,0.40605438232421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,32,1,balanced,1.6383180236816408
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,32,1,balanced,0.4111756896972656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,32,1,balanced,0.41509632110595707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,32,1,balanced,0.41992446899414065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,32,1,balanced,0.42449089050292965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,32,1,balanced,0.4303577423095703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,32,1,balanced,2.2673362731933593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,32,1,balanced,0.43234622955322266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,32,1,balanced,0.4386278533935547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,32,1,balanced,0.4093548965454102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,32,1,balanced,0.5198611068725586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,32,1,balanced,0.5372998428344726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,32,1,balanced,0.5523884963989258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,32,1,balanced,0.589051513671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,32,1,balanced,0.6233350372314452
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,32,1,balanced,3.0992446899414063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,32,1,balanced,0.6972480010986328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,32,1,balanced,0.7678150177001953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,32,1,balanced,0.916332778930664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,32,1,balanced,1.4082733154296876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,32,1,balanced,1.6940985107421873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,32,1,balanced,2.3156095886230466
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,16,2,power_law_1.2,3.0760577392578123
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,32,1,balanced,0.09309696197509766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,32,1,balanced,0.09229632377624511
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,32,1,balanced,0.09388928413391114
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,32,1,balanced,0.09518848419189453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,32,1,balanced,0.12608896255493166
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,32,1,balanced,0.1324947166442871
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,32,1,balanced,0.105415678024292
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,32,1,balanced,0.14599103927612306
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,32,1,balanced,0.1118278408050537
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,32,1,balanced,0.09833855628967285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,32,1,balanced,0.11888383865356444
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,32,1,balanced,0.16512063980102537
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,32,1,balanced,0.18937856674194337
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,32,1,balanced,0.24218879699707033
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,32,1,balanced,0.30081024169921877
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,32,1,balanced,0.34093246459960935
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,32,1,balanced,0.43102977752685545
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,32,1,balanced,0.6047788619995117
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,32,1,balanced,0.7727168273925782
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,32,1,balanced,1.114172134399414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,32,1,balanced,1.4651097106933595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,32,1,power_law_1.01,0.048886399269104
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,32,1,power_law_1.01,0.05046783924102784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,32,1,power_law_1.01,0.07886079788208009
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,32,1,balanced,2.148603515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,32,1,power_law_1.01,0.10225983619689942
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,32,1,power_law_1.01,0.1329843235015869
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,32,1,power_law_1.01,0.17180992126464845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,32,1,power_law_1.01,0.19871488571166993
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,32,1,power_law_1.01,0.22985088348388674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,32,1,power_law_1.01,0.2399795150756836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,32,1,power_law_1.01,0.25488895416259766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,32,1,balanced,2.8414016723632813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,32,1,power_law_1.01,0.26129600524902347
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,32,1,power_law_1.01,0.2457209587097168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,32,1,power_law_1.01,0.3033510398864746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,32,1,power_law_1.01,0.32194496154785157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,32,1,power_law_1.01,0.27827264785766603
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,32,1,power_law_1.01,0.33942657470703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,32,1,power_law_1.01,0.4771027374267578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,32,1,power_law_1.01,0.5101785659790039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,32,1,power_law_1.01,0.5452998352050782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,32,1,power_law_1.01,0.6161894226074219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,32,1,power_law_1.01,0.6998636627197266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,32,1,balanced,4.233056640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,32,1,power_law_1.01,0.9032479858398437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,32,1,power_law_1.01,1.0405677032470702
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,32,1,power_law_1.01,0.049677438735961914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,32,1,power_law_1.01,1.361925811767578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,32,1,power_law_1.01,0.057414398193359376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,32,1,power_law_1.01,1.6859315490722657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,32,1,power_law_1.01,0.0693830394744873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,32,1,power_law_1.01,0.1037664031982422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,32,1,power_law_1.01,0.1558790397644043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,32,1,power_law_1.01,0.2222265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,32,1,power_law_1.01,2.338429412841797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,32,1,power_law_1.01,0.27541759490966794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,32,1,power_law_1.01,0.31349119186401364
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,32,1,power_law_1.01,0.3232064056396484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,32,1,power_law_1.01,0.33270145416259767
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,32,1,power_law_1.01,0.3448889541625976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,32,1,power_law_1.01,0.3478579330444336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,32,1,power_law_1.01,0.36581825256347655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,32,1,power_law_1.01,3.0049786376953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,32,1,power_law_1.01,0.3814611053466797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,32,1,power_law_1.01,0.38335617065429684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,32,1,power_law_1.01,0.39912895202636717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,32,1,power_law_1.01,0.46796031951904293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,32,1,power_law_1.01,0.4927155303955078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,32,1,power_law_1.01,0.5191059112548828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,32,1,power_law_1.01,0.5769593429565429
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,32,1,power_law_1.01,0.6351660919189454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,32,1,power_law_1.01,0.7725209808349609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,32,1,power_law_1.01,0.8892691040039062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,32,1,power_law_1.01,1.1468831634521484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,32,1,power_law_1.01,1.3912512207031251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,32,1,power_law_1.01,1.8867001342773437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,32,1,power_law_1.01,0.09276224136352539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,32,1,power_law_1.01,2.390614471435547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,32,1,power_law_1.01,0.09262528419494628
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,32,1,power_law_1.01,0.09387328147888183
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,32,1,power_law_1.01,0.09470527648925782
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,32,1,power_law_1.01,0.09828607559204101
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,32,1,power_law_1.01,0.1118886375427246
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,32,1,power_law_1.01,0.12581376075744627
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,32,1,power_law_1.01,0.10532159805297851
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,32,1,power_law_1.01,0.11893823623657226
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,32,1,power_law_1.01,0.13324031829833985
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,32,1,power_law_1.01,0.14623744010925294
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,32,1,power_law_1.01,0.16887168884277343
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,32,1,power_law_1.01,0.1896723175048828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,32,1,power_law_1.01,0.3397804641723633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,32,1,power_law_1.01,0.29722368240356445
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,32,1,power_law_1.01,0.43153854370117184
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,32,1,power_law_1.01,0.24162752151489258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,32,1,power_law_1.01,0.6036006546020507
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,32,1,power_law_1.01,0.7762489318847656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,32,1,power_law_1.01,1.1227302551269531
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,32,1,power_law_1.01,1.453187255859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,32,1,power_law_1.01,2.1493907165527344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,32,1,power_law_1.01,2.8368377685546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,32,1,power_law_1.2,0.04476480007171631
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,32,1,power_law_1.01,4.236126098632813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,32,1,power_law_1.2,0.048856959342956544
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,32,1,power_law_1.2,0.10159168243408204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,32,1,power_law_1.2,0.08325759887695312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,32,1,power_law_1.2,0.12836992263793945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,32,1,power_law_1.2,0.16054271697998046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,32,1,power_law_1.2,0.18755584716796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,32,1,power_law_1.2,0.2244486427307129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,32,1,power_law_1.2,0.23325759887695313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,32,1,power_law_1.2,0.24747327804565428
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,32,1,power_law_1.2,0.26443647384643554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,32,1,power_law_1.2,0.27827903747558597
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,32,1,power_law_1.2,0.3019840049743652
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,32,1,power_law_1.2,0.32620033264160153
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,32,1,power_law_1.2,0.3402809524536133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,32,1,power_law_1.2,0.4777907180786133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,32,1,power_law_1.2,0.514073600769043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,32,1,power_law_1.2,0.5484076690673828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,32,1,power_law_1.2,0.6272735977172852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,32,1,power_law_1.2,0.7236025238037109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,32,1,power_law_1.2,0.9101087951660156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,32,1,power_law_1.2,0.23595775604248045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,32,1,power_law_1.2,1.0526822662353514
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,32,1,power_law_1.2,1.374832000732422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,32,1,power_law_1.2,0.0488262414932251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,32,1,power_law_1.2,0.057019519805908206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,32,1,power_law_1.2,0.07305920124053955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,32,1,power_law_1.2,0.10334015846252442
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,32,1,power_law_1.2,2.3528582763671873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,32,1,power_law_1.2,0.14912063598632813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,32,1,power_law_1.2,3.0271615600585937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,32,1,power_law_1.2,0.26494911193847653
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,32,1,power_law_1.2,0.3100601577758789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,32,1,power_law_1.2,0.318209285736084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,32,1,power_law_1.2,0.32553150177001955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,32,1,power_law_1.2,0.210383358001709
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,32,1,power_law_1.2,0.33426624298095703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,32,1,power_law_1.2,0.34630207061767576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,32,1,power_law_1.2,0.3598067092895508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,32,1,power_law_1.2,0.3745516967773438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,32,1,power_law_1.2,0.3838719940185547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,32,1,power_law_1.2,0.3970662307739258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,32,1,power_law_1.2,0.4703372955322266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,32,1,power_law_1.2,1.6907366943359374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,32,1,power_law_1.2,0.5026278305053711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,32,1,power_law_1.2,0.531283187866211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,32,1,power_law_1.2,0.5925734329223633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,32,1,power_law_1.2,0.6512300872802734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,32,1,power_law_1.2,0.7999839782714844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,32,1,power_law_1.2,0.9166246032714843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,32,1,power_law_1.2,1.1668921661376952
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,32,1,power_law_1.2,1.4228192138671876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,32,1,power_law_1.2,1.934062042236328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,32,1,power_law_1.2,0.09267071723937989
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,32,1,power_law_1.2,2.419564208984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,32,1,power_law_1.2,0.09348544120788574
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,32,1,power_law_1.2,0.09476736068725586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,32,1,power_law_1.2,0.10540224075317384
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,32,1,power_law_1.2,0.11217215538024902
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,32,1,power_law_1.2,0.11993535995483398
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,32,1,power_law_1.2,0.1253932762145996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,32,1,power_law_1.2,0.13598143577575683
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,32,1,power_law_1.2,0.1462611198425293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,32,1,power_law_1.2,0.17655616760253906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,32,1,power_law_1.2,0.23450048446655272
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,32,1,power_law_1.2,0.09220992088317871
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,32,1,power_law_1.2,0.09870976448059082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,32,1,power_law_1.2,0.19536575317382812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,32,1,power_law_1.2,0.7759532928466797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,32,1,power_law_1.2,0.29780799865722657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,32,1,power_law_1.2,0.34414718627929686
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,32,1,power_law_1.2,0.4312255859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,32,1,power_law_1.2,0.6013343811035157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,32,1,power_law_1.2,1.1171699523925782
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,32,1,power_law_1.2,1.4554573059082032
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,32,1,power_law_1.2,2.151514892578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,32,1,power_law_1.2,2.8257080078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,32,balanced,0.024751360416412356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,32,balanced,0.024723200798034667
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,32,balanced,0.027942399978637695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,32,balanced,0.04879231929779053
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,32,balanced,0.06058368206024169
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,32,balanced,0.09171199798583984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,32,1,power_law_1.2,4.216280212402344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,32,balanced,0.09133824348449707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,32,balanced,0.09316224098205567
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,32,balanced,0.09577343940734864
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,32,balanced,0.09812992095947266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,32,balanced,0.10054143905639648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,32,balanced,0.1035200023651123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,32,balanced,0.10525952339172365
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,32,balanced,0.10964863777160645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,32,balanced,0.10184703826904298
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,32,balanced,0.10841983795166014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,32,balanced,0.11487615585327149
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,32,balanced,0.12623231887817382
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,32,balanced,0.13585920333862306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,32,balanced,0.025573120117187497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,32,balanced,0.178973445892334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,32,balanced,0.19101951599121095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,32,balanced,0.27723648071289064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,32,balanced,0.3334297561645508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,32,balanced,0.469431037902832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,32,balanced,0.02768383979797363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,32,balanced,0.5959078216552735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,32,balanced,0.027676160335540774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,32,balanced,0.02888832092285156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,32,balanced,0.028090879917144772
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,32,balanced,0.02964479923248291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,32,balanced,0.04745215892791748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,32,balanced,0.04824063777923584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,32,balanced,0.04863872051239014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,32,balanced,0.04889599800109863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,32,balanced,0.04332543849945068
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,32,balanced,0.050184960365295406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,32,balanced,0.05177855968475342
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,32,balanced,0.05140992164611816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,32,balanced,0.05638527870178223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,32,balanced,0.05917695999145508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,32,balanced,0.05318143844604493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,32,balanced,0.06437759876251221
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,32,balanced,0.05737599849700927
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,32,balanced,0.06227071762084961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,32,balanced,0.06828927993774414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,32,balanced,0.08974847793579102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,32,balanced,0.10627584457397461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,32,balanced,0.07471487998962403
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,32,balanced,0.15145343780517578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,32,balanced,0.18103040695190428
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,32,balanced,0.24751487731933594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,32,balanced,0.3233241653442383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,32,balanced,0.13274368286132812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,1,32,balanced,0.017853440046310426
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,1,32,balanced,0.01789952039718628
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,1,32,balanced,0.024245760440826415
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,1,32,balanced,0.024447999000549316
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,1,32,balanced,0.02440063953399658
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,1,32,balanced,0.024555521011352537
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,1,32,balanced,0.017859840393066408
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,1,32,balanced,0.02457472085952759
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,1,32,balanced,0.0171942400932312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,1,32,balanced,0.024515841007232666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,1,32,balanced,0.024737279415130615
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,1,32,balanced,0.024524800777435303
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,1,32,balanced,0.02453376054763794
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,1,32,balanced,0.02517888069152832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,1,32,balanced,0.024546558856964114
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,1,32,balanced,0.025199360847473144
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,1,32,balanced,0.026042881011962893
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,1,32,balanced,0.02619263887405395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,1,32,balanced,0.025912320613861083
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,1,32,balanced,0.026467840671539306
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,1,32,balanced,0.028002560138702393
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,1,32,balanced,0.030433280467987062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,1,32,balanced,0.01825152039527893
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,1,32,balanced,0.0653056001663208
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,1,32,balanced,0.05469696044921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,1,32,balanced,0.042004480361938476
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,1,32,balanced,0.04345344066619873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,32,power_law_1.01,0.05813759803771973
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,32,power_law_1.01,0.06125696182250977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,32,power_law_1.01,0.059840002059936524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,32,power_law_1.01,0.09247743606567382
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,32,power_law_1.01,0.09142527580261231
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,32,power_law_1.01,0.09017600059509277
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,32,power_law_1.01,0.08946432113647461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,32,power_law_1.01,0.0980134391784668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,32,power_law_1.01,0.09594240188598632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,32,power_law_1.01,0.10245375633239746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,32,power_law_1.01,0.10004351615905763
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,32,power_law_1.01,0.10718336105346679
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,32,power_law_1.01,0.11098112106323241
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,32,power_law_1.01,0.10557567596435546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,32,power_law_1.01,0.11198847770690919
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,32,power_law_1.01,0.1215667152404785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,32,power_law_1.01,0.14390527725219726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,32,power_law_1.01,0.18537727355957032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,32,power_law_1.01,0.23106048583984373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,32,power_law_1.01,0.2953638458251953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,32,power_law_1.01,0.35801856994628906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,32,power_law_1.01,0.1077990436553955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,32,power_law_1.01,0.6772774505615234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,32,power_law_1.01,1.0134873962402344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,32,power_law_1.01,0.5493056106567382
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,32,power_law_1.01,0.041692161560058595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,32,power_law_1.01,1.2979954528808595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,32,power_law_1.01,0.041706237792968746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,32,power_law_1.01,0.04242943763732911
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,32,power_law_1.01,0.04136576175689698
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,32,power_law_1.01,0.046900482177734376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,32,power_law_1.01,0.04525568008422852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,32,power_law_1.01,0.046132478713989254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,32,power_law_1.01,0.051868162155151366
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,32,power_law_1.01,0.04256896018981934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,32,power_law_1.01,0.04881279945373535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,32,power_law_1.01,0.0537497615814209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,32,power_law_1.01,0.06033152103424072
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,32,power_law_1.01,0.06319359779357911
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,32,power_law_1.01,0.05365119934082031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,32,power_law_1.01,0.05855616092681885
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,32,power_law_1.01,0.05700223922729493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,32,power_law_1.01,0.058178558349609374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,32,power_law_1.01,0.06506239891052246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,32,power_law_1.01,0.09944831848144531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,32,power_law_1.01,0.07683584213256836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,32,power_law_1.01,0.04787456035614014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,32,power_law_1.01,0.12044416427612306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,32,power_law_1.01,0.1500160026550293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,32,power_law_1.01,0.2043827247619629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,32,power_law_1.01,0.2495065689086914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,1,32,power_law_1.01,0.024099841117858886
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,32,power_law_1.01,0.392224006652832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,1,32,power_law_1.01,0.02395263910293579
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,1,32,power_law_1.01,0.024284160137176512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,32,power_law_1.01,0.4988991928100586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,1,32,power_law_1.01,0.02470400094985962
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,1,32,power_law_1.01,0.02430464029312134
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,1,32,power_law_1.01,0.023531520366668703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,1,32,power_law_1.01,0.023992319107055664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,32,power_law_1.01,0.7657638549804687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,1,32,power_law_1.01,0.023864319324493406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,1,32,power_law_1.01,0.023555839061737062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,1,32,power_law_1.01,0.024057600498199463
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,1,32,power_law_1.01,0.024035840034484862
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,1,32,power_law_1.01,0.02563456058502197
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,1,32,power_law_1.01,0.025614080429077146
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,1,32,power_law_1.01,0.025617918968200683
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,1,32,power_law_1.01,0.025772800445556642
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,1,32,power_law_1.01,0.030975999832153323
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,1,32,power_law_1.01,0.02392832040786743
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,1,32,power_law_1.01,0.028023040294647215
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,1,32,power_law_1.01,0.028947200775146488
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,1,32,power_law_1.01,0.03688575983047485
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,1,32,power_law_1.01,0.04157567977905273
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,1,32,power_law_1.01,0.05232511997222901
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,32,power_law_1.2,0.058303999900817874
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,1,32,power_law_1.01,0.06380671977996825
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,32,power_law_1.2,0.08798463821411133
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,1,32,power_law_1.01,0.07688831806182861
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,1,32,power_law_1.01,0.11542655944824218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,32,power_law_1.2,0.08487168312072754
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,1,32,power_law_1.01,0.09676799774169922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,32,power_law_1.2,0.09351424217224122
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,32,power_law_1.2,0.09439488410949706
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,32,power_law_1.2,0.09813247680664063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,32,power_law_1.2,0.10202752113342287
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,32,power_law_1.2,0.10223872184753417
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,1,32,power_law_1.01,0.1720332717895508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,32,power_law_1.2,0.10642560005187988
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,32,power_law_1.2,0.058175997734069826
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,32,power_law_1.2,0.11186431884765624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,32,power_law_1.2,0.10507136344909669
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,32,power_law_1.2,0.10852224349975585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,32,power_law_1.2,0.13864831924438475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,32,power_law_1.2,0.11553407669067384
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,32,power_law_1.2,0.16696191787719727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,32,power_law_1.2,0.09261183738708496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,32,power_law_1.2,0.22238719940185545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,32,power_law_1.2,0.25390207290649414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,32,power_law_1.2,0.3767155075073242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,32,power_law_1.2,0.48690048217773435
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,32,power_law_1.2,0.041923837661743166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,32,power_law_1.2,0.7272716522216797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,32,power_law_1.2,0.8823334503173829
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,32,power_law_1.2,0.043148798942565916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,32,power_law_1.2,0.04290815830230713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,32,power_law_1.2,0.04440703868865967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,32,power_law_1.2,0.046933760643005376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,32,power_law_1.2,0.04323455810546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,32,power_law_1.2,0.045564160346984864
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,32,power_law_1.2,0.051324157714843756
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,32,power_law_1.2,0.047997441291809086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,32,power_law_1.2,0.04993152141571045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,32,power_law_1.2,0.04855807781219482
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,32,power_law_1.2,0.05511295795440674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,32,power_law_1.2,0.05822847843170166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,32,power_law_1.2,0.05877247810363769
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,32,power_law_1.2,0.055696640014648434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,32,power_law_1.2,0.056156158447265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,32,power_law_1.2,0.06165887832641601
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,32,power_law_1.2,0.07655680179595947
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,32,power_law_1.2,0.09464192390441895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,32,power_law_1.2,1.9126860046386718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,32,power_law_1.2,0.1328947162628174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,32,power_law_1.2,0.18800256729125978
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,32,power_law_1.2,0.24708223342895508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,32,power_law_1.2,1.3448410034179688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,32,power_law_1.2,0.3688691329956055
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,1,32,power_law_1.2,0.024284160137176512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,32,power_law_1.2,0.501605110168457
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,1,32,power_law_1.2,0.024522240161895754
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,1,32,power_law_1.2,0.024691200256347655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,32,power_law_1.2,0.11676287651062012
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,1,32,power_law_1.2,0.024208641052246092
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,1,32,power_law_1.2,0.021856000423431395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,1,32,power_law_1.2,0.02348288059234619
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,32,power_law_1.2,0.9741824340820312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,1,32,power_law_1.2,0.02355839967727661
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,1,32,power_law_1.2,0.024312319755554198
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,1,32,power_law_1.2,0.023486719131469727
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,1,32,power_law_1.2,0.024285440444946287
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,1,32,power_law_1.2,0.025153279304504395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,1,32,power_law_1.2,0.025498878955841065
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,1,32,power_law_1.2,0.025569279193878174
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,1,32,power_law_1.2,0.02593024015426636
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,1,32,power_law_1.2,0.02759936094284058
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,1,32,power_law_1.2,0.028409600257873535
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,1,32,power_law_1.2,0.031115520000457763
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,1,32,power_law_1.2,0.034773759841918946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,32,power_law_1.2,0.6747071838378906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,32,power_law_1.2,0.05978879928588867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,32,power_law_1.2,0.058790397644042966
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,1,32,power_law_1.2,0.04482816219329834
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,1,32,power_law_1.2,0.05183104038238525
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,1,32,power_law_1.2,0.06435840129852295
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,1,32,power_law_1.2,0.08068608283996583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,2,16,balanced,0.023568639755249025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,2,16,balanced,0.024780800342559813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,2,16,balanced,0.02627583980560303
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,2,16,balanced,0.04502399921417236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,2,16,balanced,0.06127871990203857
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,2,16,balanced,0.08718720436096192
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,2,16,balanced,0.023500800132751465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,2,16,balanced,0.09095295906066894
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,1,32,power_law_1.2,0.10747008323669434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,2,16,balanced,0.0919654369354248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,2,16,balanced,0.09379327774047852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,2,16,balanced,0.09234432220458985
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,1,32,power_law_1.2,0.024058880805969237
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,2,16,balanced,0.10142975807189943
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,2,16,balanced,0.09552895545959472
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,2,16,balanced,0.10272000312805177
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,2,16,balanced,0.09734911918640136
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,2,16,balanced,0.0881766414642334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,2,16,balanced,0.1183347225189209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,2,16,balanced,0.09816448211669922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,2,16,balanced,0.13912832260131835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,2,16,balanced,0.11068415641784668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,2,16,balanced,0.20569856643676757
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,2,16,balanced,0.24402559280395505
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,2,16,balanced,0.1568998432159424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,2,16,balanced,0.34941055297851564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,2,16,balanced,0.434785270690918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,2,16,balanced,0.620761604309082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,2,16,balanced,0.026771199703216553
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,2,16,balanced,0.026785280704498295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,2,16,balanced,0.7995724487304687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,2,16,balanced,0.028419840335845947
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,2,16,balanced,0.030229759216308594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,2,16,balanced,0.02804608106613159
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,2,16,balanced,0.048728318214416505
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,2,16,balanced,0.06074495792388916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,2,16,balanced,0.056696319580078126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,2,16,balanced,0.05700352191925049
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,2,16,balanced,0.05421311855316162
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,2,16,balanced,0.057644801139831545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,2,16,balanced,0.06356480121612548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,2,16,balanced,0.0573145580291748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,2,16,balanced,0.05885695934295654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,2,16,balanced,0.06173952102661133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,2,16,balanced,0.06438528060913086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,2,16,balanced,0.07140863895416259
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,2,16,balanced,0.08237312316894532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,2,16,balanced,0.0957209587097168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,2,16,balanced,0.12082943916320801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,2,16,balanced,0.14518400192260744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,2,16,balanced,0.21364608764648438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,2,16,balanced,0.26183807373046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,2,16,balanced,0.3607577514648438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,2,16,balanced,0.061795840263366694
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,2,16,balanced,0.47932926177978513
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,2,16,balanced,0.05380735874176026
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,1,32,power_law_1.2,0.1336473560333252
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,1,32,power_law_1.2,0.19937919616699218
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,1,32,power_law_1.2,0.29124351501464846
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,2,16,balanced,0.018592000007629395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,2,16,balanced,0.027568640708923342
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,2,16,balanced,0.019063040018081664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,2,16,balanced,0.02785536050796509
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,2,16,balanced,0.027485439777374266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,2,16,balanced,0.02764287948608398
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,2,16,balanced,0.027400960922241212
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,2,16,balanced,0.02819711923599243
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,2,16,balanced,0.02797951936721802
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,2,16,balanced,0.028458240032196047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,2,16,balanced,0.02754816055297852
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,2,16,balanced,0.028880639076232912
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,2,16,balanced,0.028956160545349118
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,2,16,balanced,0.030161919593811037
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,2,16,balanced,0.030920960903167725
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,2,16,balanced,0.033949439525604246
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,2,16,balanced,0.020679678916931152
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,2,16,balanced,0.03419136047363282
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,2,16,balanced,0.017994240522384644
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,2,16,balanced,0.01920639991760254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,2,16,power_law_1.01,0.0448089599609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,2,16,power_law_1.01,0.05791744232177735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,2,16,power_law_1.01,0.05741184234619141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,2,16,power_law_1.01,0.05810815811157226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,2,16,power_law_1.01,0.061169919967651365
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,2,16,power_law_1.01,0.08398719787597655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,2,16,power_law_1.01,0.08621952056884766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,2,16,power_law_1.01,0.08726783752441405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,2,16,power_law_1.01,0.08904319763183594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,2,16,balanced,0.02601855993270874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,2,16,power_law_1.01,0.09112064361572265
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,2,16,balanced,0.026078720092773438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,2,16,power_law_1.01,0.09412096023559571
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,2,16,power_law_1.01,0.09886848449707031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,2,16,power_law_1.01,0.09229567527770996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,2,16,power_law_1.01,0.09561727523803712
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,2,16,power_law_1.01,0.10272895812988281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,2,16,power_law_1.01,0.09959936141967773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,2,16,power_law_1.01,0.10552319526672363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,2,16,power_law_1.01,0.12020735740661621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,2,16,power_law_1.01,0.1382361602783203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,2,16,power_law_1.01,0.17446783065795898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,2,16,power_law_1.01,0.22195711135864257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,2,16,power_law_1.01,0.2929587173461914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,2,16,power_law_1.01,0.36451839447021483
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,2,16,power_law_1.01,0.5050956726074218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,2,16,power_law_1.01,0.6186764907836915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,2,16,power_law_1.01,1.037693405151367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,2,16,power_law_1.01,0.030072319507598876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,2,16,power_law_1.01,0.044431362152099606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,2,16,power_law_1.01,1.3152319335937501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,2,16,power_law_1.01,0.04680191993713379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,2,16,power_law_1.01,0.04596992015838623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,2,16,power_law_1.01,0.04828544139862061
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,2,16,power_law_1.01,0.048095998764038084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,2,16,power_law_1.01,0.051212801933288574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,2,16,power_law_1.01,0.05416319847106934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,2,16,power_law_1.01,0.05531263828277588
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,2,16,power_law_1.01,0.057396478652954094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,2,16,power_law_1.01,0.061283841133117675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,2,16,power_law_1.01,0.04940800189971924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,2,16,power_law_1.01,0.062497282028198244
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,2,16,balanced,0.026860799789428715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,2,16,power_law_1.01,0.057839360237121586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,2,16,power_law_1.01,0.05737343788146972
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,2,16,power_law_1.01,0.059169278144836426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,2,16,power_law_1.01,0.061571841239929204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,2,16,power_law_1.01,0.07016064167022705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,2,16,power_law_1.01,0.07685887813568115
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,2,16,power_law_1.01,0.09996159553527832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,2,16,power_law_1.01,0.12240384101867677
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,2,16,power_law_1.01,0.16467199325561524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,2,16,power_law_1.01,0.1931443214416504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,2,16,power_law_1.01,0.29966848373413085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,2,16,power_law_1.01,0.33774337768554685
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,2,16,power_law_1.01,0.020665600299835205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,2,16,power_law_1.01,0.546316795349121
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,2,16,power_law_1.01,0.02690176010131836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,2,16,power_law_1.01,0.0271616005897522
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,2,16,power_law_1.01,0.7024960327148437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,2,16,power_law_1.01,0.02722304105758667
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,2,16,power_law_1.01,0.027399680614471435
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,2,16,power_law_1.01,0.024764161109924316
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,2,16,power_law_1.01,0.027408640384674072
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,2,16,power_law_1.01,0.02689536094665527
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,2,16,power_law_1.01,0.02763391971588135
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,2,16,power_law_1.01,0.02715519905090332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,2,16,power_law_1.01,0.0273907208442688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,2,16,power_law_1.01,0.028417279720306398
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,2,16,power_law_1.01,0.028195838928222656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,2,16,power_law_1.01,0.028933119773864747
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,2,16,power_law_1.01,0.02887295961380005
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,2,16,power_law_1.01,0.027228159904479982
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,2,16,balanced,0.049387521743774414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,2,16,balanced,0.06540800094604492
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,2,16,balanced,0.05310207843780518
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,2,16,balanced,0.08110079765319825
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,2,16,power_law_1.01,0.03054080009460449
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,2,16,power_law_1.01,0.03657344102859497
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,2,16,power_law_1.01,0.03320192098617554
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,2,16,power_law_1.01,0.043150081634521484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,2,16,power_law_1.2,0.04442368030548095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,2,16,power_law_1.2,0.05742080211639404
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,2,16,power_law_1.2,0.05523712158203124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,2,16,power_law_1.2,0.057538561820983894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,2,16,power_law_1.2,0.06051583766937256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,2,16,power_law_1.2,0.08443391799926758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,2,16,power_law_1.2,0.0884006404876709
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,2,16,power_law_1.2,0.0886400032043457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,2,16,power_law_1.2,0.08979711532592774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,2,16,power_law_1.2,0.09077759742736816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,2,16,power_law_1.2,0.09346816062927246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,2,16,power_law_1.2,0.10162303924560548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,2,16,power_law_1.2,0.10464384078979491
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,2,16,power_law_1.2,0.0930073642730713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,2,16,power_law_1.2,0.09960831642150879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,2,16,power_law_1.2,0.10288255691528321
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,2,16,power_law_1.2,0.11133952140808105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,2,16,power_law_1.2,0.12515583992004395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,2,16,power_law_1.2,0.1427340793609619
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,2,16,power_law_1.01,0.04723968029022217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,2,16,power_law_1.2,0.18273664474487306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,2,16,power_law_1.2,0.25743999481201174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,2,16,power_law_1.2,0.3213030242919922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,2,16,power_law_1.2,0.4154777526855469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,2,16,power_law_1.01,0.09007488250732422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,2,16,power_law_1.2,0.6080332946777344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,2,16,power_law_1.01,0.057918720245361335
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,2,16,power_law_1.01,0.07018112182617188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,2,16,power_law_1.2,0.7632959747314454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,2,16,power_law_1.2,0.02969599962234497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,2,16,power_law_1.2,0.043612160682678223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,2,16,power_law_1.2,0.043452157974243164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,2,16,power_law_1.2,0.048439040184020996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,2,16,power_law_1.2,1.2369996643066405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,2,16,power_law_1.2,0.042758398056030274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,2,16,power_law_1.2,0.04659071922302246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,2,16,power_law_1.2,0.05268735885620117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,2,16,power_law_1.2,0.051022081375122076
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,2,16,power_law_1.2,0.052133121490478516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,2,16,power_law_1.2,0.05679872035980225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,2,16,power_law_1.2,0.058246397972106935
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,2,16,power_law_1.2,0.06348415851593017
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,2,16,power_law_1.2,0.06283775806427003
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,2,16,power_law_1.2,0.055770878791809086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,2,16,power_law_1.2,0.058869757652282716
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,2,16,power_law_1.2,1.6823744201660156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,2,16,power_law_1.2,0.05995007991790772
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,2,16,power_law_1.2,0.06349311828613281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,2,16,power_law_1.2,0.07361279964447022
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,2,16,power_law_1.2,0.08469504356384278
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,2,16,power_law_1.2,0.10225919723510742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,2,16,power_law_1.2,0.1303667163848877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,2,16,power_law_1.2,0.17237888336181642
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,2,16,power_law_1.2,0.24065664291381833
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,2,16,power_law_1.2,0.35921024322509765
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,2,16,power_law_1.2,0.020700159072875975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,2,16,power_law_1.2,0.4495206451416015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,2,16,power_law_1.2,0.027308800220489503
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,2,16,power_law_1.2,0.02718591928482056
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,2,16,power_law_1.2,0.6367398452758789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,2,16,power_law_1.2,0.02717184066772461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,2,16,power_law_1.2,0.026435840129852294
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,2,16,power_law_1.2,0.027216639518737797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,2,16,power_law_1.2,0.02608383893966675
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,2,16,power_law_1.2,0.028025600910186764
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,2,16,power_law_1.2,0.9101324462890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,2,16,power_law_1.2,0.027127039432525635
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,2,16,power_law_1.2,0.027630081176757814
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,2,16,power_law_1.2,0.027733759880065916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,2,16,power_law_1.2,0.028012800216674804
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,2,16,power_law_1.2,0.028188159465789796
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,2,16,power_law_1.2,0.02961024045944214
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,2,16,power_law_1.2,0.03070208072662354
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,2,16,power_law_1.2,0.02845567941665649
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,2,16,power_law_1.01,0.11898112297058105
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,2,16,power_law_1.01,0.15543168067932128
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,2,16,power_law_1.01,0.23024127960205076
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,2,16,power_law_1.2,0.03224319934844971
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,2,16,power_law_1.2,0.03590656042098999
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,2,16,power_law_1.2,0.03938047885894776
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,2,16,power_law_1.2,0.049689598083496093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,4,8,balanced,0.02478208065032959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,4,8,balanced,0.023508479595184324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,4,8,balanced,0.024940800666809083
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,4,8,balanced,0.0327948808670044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,4,8,balanced,0.046489601135253904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,4,8,balanced,0.06573823928833009
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,2,16,power_law_1.2,0.05800064086914063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,4,8,balanced,0.09142271995544433
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,4,8,balanced,0.0935756778717041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,4,8,balanced,0.0961023998260498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,4,8,balanced,0.10027263641357423
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,4,8,balanced,0.09071231842041015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,4,8,balanced,0.09332351684570313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,4,8,balanced,0.09618176460266113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,4,8,balanced,0.09959936141967773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,4,8,balanced,0.10261631965637208
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,4,8,balanced,0.10641023635864258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,4,8,balanced,0.11300479888916015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,4,8,balanced,0.1293951988220215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,4,8,balanced,0.14617600440979003
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,4,8,balanced,0.1774323272705078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,4,8,balanced,0.2098624038696289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,4,8,balanced,0.2813145637512207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,4,8,balanced,0.34778751373291017
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,4,8,balanced,0.5113433456420898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,4,8,balanced,0.645318374633789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,4,8,balanced,0.9400243377685547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,4,8,balanced,0.026062080860137938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,4,8,balanced,0.02682751893997192
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,4,8,balanced,1.2219967651367187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,4,8,balanced,0.02904320001602173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,4,8,balanced,0.03127808094024658
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,4,8,balanced,0.035100159645080564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,4,8,balanced,0.055080962181091306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,4,8,balanced,0.06118271827697754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,4,8,balanced,0.0636288022994995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,4,8,balanced,0.06737279891967773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,4,8,balanced,0.06828288078308106
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,4,8,balanced,0.058872318267822264
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,2,16,power_law_1.2,0.06623104095458984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,4,8,balanced,0.06006783962249755
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,4,8,balanced,0.062045440673828126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,4,8,balanced,0.0640012788772583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,4,8,balanced,0.06697728157043456
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,2,16,power_law_1.2,0.08207103729248047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,4,8,balanced,0.06820608139038085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,4,8,balanced,0.07430399894714355
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,2,16,power_law_1.2,0.1257408046722412
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,4,8,balanced,0.08425600051879882
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,4,8,balanced,0.09477631568908691
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,4,8,balanced,0.11644288063049317
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,4,8,balanced,0.13862272262573244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,4,8,balanced,0.18317823410034179
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,4,8,balanced,0.22482559204101565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,4,8,balanced,0.3348275375366211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,4,8,balanced,0.41808513641357425
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,4,8,balanced,0.5925363159179687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,4,8,balanced,0.7812544250488281
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,2,16,power_law_1.2,0.2753843116760254
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,2,16,power_law_1.2,0.1554969596862793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,2,16,power_law_1.2,0.33637630462646484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,4,8,balanced,0.021067519187927247
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,4,8,balanced,0.021274878978729247
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,4,8,balanced,0.03534463882446289
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,4,8,balanced,0.03524480104446411
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,4,8,balanced,0.022695679664611817
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,4,8,balanced,0.03575551986694336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,4,8,balanced,0.021505279541015627
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,4,8,balanced,0.03570303916931152
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,4,8,balanced,0.03614720106124878
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,4,8,balanced,0.03623296022415161
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,4,8,balanced,0.03689215898513794
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,4,8,balanced,0.0371891188621521
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,4,8,balanced,0.038545920848846435
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,4,8,balanced,0.04033664226531982
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,4,8,balanced,0.04128896236419678
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,4,8,balanced,0.04339712142944336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,4,8,balanced,0.046163201332092285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,4,8,balanced,0.0496665620803833
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,4,8,balanced,0.02477695941925049
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,4,8,balanced,0.03609855890274048
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,4,8,power_law_1.01,0.027635200023651125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,4,8,power_law_1.01,0.045416960716247554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,4,8,power_law_1.01,0.04868735790252686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,4,8,power_law_1.01,0.052540159225463866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,4,8,power_law_1.01,0.058661122322082516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,4,8,power_law_1.01,0.06450687885284424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,4,8,power_law_1.01,0.08740863800048829
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,4,8,power_law_1.01,0.08888064384460449
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,4,8,power_law_1.01,0.08995327949523926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,4,8,power_law_1.01,0.09476096153259277
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,4,8,power_law_1.01,0.09971072196960448
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,4,8,power_law_1.01,0.09499391555786132
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,4,8,power_law_1.01,0.09134207725524902
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,4,8,power_law_1.01,0.09462528228759766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,4,8,power_law_1.01,0.09772928237915039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,4,8,power_law_1.01,0.10057727813720703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,4,8,power_law_1.01,0.10759807586669923
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,4,8,power_law_1.01,0.1173132801055908
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,4,8,power_law_1.01,0.13722880363464357
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,4,8,power_law_1.01,0.1650111961364746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,4,8,power_law_1.01,0.19806720733642577
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,4,8,balanced,0.03537535905838013
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,4,8,power_law_1.01,0.2987660789489746
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,4,8,balanced,0.03503999948501587
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,4,8,balanced,0.035392000675201415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,4,8,power_law_1.01,0.3467264175415039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,4,8,power_law_1.01,0.4989772796630859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,4,8,power_law_1.01,0.028471040725708007
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,4,8,power_law_1.01,0.6665715026855469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,4,8,power_law_1.01,0.03274496078491211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,4,8,power_law_1.01,0.9622592163085937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,4,8,power_law_1.01,0.037857279777526856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,4,8,power_law_1.01,0.040755200386047366
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,4,8,power_law_1.01,0.05147391796112061
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,4,8,power_law_1.01,1.1161420440673828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,4,8,power_law_1.01,0.054009599685668944
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,4,8,power_law_1.01,0.0567142391204834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,4,8,power_law_1.01,0.05861248016357422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,4,8,power_law_1.01,0.06128767967224121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,4,8,power_law_1.01,0.0655679988861084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,4,8,power_law_1.01,0.06601344108581543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,4,8,power_law_1.01,0.05704832077026367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,4,8,power_law_1.01,0.06110335826873779
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,4,8,power_law_1.01,0.059079680442810055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,4,8,power_law_1.01,0.0631654405593872
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,4,8,power_law_1.01,0.06739071846008302
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,4,8,power_law_1.01,0.07647232055664063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,4,8,power_law_1.01,0.05232511997222901
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,4,8,power_law_1.01,0.09127552032470702
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,4,8,power_law_1.01,0.1103718376159668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,4,8,power_law_1.01,0.1374681568145752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,4,8,power_law_1.01,0.17740287780761718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,4,8,power_law_1.01,0.2206630325317383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,4,8,power_law_1.01,0.32042240142822265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,4,8,power_law_1.01,0.3682585525512695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,4,8,power_law_1.01,0.023077120780944822
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,4,8,power_law_1.01,0.5431423950195312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,4,8,power_law_1.01,0.024358398914337158
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,4,8,power_law_1.01,0.027307519912719725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,4,8,power_law_1.01,0.6763673400878906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,4,8,power_law_1.01,0.028304638862609866
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,4,8,power_law_1.01,0.03110016107559204
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,4,8,power_law_1.01,0.029795839786529544
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,4,8,power_law_1.01,0.032254719734191896
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,4,8,power_law_1.01,0.03237760066986084
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,4,8,power_law_1.01,0.032071681022644044
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,4,8,power_law_1.01,0.032111361026763915
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,4,8,power_law_1.01,0.03282047986984253
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,4,8,power_law_1.01,0.0331110405921936
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,4,8,power_law_1.01,0.03229952096939087
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,4,8,power_law_1.01,0.03352960109710693
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,4,8,power_law_1.01,0.03425920009613037
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,4,8,power_law_1.01,0.03581568002700806
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,4,8,power_law_1.01,0.03824127912521362
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,4,8,balanced,0.06706943988800049
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,4,8,balanced,0.07281792163848877
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,4,8,balanced,0.12150015830993652
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,4,8,balanced,0.09336319923400879
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,4,8,power_law_1.01,0.04242688179016113
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,4,8,power_law_1.01,0.0546227216720581
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,4,8,power_law_1.01,0.046708478927612304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,4,8,power_law_1.2,0.027583999633789064
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,4,8,power_law_1.01,0.06494719982147217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,4,8,power_law_1.2,0.04387328147888184
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,4,8,power_law_1.2,0.04506239891052246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,4,8,power_law_1.2,0.050647039413452145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,4,8,power_law_1.2,0.06047999858856201
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,4,8,power_law_1.2,0.060392961502075196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,4,8,power_law_1.2,0.08501888275146484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,4,8,power_law_1.2,0.08577152252197265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,4,8,power_law_1.2,0.09198464393615723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,4,8,power_law_1.2,0.09468544006347657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,4,8,power_law_1.2,0.0974233627319336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,4,8,power_law_1.2,0.09362943649291992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,4,8,power_law_1.2,0.09241087913513184
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,4,8,power_law_1.2,0.09604351997375489
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,4,8,power_law_1.2,0.10034175872802735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,4,8,power_law_1.2,0.10364031791687012
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,4,8,power_law_1.2,0.10801152229309081
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,4,8,power_law_1.2,0.12303487777709961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,4,8,power_law_1.2,0.13942272186279298
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,4,8,power_law_1.2,0.18968576431274414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,4,8,power_law_1.01,0.07600384235382081
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,4,8,power_law_1.2,0.21938688278198243
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,4,8,power_law_1.01,0.08865535736083985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,4,8,power_law_1.2,0.3223244857788086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,4,8,power_law_1.2,0.39186561584472657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,4,8,power_law_1.01,0.11738752365112304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,4,8,power_law_1.2,0.5197324752807617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,4,8,power_law_1.2,0.7269158172607422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,4,8,power_law_1.2,0.029300479888916015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,4,8,power_law_1.2,0.03256831884384155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,4,8,power_law_1.2,0.03535360097885132
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,4,8,power_law_1.2,0.967462387084961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,4,8,power_law_1.2,0.039607040882110596
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,4,8,power_law_1.2,0.048186879158020016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,4,8,power_law_1.2,0.05369984149932862
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,4,8,power_law_1.2,0.055694079399108885
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,4,8,power_law_1.2,0.05684224128723144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,4,8,power_law_1.2,1.4406553649902345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,4,8,power_law_1.2,0.05771903991699219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,4,8,power_law_1.2,0.06119808197021485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,4,8,power_law_1.2,0.06579584121704102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,4,8,power_law_1.2,0.06122111797332763
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,4,8,power_law_1.2,0.05681791782379151
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,4,8,power_law_1.2,0.059211521148681646
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,4,8,power_law_1.2,0.061146879196166994
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,4,8,power_law_1.2,0.062449917793273926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,4,8,power_law_1.2,0.06975488185882568
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,4,8,power_law_1.2,0.07875711917877197
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,4,8,power_law_1.2,0.09365887641906738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,4,8,power_law_1.2,0.11162879943847656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,4,8,power_law_1.2,0.14683135986328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,4,8,power_law_1.2,0.20083328247070314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,4,8,power_law_1.2,0.2398489570617676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,4,8,power_law_1.2,0.3256332778930664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,4,8,power_law_1.2,0.02289535999298096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,4,8,power_law_1.2,0.43199615478515624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,4,8,power_law_1.2,0.02439039945602417
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,4,8,power_law_1.2,0.608851203918457
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,4,8,power_law_1.2,0.028017919063568115
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,4,8,power_law_1.2,0.031296000480651856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,4,8,power_law_1.2,0.8013798522949219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,4,8,power_law_1.2,0.03304831981658936
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,4,8,power_law_1.2,0.031059200763702395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,4,8,power_law_1.01,0.15470848083496094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,4,8,power_law_1.2,0.031946239471435545
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,4,8,power_law_1.2,0.0318668794631958
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,4,8,power_law_1.2,0.03329279899597168
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,4,8,power_law_1.2,0.03439487934112549
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,4,8,power_law_1.2,0.032838399410247805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,4,8,power_law_1.2,0.035750401020050046
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,4,8,power_law_1.2,0.033602559566497804
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,4,8,power_law_1.2,0.03702271938323974
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,4,8,power_law_1.2,0.05175551891326904
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,4,8,power_law_1.2,0.026019840240478515
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,4,8,power_law_1.2,0.03229824066162109
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,4,8,power_law_1.01,0.31878143310546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,4,8,power_law_1.01,0.22204671859741212
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,4,8,power_law_1.2,0.046647038459777836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,4,8,power_law_1.2,0.04106368064880371
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,4,8,power_law_1.2,0.08117504119873047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,4,8,power_law_1.2,0.10063615798950196
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,4,8,power_law_1.2,0.06258944034576416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,8,4,balanced,0.021932799816131592
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,8,4,balanced,0.0256166410446167
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,8,4,balanced,0.028076798915863038
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,8,4,balanced,0.048922882080078126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,8,4,balanced,0.06388095855712891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,8,4,balanced,0.06427648067474365
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,8,4,balanced,0.06399104118347168
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,4,8,power_law_1.2,0.12238335609436035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,8,4,balanced,0.09250816345214843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,8,4,balanced,0.09288576126098633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,8,4,balanced,0.09549951553344727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,8,4,balanced,0.09744256019592286
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,8,4,balanced,0.10031871795654297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,8,4,balanced,0.10240768432617187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,8,4,balanced,0.1056601619720459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,8,4,balanced,0.10879615783691406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,8,4,balanced,0.11468159675598144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,8,4,balanced,0.12852352142333984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,8,4,balanced,0.14456959724426272
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,8,4,balanced,0.17446271896362303
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,8,4,balanced,0.20504703521728515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,8,4,balanced,0.2688230323791504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,8,4,balanced,0.33111297607421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,8,4,balanced,0.4779724884033203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,8,4,balanced,0.6197350311279297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,8,4,balanced,0.022672638893127442
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,8,4,balanced,0.026397440433502194
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,8,4,balanced,0.026004478931427
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,8,4,balanced,0.9002444458007812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,8,4,balanced,0.02964992046356201
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,8,4,balanced,0.03586688041687012
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,8,4,balanced,0.04322048187255859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,8,4,balanced,0.0627942419052124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,8,4,balanced,0.06264959812164307
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,8,4,balanced,0.06365952014923096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,8,4,balanced,0.07160831928253174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,8,4,balanced,0.07193984031677246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,8,4,balanced,1.1702272033691405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,8,4,balanced,0.07342463970184326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,8,4,balanced,0.07481472015380859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,8,4,balanced,0.07598720073699952
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,8,4,balanced,0.07804287910461426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,8,4,balanced,0.08047231674194336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,8,4,balanced,0.08249216079711914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,8,4,balanced,0.08596735954284668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,8,4,balanced,0.09630720138549805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,8,4,balanced,0.10585087776184081
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,8,4,balanced,0.12609024047851564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,8,4,balanced,0.14439423561096193
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,8,4,balanced,0.1836172866821289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,8,4,balanced,0.22536575317382815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,8,4,balanced,0.33630592346191407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,8,4,balanced,0.41579647064208985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,8,4,balanced,0.6053747177124024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,8,4,balanced,0.7909862518310546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,4,8,power_law_1.2,0.15829888343811035
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,4,8,power_law_1.2,0.232806396484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,4,8,power_law_1.2,0.45443073272705076
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,4,8,power_law_1.2,0.3361856079101563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,8,4,balanced,0.027720959186553956
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,8,4,balanced,0.04970240116119385
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,8,4,balanced,0.028485119342803955
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,8,4,balanced,0.04967936038970947
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,8,4,balanced,0.05004032135009766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,8,4,balanced,0.050382080078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,8,4,balanced,0.029646079540252685
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,8,4,balanced,0.05079936027526856
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,8,4,balanced,0.050476799011230464
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,8,4,balanced,0.050597119331359866
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,8,4,balanced,0.051385598182678224
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,8,4,balanced,0.05219967842102051
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,8,4,balanced,0.05313407897949218
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,8,4,balanced,0.05497471809387207
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,8,4,balanced,0.05726208209991455
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,8,4,balanced,0.0597657585144043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,8,4,balanced,0.06550015926361083
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,8,4,balanced,0.07054080009460449
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,8,4,balanced,0.03018496036529541
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,8,4,balanced,0.035033600330352785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,8,4,power_law_1.01,0.023521280288696288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,8,4,power_law_1.01,0.026812160015106203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,8,4,balanced,0.049630718231201174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,8,4,power_law_1.01,0.03937279939651489
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,8,4,power_law_1.01,0.04323455810546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,8,4,power_law_1.01,0.049964799880981445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,8,4,power_law_1.01,0.05421184062957764
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,8,4,power_law_1.01,0.059157757759094244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,8,4,power_law_1.01,0.059211521148681646
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,8,4,power_law_1.01,0.08698495864868164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,8,4,power_law_1.01,0.08792832374572754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,8,4,power_law_1.01,0.09099007606506347
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,8,4,power_law_1.01,0.09269120216369628
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,8,4,power_law_1.01,0.09576191902160644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,8,4,power_law_1.01,0.0990015983581543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,8,4,power_law_1.01,0.10157952308654786
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,8,4,power_law_1.01,0.10463616371154785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,8,4,power_law_1.01,0.11238016128540038
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,8,4,power_law_1.01,0.12781696319580077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,8,4,power_law_1.01,0.14421631813049315
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,8,4,power_law_1.01,0.17430784225463866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,8,4,power_law_1.01,0.20113151550292968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,8,4,power_law_1.01,0.2788006401062012
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,8,4,power_law_1.01,0.3665779113769531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,8,4,power_law_1.01,0.5013875198364258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,8,4,power_law_1.01,0.6420582580566406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,8,4,power_law_1.01,0.9404684448242188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,8,4,power_law_1.01,0.026794240474700925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,8,4,power_law_1.01,0.03130496025085449
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,8,4,power_law_1.01,1.1955213165283203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,8,4,power_law_1.01,0.03415424108505249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,8,4,power_law_1.01,0.041521921157836914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,8,4,power_law_1.01,0.04611839771270752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,8,4,power_law_1.01,0.04739583969116211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,8,4,power_law_1.01,0.0553433609008789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,8,4,power_law_1.01,0.05602816104888916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,8,4,power_law_1.01,0.06585343837738036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,8,4,power_law_1.01,0.06539648056030274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,8,4,power_law_1.01,0.06792960166931153
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,8,4,power_law_1.01,0.06857215881347656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,8,4,power_law_1.01,0.07056384086608887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,8,4,power_law_1.01,0.07529344081878661
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,8,4,power_law_1.01,0.07648640155792237
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,8,4,power_law_1.01,0.0784819221496582
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,8,4,balanced,0.04998144149780274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,8,4,power_law_1.01,0.08277376174926758
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,8,4,balanced,0.04975359916687012
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,8,4,power_law_1.01,0.1310848045349121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,8,4,power_law_1.01,0.10682239532470703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,8,4,power_law_1.01,0.16246784210205079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,8,4,power_law_1.01,0.20403072357177737
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,8,4,power_law_1.01,0.2509760093688965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,8,4,power_law_1.01,0.35867263793945314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,8,4,power_law_1.01,0.45672958374023437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,8,4,power_law_1.01,0.09310336112976074
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,8,4,power_law_1.01,0.029702401161193846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,8,4,power_law_1.01,0.6542259216308594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,8,4,power_law_1.01,0.03063807964324951
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,8,4,power_law_1.01,0.03247872114181519
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,8,4,power_law_1.01,0.03508608102798462
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,8,4,power_law_1.01,0.03702399969100952
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,8,4,power_law_1.01,0.813091812133789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,8,4,power_law_1.01,0.038251519203186035
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,8,4,power_law_1.01,0.04399744033813476
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,8,4,power_law_1.01,0.04485375881195068
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,8,4,power_law_1.01,0.0440064001083374
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,8,4,power_law_1.01,0.0458892822265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,8,4,power_law_1.01,0.04673664093017578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,8,4,power_law_1.01,0.04769536018371582
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,8,4,power_law_1.01,0.04950784206390381
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,8,4,power_law_1.01,0.050991358757019045
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,8,4,power_law_1.01,0.05149439811706543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,8,4,power_law_1.01,0.05367551803588867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,8,4,power_law_1.01,0.04316671848297119
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,8,4,balanced,0.1096947193145752
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,8,4,balanced,0.14658944129943846
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,8,4,balanced,0.187508487701416
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,8,4,balanced,0.09808511734008789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,8,4,power_law_1.01,0.06231808185577392
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,8,4,power_law_1.01,0.06883711814880371
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,8,4,power_law_1.01,0.0803046417236328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,8,4,power_law_1.2,0.02311808109283447
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,8,4,power_law_1.2,0.026023681163787844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,8,4,power_law_1.2,0.03834624052047729
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,8,4,power_law_1.01,0.09315711975097656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,8,4,power_law_1.2,0.04209536075592041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,8,4,power_law_1.2,0.051302399635314945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,8,4,power_law_1.2,0.05194623947143555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,8,4,power_law_1.01,0.13753600120544435
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,8,4,power_law_1.01,0.1124351978302002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,8,4,power_law_1.2,0.05896704196929932
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,8,4,power_law_1.2,0.08469120025634766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,8,4,power_law_1.2,0.08814080238342285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,8,4,power_law_1.2,0.09082367897033691
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,8,4,power_law_1.01,0.19046911239624023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,8,4,power_law_1.2,0.09370623588562012
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,8,4,power_law_1.2,0.0978112030029297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,8,4,power_law_1.2,0.10003199577331542
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,8,4,power_law_1.2,0.1040998363494873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,8,4,power_law_1.2,0.10813055992126466
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,8,4,power_law_1.2,0.11427328109741211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,8,4,power_law_1.2,0.1277900791168213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,8,4,power_law_1.2,0.14484352111816406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,8,4,power_law_1.2,0.17780223846435547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,8,4,power_law_1.2,0.05754496097564697
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,8,4,power_law_1.2,0.20973440170288088
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,8,4,power_law_1.2,0.2930303955078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,8,4,power_law_1.01,0.2501811218261719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,8,4,power_law_1.2,0.37510654449462894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,8,4,power_law_1.2,0.027630081176757814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,8,4,power_law_1.2,0.5201676940917969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,8,4,power_law_1.2,0.6576012420654297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,8,4,power_law_1.2,0.03237760066986084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,8,4,power_law_1.2,0.03273600101470947
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,8,4,power_law_1.2,0.04153855800628662
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,8,4,power_law_1.2,0.9497100830078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,8,4,power_law_1.2,0.04640128135681153
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,8,4,power_law_1.2,0.05667967796325683
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,8,4,power_law_1.2,0.058501119613647456
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,8,4,power_law_1.2,0.06401279926300049
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,8,4,power_law_1.2,0.06731904029846192
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,8,4,power_law_1.2,1.2933389282226562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,8,4,power_law_1.2,0.07008895874023438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,8,4,power_law_1.2,0.07303296089172363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,8,4,power_law_1.2,0.07509888172149658
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,8,4,power_law_1.2,0.0787391996383667
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,8,4,power_law_1.2,0.07803008079528809
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,8,4,power_law_1.2,0.0853542423248291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,8,4,power_law_1.2,0.0945587158203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,8,4,power_law_1.2,0.1131161594390869
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,8,4,power_law_1.2,0.1408358383178711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,8,4,power_law_1.2,0.16709632873535157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,8,4,power_law_1.2,0.05473919868469238
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,8,4,power_law_1.2,0.21241344451904295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,8,4,power_law_1.2,0.26957824707031247
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,8,4,power_law_1.2,0.06688640117645264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,8,4,power_law_1.2,0.35697792053222654
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,8,4,power_law_1.2,0.029845759868621823
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,8,4,power_law_1.2,0.03094016075134277
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,8,4,power_law_1.2,0.0324070405960083
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,8,4,power_law_1.2,0.03420543909072876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,8,4,power_law_1.2,0.6609407806396485
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,8,4,power_law_1.2,0.036266241073608395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,8,4,power_law_1.2,0.03737087965011597
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,8,4,power_law_1.2,0.8578956604003907
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,8,4,power_law_1.2,0.04792319774627686
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,8,4,power_law_1.2,0.04197375774383545
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,8,4,power_law_1.2,0.042766079902648926
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,8,4,power_law_1.2,0.04749311923980713
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,8,4,power_law_1.2,0.04718463897705078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,8,4,power_law_1.2,0.04819327831268311
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,8,4,power_law_1.2,0.050240001678466796
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,8,4,power_law_1.2,0.05278848171234131
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,8,4,power_law_1.2,0.054490880966186525
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,8,4,power_law_1.2,0.06746240139007569
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,8,4,power_law_1.2,0.47257984161376954
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,8,4,power_law_1.2,0.043738880157470704
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,8,4,power_law_1.01,0.34467838287353514
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,8,4,power_law_1.01,0.4461516952514648
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,8,4,power_law_1.2,0.06009727954864502
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,8,4,power_law_1.2,0.07745920181274414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,8,4,power_law_1.2,0.0921241569519043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,8,4,power_law_1.2,0.11794943809509277
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,8,4,power_law_1.2,0.15528575897216795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,16,2,balanced,0.02192768096923828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,16,2,balanced,0.02439039945602417
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,16,2,balanced,0.025163519382476806
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,8,4,power_law_1.2,0.19116416931152344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,16,2,balanced,0.039312639236450196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,16,2,balanced,0.05302656173706055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,16,2,balanced,0.07482624053955078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,16,2,balanced,0.07392127990722656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,16,2,balanced,0.07512703895568848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,16,2,balanced,0.07609856128692627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,16,2,balanced,0.07663871765136718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,16,2,balanced,0.07778560161590577
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,16,2,balanced,0.11448575973510741
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,16,2,balanced,0.11498496055603027
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,16,2,balanced,0.11854975700378419
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,16,2,balanced,0.12094079971313476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,16,2,balanced,0.12394880294799804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,16,2,balanced,0.13011584281921387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,16,2,balanced,0.14447615623474122
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,16,2,balanced,0.15845376014709472
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,16,2,balanced,0.18710399627685548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,16,2,balanced,0.21607423782348634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,16,2,balanced,0.2791948890686035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,16,2,balanced,0.3372313690185547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,16,2,balanced,0.49027969360351564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,16,2,balanced,0.618021125793457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,16,2,balanced,0.025207040309906004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,16,2,balanced,0.027662079334259033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,16,2,balanced,0.03100672006607056
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,16,2,balanced,0.8895897674560548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,16,2,balanced,0.03951359987258911
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,16,2,balanced,0.05751039981842041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,16,2,balanced,0.0818496036529541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,16,2,balanced,0.0812070369720459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,16,2,balanced,0.08210176467895508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,16,2,balanced,1.1572402954101562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,16,2,balanced,0.08208895683288574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,16,2,balanced,0.08298368453979492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,16,2,balanced,0.08341504096984863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,16,2,balanced,0.09675519943237305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,16,2,balanced,0.09870847702026367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,16,2,balanced,0.10033791542053223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,16,2,balanced,0.10162176132202147
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,16,2,balanced,0.10330752372741699
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,16,2,balanced,0.10745599746704101
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,16,2,balanced,0.11476096153259277
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,16,2,balanced,0.12359552383422852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,16,2,balanced,0.14081024169921874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,16,2,balanced,0.16192895889282227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,16,2,balanced,0.19856767654418944
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,16,2,balanced,0.2358208084106445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,16,2,balanced,0.35803775787353515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,16,2,balanced,0.4294118499755859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,16,2,balanced,0.6228224182128906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,16,2,balanced,0.8193497467041017
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,8,4,power_law_1.2,0.2880550384521484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,8,4,power_law_1.2,0.36560768127441406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,8,4,power_law_1.2,0.5979622268676759
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,8,4,power_law_1.2,0.818481903076172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,16,2,balanced,0.041991682052612306
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,16,2,balanced,0.04316671848297119
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,16,2,balanced,0.07778048038482666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,16,2,balanced,0.07620096206665039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,16,2,balanced,0.07694719791412354
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,16,2,balanced,0.07740543842315674
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,16,2,balanced,0.04405375957489014
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,16,2,balanced,0.04696832180023193
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,16,2,balanced,0.07876095771789551
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,16,2,balanced,0.0778214406967163
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,16,2,balanced,0.07989120006561279
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,16,2,balanced,0.08309247970581055
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,16,2,balanced,0.08233983993530274
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,16,2,balanced,0.08300288200378418
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,16,2,balanced,0.0866483211517334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,16,2,balanced,0.09326720237731934
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,16,2,balanced,0.09756928443908691
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,16,2,balanced,0.10628992080688478
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,16,2,balanced,0.11984000205993653
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,16,2,balanced,0.05604735851287842
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,16,2,balanced,0.07571839809417724
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,16,2,power_law_1.01,0.02514688014984131
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,16,2,power_law_1.01,0.03755520105361938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,16,2,power_law_1.01,0.04271872043609619
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,16,2,power_law_1.01,0.050728960037231444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,16,2,power_law_1.01,0.02268671989440918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,16,2,power_law_1.01,0.05820672035217285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,16,2,power_law_1.01,0.0673804807662964
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,16,2,power_law_1.01,0.06883711814880371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,16,2,power_law_1.01,0.0696665620803833
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,16,2,power_law_1.01,0.07335040092468262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,16,2,power_law_1.01,0.11051263809204101
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,16,2,power_law_1.01,0.11281279563903808
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,16,2,power_law_1.01,0.1165734386444092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,16,2,power_law_1.01,0.11894144058227538
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,16,2,power_law_1.01,0.12282496452331544
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,16,2,power_law_1.01,0.13016575813293457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,16,2,power_law_1.01,0.06452479839324951
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,16,2,power_law_1.01,0.1437286376953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,16,2,power_law_1.01,0.1607961654663086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,16,2,power_law_1.01,0.19183231353759767
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,16,2,power_law_1.01,0.22191232681274414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,16,2,power_law_1.01,0.29074176788330075
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,16,2,balanced,0.07648896217346192
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,16,2,balanced,0.07560448169708252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,16,2,power_law_1.01,0.3646809768676758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,16,2,power_law_1.01,0.504060173034668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,16,2,power_law_1.01,0.642081298828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,16,2,power_law_1.01,0.026888959407806397
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,16,2,power_law_1.01,0.030021119117736816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,16,2,power_law_1.01,0.03453824043273926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,16,2,power_law_1.01,0.04347008228302002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,16,2,power_law_1.01,0.05583871841430664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,16,2,power_law_1.01,1.2009037017822266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,16,2,power_law_1.01,0.062073597908020015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,16,2,power_law_1.01,0.06815616130828858
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,16,2,power_law_1.01,0.07026688098907471
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,16,2,power_law_1.01,0.07162623882293702
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,16,2,power_law_1.01,0.07420032024383545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,16,2,power_law_1.01,0.07647488117218018
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,16,2,power_law_1.01,0.08697471618652344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,16,2,power_law_1.01,0.08850687980651856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,16,2,power_law_1.01,0.9165670776367187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,16,2,power_law_1.01,0.09457663536071778
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,16,2,power_law_1.01,0.09645824432373047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,16,2,power_law_1.01,0.09807744026184081
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,16,2,power_law_1.01,0.10255871772766115
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,16,2,power_law_1.01,0.11095808029174806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,16,2,power_law_1.01,0.12558079719543458
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,16,2,power_law_1.01,0.1492031955718994
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,16,2,power_law_1.01,0.18198015213012694
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,16,2,power_law_1.01,0.22853631973266603
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,16,2,power_law_1.01,0.27704959869384765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,16,2,power_law_1.01,0.380263671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,16,2,power_law_1.01,0.4797862243652344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,16,2,power_law_1.01,0.043624958992004394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,16,2,power_law_1.01,0.6777292633056641
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,16,2,power_law_1.01,0.04482687950134277
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,16,2,power_law_1.01,0.04644480228424072
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,16,2,power_law_1.01,0.05126912117004394
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,16,2,power_law_1.01,0.05492479801177978
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,16,2,power_law_1.01,0.8786637115478516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,16,2,power_law_1.01,0.06756735801696777
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,16,2,power_law_1.01,0.06940288066864013
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,16,2,power_law_1.01,0.07055359840393066
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,16,2,power_law_1.01,0.07246719837188721
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,16,2,power_law_1.01,0.07309440135955811
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,16,2,power_law_1.01,0.07489920139312745
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,16,2,balanced,0.1991654396057129
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,16,2,balanced,0.17673343658447266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,16,2,power_law_1.01,0.07574016094207764
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,16,2,balanced,0.2797209548950195
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,16,2,power_law_1.01,0.07832575798034667
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,16,2,power_law_1.01,0.08163968086242676
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,16,2,balanced,0.35642112731933595
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,16,2,power_law_1.01,0.08592767715454101
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,16,2,power_law_1.01,0.05815807819366455
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,16,2,power_law_1.01,0.11261311531066895
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,16,2,power_law_1.01,0.1576908779144287
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,16,2,power_law_1.01,0.13436800003051758
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,16,2,power_law_1.01,0.10225664138793947
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,16,2,power_law_1.01,0.19473024368286135
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,16,2,power_law_1.01,0.26302080154418944
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,16,2,power_law_1.2,0.022699520587921143
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,16,2,power_law_1.01,0.34281726837158205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,16,2,power_law_1.2,0.03827199935913086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,16,2,power_law_1.2,0.026809599399566654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,16,2,power_law_1.2,0.049770240783691404
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,16,2,power_law_1.2,0.06489088058471679
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,16,2,power_law_1.2,0.062186241149902344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,16,2,power_law_1.2,0.04160768032073975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,16,2,power_law_1.2,0.06708992004394532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,16,2,power_law_1.2,0.07064703941345216
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,16,2,power_law_1.2,0.07312767982482911
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,16,2,power_law_1.2,0.11027839660644531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,16,2,power_law_1.2,0.11656191825866699
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,16,2,power_law_1.2,0.12030336380004883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,16,2,power_law_1.2,0.12329600334167481
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,16,2,power_law_1.2,0.1304652786254883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,16,2,power_law_1.2,0.14511232376098632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,16,2,power_law_1.2,0.16219776153564452
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,16,2,power_law_1.2,0.1921766471862793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,16,2,power_law_1.2,0.1136729621887207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,16,2,power_law_1.2,0.22237056732177735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,16,2,power_law_1.2,0.29519487380981446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,16,2,power_law_1.2,0.3742144012451172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,16,2,power_law_1.2,0.5090035247802734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,16,2,power_law_1.2,0.06941952228546142
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,16,2,power_law_1.2,0.6434815979003906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,16,2,power_law_1.2,0.026973440647125247
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,16,2,power_law_1.2,0.030195200443267824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,16,2,power_law_1.2,0.9176563262939453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,16,2,power_law_1.2,0.03295360088348388
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,16,2,power_law_1.2,0.04046592235565186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,16,2,power_law_1.2,0.05468160152435303
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,16,2,power_law_1.2,1.2174668884277344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,16,2,power_law_1.2,0.06440063953399658
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,16,2,power_law_1.2,0.0677888011932373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,16,2,power_law_1.2,0.07038976192474365
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,16,2,power_law_1.2,0.07163904190063478
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,16,2,power_law_1.2,0.07415552139282226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,16,2,power_law_1.2,0.07546495914459228
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,16,2,power_law_1.2,0.0876966381072998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,16,2,power_law_1.2,0.08966912269592285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,16,2,power_law_1.2,0.09308416366577148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,16,2,power_law_1.2,0.09560064315795899
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,16,2,power_law_1.2,0.09836928367614746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,16,2,power_law_1.2,0.10482303619384765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,16,2,power_law_1.2,0.11564543724060058
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,16,2,power_law_1.2,0.1312512016296387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,16,2,power_law_1.2,0.16074880599975586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,16,2,power_law_1.2,0.18966400146484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,16,2,power_law_1.2,0.2376972770690918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,16,2,power_law_1.2,0.28462848663330076
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,16,2,power_law_1.2,0.39129985809326173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,16,2,power_law_1.2,0.4964134216308594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,16,2,power_law_1.2,0.043170561790466314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,16,2,power_law_1.2,0.6910169219970703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,16,2,power_law_1.2,0.04472447872161865
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,16,2,power_law_1.2,0.046383361816406246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,16,2,power_law_1.2,0.8983654022216797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,16,2,power_law_1.2,0.0496563196182251
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,16,2,power_law_1.2,0.052863998413085936
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,16,2,power_law_1.2,0.05751423835754395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,16,2,power_law_1.2,0.066975998878479
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,16,2,power_law_1.2,0.06882688045501709
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,16,2,power_law_1.2,0.0701196813583374
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,16,2,power_law_1.2,0.07133183956146241
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,16,2,power_law_1.2,0.07351935863494873
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,16,2,power_law_1.2,0.07484159946441651
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,16,2,power_law_1.2,0.08010623931884767
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,16,2,power_law_1.2,0.07713151931762695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,16,2,power_law_1.2,0.08294912338256835
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,16,2,power_law_1.2,0.08684160232543944
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,16,2,power_law_1.2,0.0984166431427002
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,16,2,power_law_1.01,0.44986881256103517
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,16,2,power_law_1.01,0.7694156646728516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,16,2,power_law_1.01,0.9306816101074219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,16,2,power_law_1.2,0.11345919609069824
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,16,2,power_law_1.2,0.13273344039916993
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,16,2,power_law_1.2,0.19096191406250002
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,16,2,power_law_1.2,0.18016895294189453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,32,1,balanced,0.030846719741821287
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,32,1,balanced,0.034106879234313964
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,32,1,balanced,0.03696320056915283
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,32,1,balanced,0.046813440322875974
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,32,1,balanced,0.07047616004943848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,32,1,balanced,0.10811903953552246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,32,1,balanced,0.10222399711608887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,32,1,balanced,0.10249600410461426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,32,1,balanced,0.10299327850341797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,32,1,balanced,0.10449088096618653
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,32,1,balanced,0.10607232093811034
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,32,1,balanced,0.10854975700378418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,32,1,balanced,0.11100480079650879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,32,1,balanced,0.11590080261230469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,32,1,balanced,0.17082048416137696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,32,1,balanced,0.17403711318969725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,32,1,balanced,0.1798361587524414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,32,1,balanced,0.19251264572143553
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,32,1,balanced,0.207006721496582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,32,1,balanced,0.23243072509765622
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,32,1,balanced,0.2595699119567871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,32,1,balanced,0.3332032012939453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,32,1,balanced,0.3938790512084961
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,16,2,power_law_1.2,0.26913408279418943
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,32,1,balanced,0.54508544921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,16,2,power_law_1.2,0.3454054260253906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,32,1,balanced,0.6801670074462891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,32,1,balanced,0.03369152069091797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,32,1,balanced,0.038191359043121334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,16,2,power_law_1.2,0.5168537521362305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,32,1,balanced,0.974368667602539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,32,1,balanced,0.04308032035827637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,32,1,balanced,0.0578547191619873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,32,1,balanced,0.0806937599182129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,32,1,balanced,0.13446208000183107
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,32,1,balanced,0.13372863769531249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,32,1,balanced,0.1360383987426758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,32,1,balanced,0.13521151542663573
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,32,1,balanced,0.13680768013000488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,32,1,balanced,0.13922816276550293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,32,1,balanced,0.14034175872802734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,32,1,balanced,0.14258432388305664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,32,1,balanced,0.14597248077392577
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,32,1,balanced,0.17488063812255858
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,32,1,balanced,0.17675647735595704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,32,1,balanced,0.17990079879760743
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,32,1,balanced,0.18703231811523438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,32,1,balanced,1.2653644561767579
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,32,1,balanced,0.19630847930908205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,32,1,balanced,0.20986751556396482
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,32,1,balanced,0.22658111572265627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,32,1,balanced,0.26110591888427737
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,32,1,balanced,0.2960345649719238
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,32,1,balanced,0.46975360870361327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,32,1,balanced,0.5413894271850586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,32,1,power_law_1.01,0.030815999507904052
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,32,1,balanced,0.7858783721923828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,32,1,balanced,1.0181056213378905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,32,1,power_law_1.01,0.03285887956619263
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,32,1,power_law_1.01,0.03951103925704956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,32,1,power_law_1.01,0.06160384178161621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,32,1,power_law_1.01,0.07696512222290039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,32,1,power_law_1.01,0.0877235221862793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,32,1,power_law_1.01,0.09042176246643066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,32,1,power_law_1.01,0.09246399879455566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,32,1,power_law_1.01,0.09356415748596192
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,32,1,power_law_1.01,0.09843647956848145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,32,1,power_law_1.01,0.1041823959350586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,32,1,power_law_1.01,0.10757184028625488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,32,1,power_law_1.01,0.11570303916931153
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,32,1,power_law_1.01,0.05631807804107666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,32,1,power_law_1.01,0.17071039199829102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,32,1,power_law_1.01,0.1790220832824707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,32,1,power_law_1.01,0.2096607971191406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,32,1,power_law_1.01,0.19350719451904297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,32,1,power_law_1.01,0.24035648345947264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,32,1,power_law_1.01,0.27308479309082034
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,32,1,power_law_1.01,0.3510220718383789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,32,1,power_law_1.01,0.16718528747558595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,32,1,power_law_1.01,0.4334534454345703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,32,1,power_law_1.01,0.5755948638916015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,32,1,power_law_1.01,0.727781753540039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,32,1,power_law_1.01,0.03368639945983887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,32,1,power_law_1.01,0.039835519790649414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,32,1,power_law_1.01,1.0156038665771485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,32,1,power_law_1.01,0.04325695991516113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,32,1,power_law_1.01,0.05956992149353028
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,32,1,power_law_1.01,0.07664256095886231
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,32,1,power_law_1.01,0.097706880569458
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,32,1,power_law_1.01,1.3069267272949219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,32,1,power_law_1.01,0.1128435230255127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,32,1,power_law_1.01,0.11508543968200684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,32,1,power_law_1.01,0.11691904067993164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,32,1,power_law_1.01,0.11814847946166993
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,32,1,power_law_1.01,0.1240511989593506
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,32,1,power_law_1.01,0.12740480422973632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,32,1,power_law_1.01,0.12975232124328612
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,32,1,power_law_1.01,0.1338150405883789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,16,2,power_law_1.2,0.7306073760986328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,32,1,power_law_1.01,0.16430335998535156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,32,1,power_law_1.01,0.16462591171264648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,32,1,power_law_1.01,0.17045568466186523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,32,1,power_law_1.01,0.18150335311889648
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,16,2,power_law_1.2,1.3318707275390627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,32,1,power_law_1.01,0.19612159729003906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,32,1,power_law_1.01,0.222857608795166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,32,1,power_law_1.01,0.2586739158630371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,32,1,power_law_1.01,0.31448768615722655
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,16,2,power_law_1.2,1.5296231079101563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,32,1,power_law_1.01,0.37037120819091796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,32,1,power_law_1.2,0.03082560062408447
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,32,1,power_law_1.2,0.042686719894409184
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,32,1,power_law_1.01,0.48669185638427737
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,32,1,power_law_1.01,0.6174297714233399
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,32,1,power_law_1.2,0.05226751804351807
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,32,1,power_law_1.2,0.05970240116119385
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,32,1,power_law_1.2,0.08839679718017578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,32,1,power_law_1.01,0.8586681365966797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,32,1,power_law_1.2,0.07552832126617431
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,32,1,power_law_1.2,0.08938943862915039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,32,1,power_law_1.2,0.09163264274597169
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,32,1,power_law_1.01,1.0983225250244142
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,32,1,power_law_1.2,0.10435839653015137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,32,1,power_law_1.2,0.09971967697143555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,32,1,power_law_1.2,0.10822655677795409
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,32,1,power_law_1.2,0.11773311614990234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,32,1,power_law_1.2,0.03162496089935303
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,32,1,power_law_1.2,0.16815168380737305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,32,1,power_law_1.2,0.17148351669311523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,32,1,power_law_1.2,0.18024511337280275
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,32,1,power_law_1.2,0.1950547218322754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,32,1,power_law_1.2,0.21102720260620117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,32,1,power_law_1.2,0.2762636756896973
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,32,1,power_law_1.2,0.35200958251953124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,32,1,power_law_1.2,0.43763263702392574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,32,1,power_law_1.2,0.09561344146728515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,32,1,power_law_1.2,0.5795455932617187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,32,1,power_law_1.2,0.03983488082885742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,32,1,power_law_1.2,0.7317593383789063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,32,1,power_law_1.2,0.04349440097808838
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,32,1,power_law_1.2,0.05496767997741699
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,32,1,power_law_1.2,0.03534208059310913
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,32,1,power_law_1.2,0.07048831939697266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,32,1,power_law_1.2,0.09554047584533691
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,32,1,power_law_1.2,0.11224384307861328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,32,1,power_law_1.2,0.1145798397064209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,32,1,power_law_1.2,0.11527872085571289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,32,1,power_law_1.2,1.3113478088378907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,32,1,power_law_1.2,0.11887424468994141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,32,1,power_law_1.2,0.1233894443511963
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,32,1,power_law_1.2,0.12735360145568847
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,32,1,power_law_1.2,0.24394432067871094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,32,1,power_law_1.2,0.13569024085998535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,32,1,power_law_1.2,0.16495168685913086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,32,1,power_law_1.2,0.16845184326171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,32,1,power_law_1.2,0.17348543167114258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,32,1,power_law_1.2,1.0216896057128906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,32,1,power_law_1.2,0.1882815933227539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,32,1,power_law_1.2,0.20146240234374999
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,32,1,power_law_1.2,0.23674560546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,32,1,power_law_1.2,0.2648006439208984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,32,1,power_law_1.2,0.32338241577148436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,32,1,power_law_1.2,0.3914259338378906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,32,1,power_law_1.2,0.4933875274658203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,32,1,power_law_1.2,0.6257849502563476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,32,1,power_law_1.2,0.8658924865722657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,32,1,power_law_1.2,0.12966400146484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,32,balanced,0.025134079456329346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,32,1,power_law_1.2,1.1075769805908204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,32,balanced,0.026429440975189213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,32,balanced,0.078056960105896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,32,balanced,0.08050304412841797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,32,balanced,0.08122112274169921
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,32,balanced,0.07008512020111084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,32,balanced,0.07878528118133546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,32,balanced,0.0838425636291504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,32,balanced,0.08390015602111817
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,32,balanced,0.08428928375244141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,32,balanced,0.08873087882995605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,32,balanced,0.08604928016662597
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,32,balanced,0.0907686424255371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,32,balanced,0.09385727882385254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,32,balanced,0.08000255584716796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,32,balanced,0.10440320014953612
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,32,balanced,0.09599743843078613
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,32,balanced,0.10736512184143066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,32,balanced,0.1227289581298828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,32,balanced,0.13557375907897948
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,32,balanced,0.02721407890319824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,32,balanced,0.19514112472534179
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,32,balanced,0.2304691123962402
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,32,balanced,0.3137727928161621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,32,balanced,0.39799678802490235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,32,balanced,0.10082816123962401
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,32,balanced,0.04105472087860108
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,1,32,balanced,0.018627840280532836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,1,32,balanced,0.018670079708099366
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,1,32,balanced,0.018173439502716066
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,1,32,balanced,0.019023360013961793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,1,32,balanced,0.01904639959335327
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,1,32,balanced,0.0191103994846344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,1,32,balanced,0.019404799938201905
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,1,32,balanced,0.018845440149307252
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,1,32,balanced,0.01920639991760254
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,1,32,balanced,0.01939072012901306
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,1,32,balanced,0.019808000326156615
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,1,32,balanced,0.01984768033027649
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,1,32,balanced,0.019883519411087035
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,1,32,balanced,0.020661759376525878
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,1,32,balanced,0.02102400064468384
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,1,32,balanced,0.01900928020477295
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,1,32,balanced,0.01868927955627441
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,1,32,balanced,0.01945855975151062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,1,32,balanced,0.019429119825363158
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,1,32,balanced,0.01941248059272766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,32,power_law_1.01,0.04910208225250244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,32,power_law_1.01,0.04918399810791015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,32,power_law_1.01,0.05006080150604249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,32,power_law_1.01,0.07716608047485352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,32,power_law_1.01,0.07648896217346192
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,32,power_law_1.01,0.07875199794769287
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,32,power_law_1.01,0.07953152179718018
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,32,power_law_1.01,0.08211456298828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,32,power_law_1.01,0.08310015678405762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,32,power_law_1.01,0.08431103706359863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,32,power_law_1.01,0.08796799659729004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,32,power_law_1.01,0.09059455871582031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,32,power_law_1.01,0.0929702377319336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,32,power_law_1.01,0.09883135795593262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,32,power_law_1.01,0.1022374439239502
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,32,power_law_1.01,0.09884415626525879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,32,power_law_1.01,0.10322943687438965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,32,power_law_1.01,0.10697471618652343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,32,power_law_1.01,0.11975680351257326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,32,power_law_1.01,0.1650137519836426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,32,power_law_1.01,0.18819583892822264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,32,power_law_1.01,0.2709990310668945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,32,power_law_1.01,0.3317580795288086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,32,power_law_1.01,0.5742489624023437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,32,power_law_1.01,0.6584575653076172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,32,power_law_1.01,1.0938585662841798
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,1,32,balanced,0.02808703899383545
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,1,32,balanced,0.029319679737091063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,1,32,balanced,0.036006400585174556
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,1,32,power_law_1.01,0.019056639671325683
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,1,32,balanced,0.03822207927703857
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,1,32,power_law_1.01,0.019038720130920412
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,1,32,power_law_1.01,0.019066879749298098
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,1,32,balanced,0.01819391965866089
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,1,32,power_law_1.01,0.018885120153427123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,32,power_law_1.01,1.2187596893310546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,1,32,power_law_1.01,0.018978559970855714
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,1,32,power_law_1.01,0.018936320543289184
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,1,32,power_law_1.01,0.018981120586395263
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,1,32,balanced,0.05984640121459961
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,1,32,power_law_1.01,0.019170559644699096
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,1,32,balanced,0.05029119968414306
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,1,32,power_law_1.01,0.01977728009223938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,1,32,power_law_1.01,0.01984768033027649
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,1,32,power_law_1.01,0.01983232021331787
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,1,32,power_law_1.01,0.019711999893188475
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,1,32,power_law_1.01,0.021857280731201172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,1,32,power_law_1.01,0.019443199634552003
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,1,32,power_law_1.01,0.019503359794616697
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,1,32,power_law_1.01,0.01983232021331787
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,1,32,power_law_1.01,0.01859071969985962
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,1,32,power_law_1.01,0.0275161600112915
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,1,32,power_law_1.01,0.02462719917297363
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,1,32,power_law_1.01,0.032920320034027103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,32,power_law_1.2,0.04988287925720215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,32,power_law_1.2,0.04943615913391113
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,1,32,power_law_1.01,0.05091584205627442
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,32,power_law_1.2,0.050347518920898435
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,1,32,power_law_1.01,0.03912832021713257
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,1,32,power_law_1.01,0.056284160614013666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,32,power_law_1.2,0.06880896091461182
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,32,power_law_1.2,0.07518208026885986
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,1,32,power_law_1.01,0.08049023628234864
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,32,power_law_1.2,0.07860735893249512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,32,power_law_1.2,0.0823423957824707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,32,power_law_1.2,0.08002943992614746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,32,power_law_1.2,0.08332672119140624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,32,power_law_1.2,0.08894335746765136
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,32,power_law_1.2,0.0912268829345703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,32,power_law_1.2,0.09541119575500488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,32,power_law_1.2,0.10274944305419922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,32,power_law_1.2,0.0996735954284668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,32,power_law_1.2,0.09869440078735352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,32,power_law_1.2,0.10141311645507814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,32,power_law_1.2,0.12526080131530762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,32,power_law_1.2,0.14532352447509767
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,32,power_law_1.2,0.19854848861694335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,32,power_law_1.2,0.2491609573364258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,32,power_law_1.2,0.3453696060180664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,1,32,power_law_1.01,0.1308403205871582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,32,power_law_1.2,0.4768972778320313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,1,32,power_law_1.2,0.01906048059463501
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,1,32,power_law_1.2,0.01907711982727051
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,1,32,power_law_1.2,0.019084800481796265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,32,power_law_1.2,0.7331404876708985
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,1,32,power_law_1.2,0.019055360555648805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,1,32,power_law_1.2,0.019048960208892823
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,1,32,power_law_1.2,0.01928447961807251
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,1,32,power_law_1.2,0.019040000438690186
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,1,32,power_law_1.2,0.019402240514755247
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,1,32,power_law_1.2,0.01942271947860718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,32,power_law_1.2,1.0224691009521485
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,1,32,power_law_1.2,0.019601919651031495
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,1,32,power_law_1.2,0.019874559640884398
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,1,32,power_law_1.2,0.018967039585113525
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,1,32,power_law_1.2,0.019503359794616697
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,1,32,power_law_1.2,0.019842560291290286
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,1,32,power_law_1.2,0.019838720560073853
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,32,power_law_1.2,1.8316018676757815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,32,power_law_1.2,1.4015347290039064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,32,power_law_1.2,0.08427904129028321
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,1,32,power_law_1.01,0.14080256462097168
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,1,32,power_law_1.01,0.2034636878967285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,1,32,power_law_1.2,0.022639360427856445
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,1,32,power_law_1.2,0.027407360076904298
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,1,32,power_law_1.2,0.021516799926757812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,1,32,power_law_1.2,0.03247488021850586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,1,32,power_law_1.2,0.03829504013061523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,16,balanced,0.02226047992706299
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,16,balanced,0.021877760887146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,16,balanced,0.023532800674438477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,16,balanced,0.025703680515289307
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,16,balanced,0.03928447961807251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,16,balanced,0.06965760231018067
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,16,balanced,0.07220608234405518
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,16,balanced,0.07295743942260742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,16,balanced,0.07426176071166993
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,16,balanced,0.07386496067047119
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,16,balanced,0.07554944038391113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,16,balanced,0.07672319889068605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,16,balanced,0.0798310422897339
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,16,balanced,0.0814080047607422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,16,balanced,0.08803071975708007
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,16,balanced,0.08956416130065918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,16,balanced,0.08038911819458008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,16,balanced,0.08309120178222656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,16,balanced,0.08806015968322753
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,16,balanced,0.10158592224121095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,16,balanced,0.11256959915161133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,16,balanced,0.13643136024475097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,16,balanced,0.1546790409088135
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,16,balanced,0.23033216476440427
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,16,balanced,0.2814092826843262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,16,balanced,0.3892736053466797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,1,32,power_law_1.2,0.048047361373901365
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,16,balanced,0.5010240173339844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,1,32,power_law_1.2,0.05975168228149415
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,1,32,power_law_1.2,0.07678336143493653
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,1,32,power_law_1.2,0.11862015724182129
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,1,32,power_law_1.2,0.15512703895568847
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,1,32,power_law_1.2,0.2096883201599121
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,1,32,power_law_1.2,0.29106687545776366
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,2,16,balanced,0.018803199529647825
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,2,16,balanced,0.021100800037384033
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,2,16,balanced,0.021425919532775877
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,2,16,balanced,0.021415679454803466
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,2,16,balanced,0.021493759155273438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,2,16,balanced,0.021416959762573243
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,2,16,balanced,0.021368319988250735
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,2,16,balanced,0.018576639890670776
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,2,16,balanced,0.02149631977081299
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,2,16,balanced,0.021845760345458983
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,2,16,balanced,0.02164096117019653
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,2,16,balanced,0.021921279430389403
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,2,16,balanced,0.022191359996795654
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,2,16,balanced,0.023075840473175048
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,2,16,balanced,0.022892799377441406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,2,16,balanced,0.019308799505233766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,2,16,balanced,0.01949695944786072
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,2,16,balanced,0.019875839948654175
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,2,16,balanced,0.02113663911819458
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,2,16,balanced,0.021172480583190916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,16,power_law_1.01,0.036287999153137206
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,2,16,balanced,0.021276159286499025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,16,power_law_1.01,0.04359039783477783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,16,power_law_1.01,0.04157567977905273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,16,power_law_1.01,0.04438784122467041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,16,power_law_1.01,0.06617599964141846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,16,power_law_1.01,0.06828159809112548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,16,power_law_1.01,0.07033599853515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,16,power_law_1.01,0.07209727764129639
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,16,power_law_1.01,0.07331840038299561
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,16,power_law_1.01,0.0760256004333496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,16,power_law_1.01,0.0786137580871582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,16,power_law_1.01,0.08073856353759765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,16,power_law_1.01,0.0823846435546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,16,power_law_1.01,0.08660991668701172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,16,power_law_1.01,0.08683263778686523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,16,power_law_1.01,0.08477952003479003
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,16,power_law_1.01,0.08490752220153809
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,16,power_law_1.01,0.09487104415893555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,16,power_law_1.01,0.1042188835144043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,16,power_law_1.01,0.12874496459960938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,16,power_law_1.01,0.17315839767456054
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,16,power_law_1.01,0.21064191818237304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,16,power_law_1.01,0.24999040603637696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,16,power_law_1.01,0.4067071914672852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,16,power_law_1.01,0.4557939147949218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,16,power_law_1.01,0.7445465850830078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,2,16,power_law_1.01,0.019848959445953367
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,2,16,power_law_1.01,0.02121216058731079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,16,power_law_1.01,0.9645760345458985
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,2,16,power_law_1.01,0.02125312089920044
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,2,16,power_law_1.01,0.021199359893798828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,2,16,power_law_1.01,0.0205132794380188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,2,16,power_law_1.01,0.020546560287475583
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,2,16,power_law_1.01,0.020770559310913088
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,2,16,power_law_1.01,0.020762879848480225
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,2,16,power_law_1.01,0.020668160915374757
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,2,16,power_law_1.01,0.02061824083328247
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,2,16,balanced,0.03212543964385987
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,2,16,balanced,0.03314304113388061
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,2,16,power_law_1.01,0.020776960849761963
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,2,16,power_law_1.01,0.02128000020980835
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,2,16,power_law_1.01,0.02102400064468384
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,2,16,power_law_1.01,0.021524479389190672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,2,16,power_law_1.01,0.024033279418945314
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,2,16,power_law_1.01,0.02691328048706055
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,2,16,power_law_1.01,0.021049599647521972
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,2,16,power_law_1.01,0.02109312057495117
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,2,16,balanced,0.043921918869018556
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,2,16,balanced,0.04677504062652588
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,2,16,balanced,0.06571775913238526
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,2,16,balanced,0.07453055858612061
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,2,16,power_law_1.01,0.029687039852142334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,2,16,power_law_1.01,0.03538687944412232
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,16,power_law_1.2,0.035873279571533204
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,2,16,power_law_1.01,0.03909888029098511
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,16,power_law_1.2,0.04428287982940674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,16,power_law_1.2,0.042375679016113284
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,16,power_law_1.2,0.044766721725463865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,16,power_law_1.2,0.06667903900146485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,16,power_law_1.2,0.06903679847717285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,2,16,power_law_1.01,0.06444032192230224
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,16,power_law_1.2,0.0686515188217163
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,16,power_law_1.2,0.0728704023361206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,16,power_law_1.2,0.07469567775726318
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,16,power_law_1.2,0.07788928031921387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,16,power_law_1.2,0.07890560150146483
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,2,16,power_law_1.01,0.0828927993774414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,16,power_law_1.2,0.08510208129882812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,16,power_law_1.2,0.08451583862304687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,16,power_law_1.2,0.08468735694885254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,16,power_law_1.2,0.0864025592803955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,16,power_law_1.2,0.08488832473754883
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,2,16,power_law_1.01,0.05167744159698486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,16,power_law_1.2,0.10348799705505371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,16,power_law_1.2,0.10812800407409669
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,16,power_law_1.2,0.16507904052734373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,16,power_law_1.2,0.2016102409362793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,16,power_law_1.2,0.2689228820800781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,16,power_law_1.2,0.3820889663696289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,16,power_law_1.2,0.07244927883148193
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,16,power_law_1.2,0.5295219039916992
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,2,16,power_law_1.2,0.01978368043899536
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,2,16,power_law_1.2,0.02110208034515381
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,2,16,power_law_1.2,0.02079744100570679
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,2,16,power_law_1.2,0.021031680107116698
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,2,16,power_law_1.2,0.020636160373687745
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,16,power_law_1.2,0.6196518325805664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,2,16,power_law_1.2,0.020584959983825684
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,2,16,power_law_1.2,0.020627200603485107
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,2,16,power_law_1.2,0.020213758945465087
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,2,16,power_law_1.2,0.020660479068756104
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,2,16,power_law_1.2,0.020661759376525878
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,2,16,power_law_1.2,0.020611839294433592
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,2,16,power_law_1.2,0.020666880607604982
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,2,16,power_law_1.2,0.020936961174011233
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,2,16,power_law_1.2,0.02175487995147705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,16,power_law_1.2,0.9511590576171874
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,2,16,power_law_1.01,0.09650815963745117
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,2,16,power_law_1.01,0.14400511741638183
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,16,power_law_1.2,1.3019200134277344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,2,16,power_law_1.01,0.19988351821899414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,2,16,power_law_1.2,0.022699520587921143
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,2,16,power_law_1.2,0.02527359962463379
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,2,16,power_law_1.2,0.02369663953781128
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,2,16,power_law_1.2,0.029100799560546876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,2,16,power_law_1.2,0.03185152053833008
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,2,16,power_law_1.2,0.03869951963424682
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,4,8,balanced,0.022310400009155275
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,4,8,balanced,0.02123647928237915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,4,8,balanced,0.022700800895690917
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,4,8,balanced,0.025597438812255856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,4,8,balanced,0.036968960762023925
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,2,16,power_law_1.2,0.050451197624206544
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,4,8,balanced,0.07106048107147217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,4,8,balanced,0.07247104167938231
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,4,8,balanced,0.0734169578552246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,4,8,balanced,0.07497600078582764
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,4,8,balanced,0.07723264217376709
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,4,8,balanced,0.08470911979675293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,4,8,balanced,0.0858675193786621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,4,8,balanced,0.0754969596862793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,4,8,balanced,0.07999104022979736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,4,8,balanced,0.07773824214935303
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,4,8,balanced,0.06808703899383545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,4,8,balanced,0.08231807708740234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,4,8,balanced,0.1011622428894043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,4,8,balanced,0.12361344337463379
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,2,16,power_law_1.2,0.060963840484619135
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,4,8,balanced,0.13925760269165038
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,4,8,balanced,0.21378175735473634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,4,8,balanced,0.31834880828857426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,4,8,balanced,0.39654529571533204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,4,8,balanced,0.09084159851074218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,4,8,balanced,0.5661619186401368
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,4,8,power_law_1.01,0.02391040086746216
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,4,8,power_law_1.01,0.035594239234924316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,4,8,balanced,0.181789436340332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,4,8,balanced,0.7412467193603516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,4,8,power_law_1.01,0.037482240200042725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,4,8,power_law_1.01,0.04434815883636475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,4,8,power_law_1.01,0.06558080196380615
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,4,8,power_law_1.01,0.06889599800109864
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,4,8,power_law_1.01,0.06695295810699463
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,4,8,power_law_1.01,0.07123583793640137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,4,8,power_law_1.01,0.07455872058868408
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,4,8,power_law_1.01,0.07634047985076904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,4,8,power_law_1.01,0.08005375862121582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,4,8,power_law_1.01,0.08276224136352539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,4,8,power_law_1.01,0.07694591999053954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,4,8,power_law_1.01,0.07776768207550049
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,4,8,power_law_1.01,0.07822463989257813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,4,8,power_law_1.01,0.08336000442504883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,4,8,power_law_1.01,0.09486080169677734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,4,8,power_law_1.01,0.10560383796691894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,4,8,power_law_1.01,0.13040127754211425
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,4,8,power_law_1.01,0.14490240097045898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,4,8,power_law_1.01,0.19922815322875978
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,4,8,power_law_1.01,0.25592832565307616
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,4,8,power_law_1.01,0.36524032592773437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,4,8,power_law_1.01,0.4983347320556641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,4,8,power_law_1.01,0.040366082191467284
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,2,16,power_law_1.2,0.08097920417785645
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,2,16,power_law_1.2,0.14043007850646974
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,4,8,power_law_1.2,0.03345792055130005
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,4,8,power_law_1.2,0.040330238342285156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,4,8,power_law_1.01,0.9303743743896484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,4,8,power_law_1.2,0.046846718788146974
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,4,8,power_law_1.2,0.06666368007659912
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,4,8,power_law_1.2,0.06899072170257568
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,4,8,power_law_1.01,0.6730738830566406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,4,8,power_law_1.2,0.07106815814971924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,4,8,power_law_1.2,0.07416831970214843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,4,8,power_law_1.2,0.03610368013381958
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,4,8,power_law_1.2,0.07616896152496337
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,4,8,power_law_1.2,0.07928832054138184
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,4,8,power_law_1.2,0.08260736465454102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,4,8,power_law_1.2,0.07769343852996827
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,2,16,power_law_1.2,0.13076095581054686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,4,8,power_law_1.2,0.08053888320922852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,4,8,power_law_1.2,0.0955673599243164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,4,8,power_law_1.2,0.06725632190704346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,4,8,power_law_1.2,0.10699392318725587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,4,8,power_law_1.2,0.1369369602203369
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,4,8,power_law_1.2,0.023947520256042482
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,4,8,power_law_1.2,0.16202495574951173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,4,8,power_law_1.2,0.07530879974365234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,4,8,power_law_1.2,0.22620159149169922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,2,16,power_law_1.2,0.29873023986816405
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,2,16,power_law_1.2,0.24810111999511716
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,8,4,balanced,0.023555839061737062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,8,4,balanced,0.02486272096633911
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,8,4,balanced,0.027224318981170652
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,4,8,power_law_1.2,0.40661376953125006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,8,4,balanced,0.03462016105651856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,8,4,balanced,0.05568895816802979
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,8,4,balanced,0.10273280143737792
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,8,4,balanced,0.1042137622833252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,8,4,balanced,0.10825599670410155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,8,4,balanced,0.10891008377075195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,8,4,balanced,0.1150233554840088
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,8,4,balanced,0.11865344047546386
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,8,4,balanced,0.10616703987121583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,4,8,power_law_1.2,0.5645196914672852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,8,4,balanced,0.11032959938049317
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,8,4,balanced,0.10845952033996582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,8,4,balanced,0.1150054359436035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,8,4,balanced,0.11126527786254883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,8,4,balanced,0.12501631736755373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,8,4,balanced,0.13551103591918945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,8,4,balanced,0.18813440322875977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,8,4,balanced,0.20821760177612303
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,8,4,balanced,0.277521915435791
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,8,4,balanced,0.3160320091247558
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,8,4,balanced,0.48430335998535157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,8,4,balanced,0.5968921661376954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,4,8,power_law_1.2,0.29038463592529296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,8,4,power_law_1.01,0.025576961040496827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,8,4,balanced,0.8660479736328124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,4,8,power_law_1.2,0.08653823852539062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,8,4,power_law_1.01,0.04155903816223144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,8,4,power_law_1.01,0.04923264026641846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,8,4,power_law_1.01,0.06699903964996337
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,8,4,balanced,1.1316480255126953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,8,4,power_law_1.01,0.09475456237792969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,8,4,power_law_1.01,0.09855999946594238
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,4,8,power_law_1.2,0.7898368072509766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,4,8,power_law_1.2,1.1144627380371095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,8,4,power_law_1.01,0.10067199707031249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,8,4,power_law_1.01,0.10536191940307618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,8,4,power_law_1.01,0.11185791969299316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,8,4,power_law_1.01,0.10699904441833497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,8,4,power_law_1.01,0.10526080131530761
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,8,4,power_law_1.01,0.11023232460021973
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,8,4,power_law_1.01,0.11773056030273436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,8,4,power_law_1.01,0.1335308837890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,8,4,power_law_1.01,0.09519231796264649
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,8,4,power_law_1.01,0.15048704147338868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,8,4,power_law_1.01,0.18124544143676757
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,8,4,power_law_1.01,0.21770496368408204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,8,4,power_law_1.01,0.03224704027175903
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,8,4,power_law_1.01,0.27847167968750003
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,8,4,power_law_1.01,0.3802214431762695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,8,4,power_law_1.01,0.10377856254577637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,8,4,power_law_1.01,0.5290982437133789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,8,4,power_law_1.2,0.030090239048004154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,8,4,power_law_1.01,0.6842368316650391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,8,4,power_law_1.2,0.03870719909667969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,8,4,power_law_1.2,0.0470143985748291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,8,4,power_law_1.2,0.06833407878875733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,8,4,power_law_1.2,0.09352191925048828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,8,4,balanced,0.07286015987396241
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,8,4,power_law_1.2,0.025998079776763917
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,8,4,power_law_1.2,0.09360896110534668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,8,4,power_law_1.2,0.09806336402893066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,8,4,power_law_1.01,1.2084786987304688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,8,4,power_law_1.2,0.10768256187438965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,8,4,power_law_1.2,0.10262144088745118
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,8,4,power_law_1.2,0.10711296081542969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,8,4,power_law_1.2,0.1131942367553711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,8,4,power_law_1.2,0.12155776023864746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,8,4,power_law_1.2,0.1359718418121338
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,8,4,power_law_1.2,0.1552947235107422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,8,4,power_law_1.01,0.05732736110687255
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,8,4,power_law_1.2,0.1962892723083496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,8,4,power_law_1.2,0.22965375900268556
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,8,4,power_law_1.2,0.1113100814819336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,8,4,power_law_1.2,0.10534784317016602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,8,4,power_law_1.2,0.30688512802124024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,8,4,power_law_1.2,0.05598720073699951
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,8,4,power_law_1.2,0.4056639862060547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,16,2,balanced,0.02395263910293579
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,16,2,balanced,0.02870527982711792
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,16,2,balanced,0.03788800001144409
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,16,2,balanced,0.060938239097595215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,8,4,power_law_1.2,0.5600102233886719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,16,2,balanced,0.08076416015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,16,2,balanced,0.08237183570861817
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,16,2,balanced,0.08288895606994628
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,16,2,balanced,0.10646016120910644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,16,2,balanced,0.024791040420532227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,16,2,balanced,0.10673279762268066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,16,2,balanced,0.10951807975769043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,16,2,balanced,0.11030015945434571
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,16,2,balanced,0.11328895568847656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,16,2,balanced,0.11473152160644531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,16,2,balanced,0.1135091209411621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,16,2,balanced,0.11662976264953613
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,16,2,balanced,0.11965951919555665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,16,2,balanced,0.12982912063598634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,16,2,balanced,0.1409126377105713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,16,2,balanced,0.22602495193481445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,16,2,balanced,0.24464128494262694
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,16,2,balanced,0.31562496185302735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,8,4,power_law_1.2,0.7042585754394531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,16,2,balanced,0.3558720016479492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,16,2,balanced,0.5247257614135743
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,16,2,power_law_1.01,0.025989120006561277
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,16,2,balanced,0.6678797149658203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,16,2,power_law_1.01,0.02886784076690674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,16,2,power_law_1.01,0.03827584028244019
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,16,2,balanced,0.984035873413086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,8,4,power_law_1.01,1.0141542053222656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,16,2,power_law_1.01,0.05530623912811279
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,8,4,power_law_1.2,1.0852185821533202
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,16,2,power_law_1.01,0.07177343845367432
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,16,2,balanced,1.2695091247558594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,8,4,power_law_1.2,0.10026240348815918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,16,2,power_law_1.01,0.09955967903137207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,16,2,power_law_1.01,0.04835968017578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,16,2,power_law_1.01,0.07502336025238036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,16,2,power_law_1.01,0.11149439811706544
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,16,2,power_law_1.01,0.11671039581298828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,16,2,power_law_1.01,0.12474111557006835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,16,2,power_law_1.01,0.1405836772918701
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,16,2,power_law_1.01,0.09667840003967285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,16,2,power_law_1.01,0.10206208229064942
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,16,2,power_law_1.01,0.15884415626525877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,8,4,power_law_1.2,1.3962905883789063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,16,2,power_law_1.01,0.10607104301452637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,16,2,power_law_1.01,0.2216307258605957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,16,2,power_law_1.01,0.2527769660949707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,16,2,power_law_1.01,0.32072704315185546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,16,2,power_law_1.01,0.4140300750732422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,16,2,power_law_1.01,0.576866569519043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,16,2,power_law_1.01,0.7188390350341797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,16,2,power_law_1.01,0.11028736114501954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,16,2,power_law_1.2,0.02845567941665649
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,16,2,power_law_1.2,0.03865983963012695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,16,2,power_law_1.2,0.054113278388977046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,16,2,power_law_1.01,1.0513279724121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,16,2,power_law_1.2,0.07130496025085449
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,16,2,power_law_1.2,0.0730073595046997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,16,2,power_law_1.2,0.07437183856964111
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,16,2,power_law_1.2,0.09955455780029296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,16,2,power_law_1.2,0.09899519920349122
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,16,2,power_law_1.2,0.10449536323547363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,16,2,power_law_1.01,1.4422770690917968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,16,2,power_law_1.2,0.1049407958984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,16,2,power_law_1.2,0.10644096374511718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,16,2,power_law_1.2,0.11460864067077636
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,16,2,power_law_1.2,0.11465984344482423
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,16,2,power_law_1.2,0.11954688072204589
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,16,2,power_law_1.2,0.02390912055969238
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,16,2,power_law_1.2,0.12796031951904296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,16,2,power_law_1.2,0.143754243850708
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,16,2,power_law_1.2,0.16236799240112304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,16,2,power_law_1.2,0.22869760513305665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,16,2,power_law_1.2,0.2583859252929687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,16,2,power_law_1.2,0.33039871215820316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,16,2,power_law_1.2,0.42333183288574217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,32,1,balanced,0.030005760192871093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,32,1,balanced,0.03524607896804809
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,32,1,balanced,0.04203904151916504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,16,2,power_law_1.2,0.04606592178344727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,16,2,power_law_1.2,0.5938713455200195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,32,1,balanced,0.05416512012481689
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,32,1,balanced,0.0922867202758789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,32,1,balanced,0.11807295799255371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,32,1,balanced,0.11837504386901856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,32,1,balanced,0.12026752471923827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,32,1,balanced,0.1206220817565918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,32,1,balanced,0.12065216064453126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,32,1,balanced,0.12435968399047852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,32,1,balanced,0.14217280387878417
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,32,1,balanced,0.1435641574859619
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,32,1,balanced,0.14750720024108888
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,16,2,power_law_1.01,0.10330368041992188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,32,1,balanced,0.14602047920227051
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,32,1,balanced,0.14813952445983886
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,32,1,balanced,0.15412927627563477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,32,1,balanced,0.16335424423217773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,32,1,balanced,0.173571834564209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,32,1,balanced,0.2661638450622559
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,32,1,balanced,0.28457664489746093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,32,1,balanced,0.36301761627197265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,32,1,balanced,0.3998828887939453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,16,2,power_law_1.2,0.7281100463867187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,32,1,balanced,0.5849951934814454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,16,2,power_law_1.01,0.0727833604812622
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,32,1,balanced,0.7317337799072265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,32,1,power_law_1.01,0.029166719913482665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,32,1,power_law_1.01,0.04559103965759277
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,32,1,power_law_1.01,0.060241918563842776
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,32,1,balanced,1.0722271728515627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,32,1,power_law_1.01,0.07381824016571045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,32,1,power_law_1.01,0.09527104377746583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,32,1,power_law_1.01,0.10682880401611328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,32,1,power_law_1.01,0.1094976043701172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,16,2,power_law_1.2,1.065279998779297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,32,1,power_law_1.01,0.11417984008789062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,32,1,power_law_1.01,0.1213804817199707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,32,1,power_law_1.01,0.135098876953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,32,1,power_law_1.01,0.13842944145202635
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,32,1,power_law_1.01,0.14348799705505372
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,32,1,power_law_1.01,0.03528575897216797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,32,1,power_law_1.01,0.14393280029296876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,32,1,power_law_1.01,0.1490892791748047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,32,1,power_law_1.01,0.15743488311767578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,32,1,power_law_1.01,0.1016057586669922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,32,1,balanced,1.3916026306152343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,32,1,power_law_1.01,0.255927677154541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,32,1,power_law_1.01,0.29235071182250977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,32,1,power_law_1.01,0.35937599182128904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,32,1,power_law_1.01,0.4559884643554687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,32,1,power_law_1.2,0.0287609601020813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,32,1,power_law_1.01,0.595294075012207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,32,1,power_law_1.01,0.19356800079345704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,32,1,power_law_1.01,0.7606380462646485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,32,1,power_law_1.2,0.03369663953781128
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,32,1,power_law_1.2,0.046210560798645016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,32,1,power_law_1.2,0.05894464015960693
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,32,1,power_law_1.2,0.07118847846984863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,32,1,power_law_1.2,0.09512767791748047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,32,1,power_law_1.2,0.10239616394042969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,32,1,power_law_1.01,1.0945184326171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,32,1,power_law_1.2,0.11209152221679688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,32,1,power_law_1.2,0.11610560417175293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,32,1,power_law_1.2,0.12335616111755372
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,32,1,power_law_1.2,0.13818943977355957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,16,2,power_law_1.2,1.4655757141113281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,32,1,power_law_1.2,0.14033984184265136
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,32,1,power_law_1.01,0.17641151428222657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,32,1,power_law_1.01,1.4352928161621095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,32,1,power_law_1.2,0.14647359848022462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,32,1,power_law_1.2,0.15180159568786622
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,32,1,power_law_1.2,0.1601433563232422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,32,1,power_law_1.2,0.1781977653503418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,32,1,power_law_1.2,0.19517120361328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,32,1,power_law_1.2,0.2597420883178711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,32,1,power_law_1.2,0.293573112487793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,32,1,power_law_1.2,0.36104000091552735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,32,1,power_law_1.2,0.10705471992492677
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,32,balanced,0.037285120487213136
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,32,balanced,0.07057407855987549
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,32,1,power_law_1.2,0.45442111968994137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,32,balanced,0.09161855697631835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,32,1,power_law_1.2,0.1462335968017578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,32,balanced,0.16527488708496094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,32,balanced,0.2389913558959961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,32,balanced,0.25673856735229494
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,32,balanced,0.3069568061828613
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,32,balanced,0.2690444755554199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,32,balanced,0.2982579231262207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,32,balanced,0.27947647094726563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,32,1,power_law_1.2,0.5973408126831055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,32,balanced,0.3211110305786133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,32,balanced,0.32960769653320315
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,32,balanced,0.33945598602294924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,32,balanced,0.3640614318847656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,32,balanced,0.3768627166748047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,32,balanced,0.3767027282714844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,32,balanced,0.40755966186523435
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,32,balanced,0.47159553527832027
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,32,balanced,0.5339635086059571
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,32,1,power_law_1.2,0.76299072265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,32,balanced,0.7178521728515624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,32,balanced,0.833544921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,32,balanced,1.1527283477783203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,32,balanced,0.035357439517974855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,32,balanced,1.4662950134277344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,32,balanced,0.04520063877105713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,32,balanced,0.05803264141082763
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,32,balanced,0.09776384353637695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,32,balanced,0.1294502353668213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,32,balanced,0.14027135848999023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,32,balanced,2.272447967529297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,32,balanced,0.1455244827270508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,32,balanced,0.15801471710205078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,32,balanced,0.14696703910827638
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,32,balanced,0.1571225643157959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,32,balanced,0.1629631996154785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,32,balanced,0.16761087417602538
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,32,balanced,0.18902528762817383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,32,balanced,0.207956485748291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,32,1,power_law_1.2,1.0992415618896483
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,32,balanced,0.19237632751464845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,32,balanced,0.2042470359802246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,32,balanced,0.245294075012207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,32,balanced,0.1484006404876709
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,32,balanced,0.29021568298339845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,32,balanced,0.35682304382324215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,32,balanced,0.43142017364501956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,32,balanced,4.43820556640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,32,balanced,0.6102207946777344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,32,balanced,0.7577471923828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,32,balanced,1.0846131134033203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,32,balanced,5.813201904296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,32,balanced,1.3683494567871093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,32,balanced,2.032801208496094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,32,balanced,2.9451544189453123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,32,balanced,2.6483889770507814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,32,1,power_law_1.2,1.4372575378417969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,1,32,balanced,0.02152575969696045
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,1,32,balanced,0.024395520687103274
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,1,32,balanced,0.07436416149139405
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,1,32,balanced,0.07345664024353027
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,1,32,balanced,0.07437439918518066
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,1,32,balanced,0.07336319923400879
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,1,32,balanced,0.059171838760375975
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,1,32,balanced,0.07502975940704346
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,1,32,balanced,0.07422592163085937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,1,32,balanced,0.07483007907867431
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,1,32,balanced,0.07509119987487793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,1,32,balanced,0.0747430419921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,1,32,balanced,0.07351808071136474
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,1,32,balanced,0.07814527988433838
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,1,32,balanced,0.07926911830902099
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,1,32,balanced,0.0824396800994873
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,1,32,balanced,0.03583103895187378
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,1,32,balanced,0.07311744213104247
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,1,32,balanced,0.07424767971038818
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,1,32,balanced,0.0729203176498413
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,1,32,balanced,0.07561728000640869
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,32,power_law_1.01,0.23462911605834963
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,32,power_law_1.01,0.23118719100952148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,32,power_law_1.01,0.2323148727416992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,32,power_law_1.01,0.2325388717651367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,32,power_law_1.01,0.24180736541748046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,32,power_law_1.01,0.26165760040283204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,32,power_law_1.01,0.27566591262817386
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,32,power_law_1.01,0.2893337631225586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,32,power_law_1.01,0.2764006423950195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,32,power_law_1.01,0.28164991378784177
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,32,power_law_1.01,0.3038041687011719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,32,power_law_1.01,0.31939456939697264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,32,power_law_1.01,0.33262081146240235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,32,power_law_1.01,0.3584409713745117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,32,power_law_1.01,0.3693580627441406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,32,power_law_1.01,0.374334716796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,32,power_law_1.01,0.43261825561523437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,32,power_law_1.01,0.540560646057129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,32,power_law_1.01,0.6043097686767578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,32,power_law_1.01,0.8259980773925781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,32,power_law_1.01,1.0937471771240235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,32,power_law_1.01,1.5067507934570312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,32,power_law_1.01,1.7743270874023438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,32,power_law_1.01,2.8847015380859373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,32,power_law_1.01,3.7024639892578124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,32,power_law_1.01,6.266257934570312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,32,power_law_1.01,0.1197606372833252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,32,power_law_1.01,0.1196940803527832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,32,power_law_1.01,0.12237695693969726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,32,power_law_1.01,0.12392704010009765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,32,power_law_1.01,0.11631232261657716
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,32,power_law_1.01,0.12017151832580567
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,32,power_law_1.01,7.831528930664062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,32,power_law_1.01,0.1287769603729248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,32,power_law_1.01,0.14805503845214846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,32,power_law_1.01,0.13426560401916504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,32,power_law_1.01,0.13575039863586427
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,32,power_law_1.01,0.1496729564666748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,32,power_law_1.01,0.15509247779846191
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,32,power_law_1.01,0.1761612892150879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,32,power_law_1.01,0.1836479949951172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,32,power_law_1.01,0.19540607452392578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,32,power_law_1.01,0.1941542434692383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,32,power_law_1.01,0.22244096755981446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,32,power_law_1.01,0.2696921539306641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,32,power_law_1.01,0.32071552276611326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,32,power_law_1.01,0.41366142272949225
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,1,32,balanced,0.10724479675292968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,1,32,balanced,0.15163647651672363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,32,power_law_1.01,0.5430156707763671
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,32,power_law_1.01,0.6850994873046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,1,32,balanced,0.1593344020843506
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,1,32,balanced,0.24091775894165038
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,32,power_law_1.01,0.9401741027832031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,1,32,power_law_1.01,0.07294464111328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,1,32,balanced,0.2907865524291992
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,1,32,balanced,0.11226240158081055
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,1,32,power_law_1.01,0.07348095893859863
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,1,32,power_law_1.01,0.07372032165527344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,32,power_law_1.01,1.368423614501953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,1,32,power_law_1.01,0.0702566385269165
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,1,32,power_law_1.01,0.06596992015838624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,1,32,power_law_1.01,0.06596479892730714
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,1,32,power_law_1.01,0.06646016120910644
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,1,32,power_law_1.01,0.0675225591659546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,1,32,power_law_1.01,0.0673804807662964
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,1,32,power_law_1.01,0.06860671997070313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,32,power_law_1.01,2.005909729003906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,1,32,power_law_1.01,0.06877056121826172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,1,32,power_law_1.01,0.07264639854431151
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,1,32,power_law_1.01,0.07109888076782227
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,1,32,power_law_1.01,0.07220736026763916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,1,32,power_law_1.01,0.07414527893066405
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,1,32,power_law_1.01,0.07445504188537597
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,1,32,power_law_1.01,0.12981632232666016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,32,power_law_1.01,2.5122137451171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,32,power_law_1.01,3.7422027587890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,1,32,power_law_1.01,0.07662335872650147
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,1,32,power_law_1.01,0.0837337589263916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,1,32,power_law_1.01,0.11233280181884767
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,1,32,power_law_1.01,0.0913907241821289
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,1,32,power_law_1.01,0.1765939140319824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,32,power_law_1.2,0.23049856185913087
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,32,power_law_1.2,0.23147008895874022
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,1,32,power_law_1.01,0.20902143478393556
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,32,power_law_1.2,0.2295065689086914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,32,power_law_1.2,0.2363520050048828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,32,power_law_1.2,0.2437068748474121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,32,power_law_1.2,0.2685312080383301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,32,power_law_1.2,0.29159807205200194
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,32,power_law_1.2,0.2845363235473633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,1,32,power_law_1.01,0.2926566314697266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,32,power_law_1.2,0.28124160766601564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,32,power_law_1.2,0.30274560928344724
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,32,power_law_1.2,0.3163929557800293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,1,32,power_law_1.01,0.4003456115722656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,32,power_law_1.2,0.34845439910888676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,32,power_law_1.2,0.3687104034423828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,32,power_law_1.2,0.3811481475830078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,32,power_law_1.2,0.39179904937744137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,32,power_law_1.2,0.4338483047485352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,32,power_law_1.2,0.5535744094848634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,32,power_law_1.2,0.7057843017578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,32,power_law_1.2,0.9169907379150392
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,32,power_law_1.2,1.1661939239501953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,32,power_law_1.2,0.23016319274902344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,32,power_law_1.2,1.7442842102050782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,32,power_law_1.2,0.12270591735839845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,32,power_law_1.2,2.3305433654785155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,32,power_law_1.2,0.12035455703735351
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,32,power_law_1.2,0.12256256103515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,32,power_law_1.2,0.12500351905822754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,32,power_law_1.2,0.11451264381408692
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,32,power_law_1.2,0.11630592346191407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,32,power_law_1.2,0.13245823860168457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,32,power_law_1.2,3.718858337402344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,32,power_law_1.2,0.13993215560913086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,32,power_law_1.2,0.13167231559753417
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,32,power_law_1.2,0.15099264144897462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,32,power_law_1.2,0.15927424430847167
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,32,power_law_1.2,0.1415564823150635
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,32,power_law_1.2,0.17321983337402344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,32,power_law_1.2,0.19189119338989258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,32,power_law_1.2,0.20254592895507811
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,32,power_law_1.2,0.20655487060546873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,32,power_law_1.2,0.22821760177612305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,32,power_law_1.2,0.2893427276611328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,32,power_law_1.2,4.893222351074219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,32,power_law_1.2,0.3675008010864258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,32,power_law_1.2,0.43158401489257814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,32,power_law_1.2,0.6076403045654297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,32,power_law_1.2,0.8270925140380859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,32,power_law_1.2,1.0557606506347657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,32,power_law_1.2,1.5356198120117188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,1,32,power_law_1.2,0.0738316822052002
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,1,32,power_law_1.2,0.0736243200302124
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,1,32,power_law_1.2,0.07387775897979736
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,1,32,power_law_1.2,0.07363711833953858
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,32,power_law_1.2,2.06308349609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,1,32,power_law_1.2,0.06571648120880128
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,32,power_law_1.2,7.75433349609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,32,power_law_1.2,2.994330749511719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,1,32,power_law_1.2,0.06462207794189453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,1,32,power_law_1.2,0.06146815776824951
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,1,32,power_law_1.2,0.06976255893707276
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,1,32,power_law_1.2,0.06865280151367188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,1,32,power_law_1.2,0.0650163221359253
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,1,32,power_law_1.2,0.06819071769714355
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,1,32,power_law_1.2,0.07022079944610596
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,1,32,power_law_1.2,0.07127423763275147
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,1,32,power_law_1.2,0.07640448093414307
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,1,32,power_law_1.2,0.07737855911254883
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,1,32,power_law_1.2,0.08418944358825683
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,1,32,power_law_1.2,0.09735936164855957
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,1,32,power_law_1.2,0.10036607742309571
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,1,32,power_law_1.01,0.536998405456543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,1,32,power_law_1.01,0.8139814758300782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,32,power_law_1.2,4.349550170898437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,32,power_law_1.2,10.95843505859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,1,32,power_law_1.2,0.07698431968688965
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,1,32,power_law_1.2,0.1361728000640869
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,16,balanced,0.03198463916778564
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,1,32,power_law_1.2,0.1583078384399414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,1,32,power_law_1.2,0.22060031890869142
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,16,balanced,0.05816448211669922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,16,balanced,0.07149184226989747
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,16,balanced,0.12360192298889161
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,16,balanced,0.1989094352722168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,16,balanced,0.23962368011474608
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,16,balanced,0.2624665641784668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,16,balanced,0.271724796295166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,16,balanced,0.28616575241088865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,16,balanced,0.2846195220947266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,16,balanced,0.30019968032836913
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,16,balanced,0.31897855758666993
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,16,balanced,0.3283404922485352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,16,balanced,0.3584345626831055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,16,balanced,0.3427033615112305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,16,balanced,0.3659737777709961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,16,balanced,0.40270336151123043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,16,balanced,0.4637337493896484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,16,balanced,0.5353817749023437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,16,balanced,0.7357657623291016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,16,balanced,0.8660569763183593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,16,balanced,1.224559326171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,16,balanced,1.4915533447265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,16,balanced,2.3112332153320314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,16,balanced,3.0506085205078124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,16,balanced,0.03260672092437744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,16,balanced,0.037864959239959715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,16,balanced,0.05372032165527344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,16,balanced,0.08343040466308593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,16,balanced,0.11080960273742675
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,1,32,power_law_1.2,0.27196928024291994
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,16,balanced,0.14299776077270507
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,16,balanced,0.15173248291015623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,16,balanced,0.15124608039855958
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,16,balanced,4.627176818847657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,1,32,power_law_1.2,0.4020467376708984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,16,balanced,0.160817928314209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,16,balanced,0.16198272705078126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,16,balanced,0.1665113639831543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,16,balanced,0.18159360885620118
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,16,balanced,0.19717887878417967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,16,balanced,0.2119820785522461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,16,balanced,0.1877427291870117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,16,balanced,0.21092479705810546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,16,balanced,0.22434944152832031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,16,balanced,0.2732019233703613
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,16,balanced,0.31936128616333004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,16,balanced,0.3915929412841797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,16,balanced,0.47963905334472656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,16,balanced,0.6916006469726563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,1,32,power_law_1.2,0.5449868774414062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,16,balanced,0.8504345703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,16,balanced,1.2468736267089844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,16,balanced,6.128114013671874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,16,balanced,1.5760665893554688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,16,balanced,2.344170227050781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,16,balanced,3.067532653808594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,1,32,power_law_1.2,0.8717273712158203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,1,32,power_law_1.2,1.1710118103027345
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,2,16,balanced,0.024159998893737794
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,2,16,balanced,0.07631103992462158
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,2,16,balanced,0.07688064098358154
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,2,16,balanced,0.03304960012435913
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,2,16,balanced,0.07679743766784668
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,2,16,balanced,0.07753215789794922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,2,16,balanced,0.07746431827545167
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,2,16,balanced,0.02470911979675293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,2,16,balanced,0.07832575798034667
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,2,16,balanced,0.07768447875976563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,2,16,balanced,0.0517580795288086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,2,16,balanced,0.07828735828399658
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,2,16,balanced,0.07981311798095703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,2,16,balanced,0.08081536293029785
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,2,16,balanced,0.08438655853271484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,2,16,balanced,0.0873100757598877
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,2,16,balanced,0.09221504211425782
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,2,16,balanced,0.06704127788543701
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,2,16,balanced,0.0775487995147705
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,2,16,balanced,0.07665152072906493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,16,power_law_1.01,0.1617932891845703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,16,power_law_1.01,0.22177536010742188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,16,power_law_1.01,0.2186188888549805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,16,power_law_1.01,0.22336511611938475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,16,power_law_1.01,0.22858112335205077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,16,power_law_1.01,0.22259071350097653
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,2,16,balanced,0.07657472133636475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,16,power_law_1.01,0.26502399444580077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,16,power_law_1.01,0.26162431716918944
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,16,power_law_1.01,0.25764736175537106
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,16,power_law_1.01,0.26956287384033206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,16,power_law_1.01,0.2803353691101074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,16,power_law_1.01,0.29609600067138675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,16,power_law_1.01,0.30943103790283205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,16,power_law_1.01,0.33254398345947267
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,16,power_law_1.01,0.32942081451416017
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,16,power_law_1.01,0.3512345504760742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,16,power_law_1.01,0.38197502136230466
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,16,power_law_1.01,0.46354049682617193
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,16,power_law_1.01,0.5474393463134766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,16,power_law_1.01,0.7181017303466797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,16,power_law_1.01,0.8906086730957032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,16,power_law_1.01,1.3299685668945311
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,16,power_law_1.01,1.6425509643554688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,16,power_law_1.01,2.470028839111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,2,16,balanced,0.17927167892456056
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,2,16,balanced,0.13246335983276367
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,2,16,balanced,0.1241983985900879
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,2,16,balanced,0.1959129524230957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,16,power_law_1.01,0.09420031547546387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,16,power_law_1.01,3.2306417846679687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,2,16,balanced,0.2962700843811035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,16,power_law_1.01,0.1259494400024414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,16,power_law_1.01,0.12599552154541016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,16,power_law_1.01,0.12926207542419435
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,16,power_law_1.01,0.13363455772399901
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,16,power_law_1.01,0.139421443939209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,16,power_law_1.01,0.1388582420349121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,16,power_law_1.01,0.15179519653320311
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,16,power_law_1.01,0.15848832130432128
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,2,16,balanced,0.3608678436279297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,16,power_law_1.01,0.17845632553100585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,16,power_law_1.01,0.20541439056396485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,16,power_law_1.01,0.1933772850036621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,16,power_law_1.01,0.19261568069458007
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,16,power_law_1.01,0.1419148826599121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,16,power_law_1.01,0.20793216705322265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,16,power_law_1.01,0.2665459251403809
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,16,power_law_1.01,0.29634815216064453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,16,power_law_1.01,0.39271423339843753
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,16,power_law_1.01,0.5001331329345703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,16,power_law_1.01,0.6708672332763672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,16,power_law_1.01,4.82342529296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,16,power_law_1.01,0.8693965148925781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,2,16,power_law_1.01,0.061066241264343256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,16,power_law_1.01,1.2396902465820312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,2,16,power_law_1.01,0.07608320236206054
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,2,16,power_law_1.01,0.07375743865966797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,16,power_law_1.01,1.6750604248046876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,2,16,power_law_1.01,0.07650432109832764
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,2,16,power_law_1.01,0.07524223804473877
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,2,16,power_law_1.01,0.0722547197341919
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,2,16,power_law_1.01,0.0760319995880127
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,2,16,power_law_1.01,0.0756492805480957
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,2,16,power_law_1.01,0.07477759838104248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,16,power_law_1.01,2.3969215393066405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,16,power_law_1.01,6.2580224609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,16,power_law_1.01,0.11950847625732422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,2,16,power_law_1.01,0.07441919803619385
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,2,16,power_law_1.01,0.07422336101531982
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,2,16,power_law_1.01,0.07635456085205078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,2,16,power_law_1.01,0.07653888225555419
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,2,16,power_law_1.01,0.07837056159973145
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,2,16,power_law_1.01,0.10614144325256347
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,2,16,power_law_1.01,0.07472127914428711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,16,power_law_1.01,3.2581771850585937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,16,power_law_1.01,0.15129599571228028
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,2,16,power_law_1.01,0.08240127563476562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,2,16,power_law_1.01,0.08553088188171387
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,2,16,power_law_1.01,0.0953164768218994
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,2,16,power_law_1.01,0.12164480209350585
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,2,16,power_law_1.01,0.14625791549682615
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,2,16,power_law_1.01,0.19347328186035156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,16,power_law_1.2,0.20924671173095702
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,2,16,power_law_1.01,0.2491916847229004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,16,power_law_1.2,0.21020671844482425
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,2,16,power_law_1.01,0.33580287933349606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,16,power_law_1.2,0.223110408782959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,16,power_law_1.2,0.19865983963012696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,16,power_law_1.2,0.22902271270751956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,16,power_law_1.2,0.25161088943481447
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,16,power_law_1.2,0.2544384002685547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,16,power_law_1.2,0.255359992980957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,16,power_law_1.2,0.2711052894592285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,16,power_law_1.2,0.2867942428588867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,16,power_law_1.2,0.3005171203613281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,16,power_law_1.2,0.31216127395629883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,16,power_law_1.2,0.33789825439453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,16,power_law_1.2,0.3297945785522461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,2,16,power_law_1.01,0.43758590698242184
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,16,power_law_1.2,0.16551040649414062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,16,power_law_1.2,0.38569087982177735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,16,power_law_1.2,0.4994841766357422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,16,power_law_1.2,0.5852915191650391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,16,power_law_1.2,0.7823436737060547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,16,power_law_1.2,0.35107070922851563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,16,power_law_1.2,1.4258930969238282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,16,power_law_1.2,1.8142489624023437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,16,power_law_1.2,0.09445631980895997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,16,power_law_1.2,2.599945068359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,16,power_law_1.2,0.11513855934143066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,16,power_law_1.2,0.1255180835723877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,16,power_law_1.2,0.12930560111999512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,16,power_law_1.2,0.12427007675170898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,16,power_law_1.2,0.12060416221618653
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,16,power_law_1.2,3.5222283935546876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,16,power_law_1.2,0.14017024040222167
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,16,power_law_1.2,0.13987071990966798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,16,power_law_1.2,0.1415513610839844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,16,power_law_1.2,0.1467123222351074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,16,power_law_1.2,0.15614336013793945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,16,power_law_1.2,0.16605184555053712
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,16,power_law_1.2,0.18102655410766602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,16,power_law_1.2,0.9978316497802734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,16,power_law_1.2,0.20840063095092773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,16,power_law_1.2,0.1787763214111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,16,power_law_1.2,0.1919398307800293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,16,power_law_1.2,0.21926528930664063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,16,power_law_1.2,0.2715660858154297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,16,power_law_1.2,0.3277798461914062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,16,power_law_1.2,5.546171875000001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,16,power_law_1.2,0.4292390441894531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,16,power_law_1.2,0.5177727890014648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,16,power_law_1.2,0.7216281890869141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,16,power_law_1.2,0.9246323394775391
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,2,16,power_law_1.2,0.06194560050964355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,16,power_law_1.2,1.3531263732910157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,2,16,power_law_1.2,0.07156352043151856
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,2,16,power_law_1.2,0.07489920139312745
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,16,power_law_1.2,1.6848204040527341
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,2,16,power_law_1.2,0.07620992183685302
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,2,16,power_law_1.2,0.07675648212432862
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,2,16,power_law_1.2,0.0674393606185913
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,2,16,power_law_1.2,0.07475456237792968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,2,16,power_law_1.2,0.0775871992111206
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,2,16,power_law_1.2,0.0731763219833374
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,2,16,power_law_1.2,0.07065599918365478
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,2,16,power_law_1.2,0.07299327850341797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,2,16,power_law_1.2,0.07425024032592772
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,2,16,power_law_1.2,0.07564671993255614
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,2,16,power_law_1.2,0.0787225580215454
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,2,16,power_law_1.2,0.08031999588012695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,16,power_law_1.2,2.4506675720214846
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,2,16,power_law_1.2,0.08814592361450195
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,2,16,power_law_1.2,0.11682304382324218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,16,power_law_1.2,3.364559326171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,16,power_law_1.2,7.154539794921876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,2,16,power_law_1.01,0.6420134735107422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,2,16,power_law_1.01,0.8611328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,2,16,power_law_1.2,0.08463744163513184
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,2,16,power_law_1.2,0.10406656265258789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,2,16,power_law_1.2,0.14548607826232912
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,4,8,balanced,0.02781183958053589
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,2,16,power_law_1.2,0.17733375549316405
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,2,16,power_law_1.2,0.24885631561279298
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,4,8,balanced,0.09819264411926269
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,4,8,balanced,0.17252607345581056
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,4,8,balanced,0.05143936157226563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,4,8,balanced,0.22789247512817382
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,4,8,balanced,0.2393190383911133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,4,8,balanced,0.241713924407959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,4,8,balanced,0.2710848045349121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,4,8,balanced,0.06552832126617432
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,4,8,balanced,0.27543296813964846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,4,8,balanced,0.2821926307678223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,4,8,balanced,0.2969779205322266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,4,8,balanced,0.3066252708435059
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,4,8,balanced,0.3301491165161133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,4,8,balanced,0.31540096282958985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,4,8,balanced,0.33115520477294924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,4,8,balanced,0.35947902679443355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,4,8,balanced,0.4209088134765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,2,16,power_law_1.2,0.31460607528686524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,4,8,balanced,0.47787521362304686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,4,8,balanced,0.6617971038818359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,4,8,balanced,0.7700723266601562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,4,8,balanced,1.0910028839111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,4,8,balanced,1.3179685974121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,4,8,balanced,1.9669273376464844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,4,8,balanced,0.030504961013793946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,4,8,balanced,0.04072703838348389
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,4,8,balanced,0.0526361608505249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,4,8,balanced,2.527158966064453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,4,8,balanced,0.08437631607055665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,4,8,balanced,0.11011839866638184
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,4,8,balanced,0.13974911689758301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,4,8,balanced,0.15050496101379393
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,4,8,balanced,0.1536755180358887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,4,8,balanced,0.17471616744995117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,4,8,balanced,0.1768832015991211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,4,8,balanced,0.18095232009887696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,4,8,balanced,0.19550079345703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,4,8,balanced,0.2050227165222168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,4,8,balanced,3.7677722167968755
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,4,8,balanced,0.2169536018371582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,4,8,balanced,0.19675392150878906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,4,8,balanced,0.21045631408691406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,4,8,balanced,0.22676479339599612
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,4,8,balanced,0.2651238441467285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,4,8,balanced,0.30099967956542967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,4,8,balanced,0.3768678283691406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,4,8,balanced,0.4417817687988281
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,2,16,power_law_1.2,0.47269119262695314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,4,8,balanced,0.6625151824951172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,4,8,balanced,5.050059509277344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,4,8,balanced,0.8055897521972657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,2,16,power_law_1.2,0.5663232040405274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,4,8,balanced,1.1568857574462892
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,4,8,balanced,1.4321945190429688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,4,8,balanced,2.1462937927246095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,4,8,balanced,2.7926080322265627
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,2,16,power_law_1.2,1.0433920288085938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,4,8,balanced,0.028844799995422367
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,4,8,balanced,0.09164159774780274
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,4,8,balanced,0.031143679618835452
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,4,8,balanced,0.09159040451049805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,4,8,balanced,0.03547391891479492
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,4,8,balanced,0.09095295906066894
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,4,8,balanced,0.09214591979980469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,4,8,balanced,0.05212927818298339
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,4,8,balanced,0.09196288108825683
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,4,8,balanced,0.0919705581665039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,4,8,balanced,0.09304960250854492
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,4,8,balanced,0.0940236759185791
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,4,8,balanced,0.0953587245941162
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,4,8,balanced,0.095665922164917
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,4,8,balanced,0.09943296432495116
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,4,8,balanced,0.10139904022216797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,4,8,balanced,0.10911744117736817
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,4,8,balanced,0.11205632209777831
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,4,8,balanced,0.07500671863555908
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,2,16,power_law_1.2,1.4327372741699218
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,4,8,balanced,0.08995072364807129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,4,8,power_law_1.01,0.08706175804138183
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,4,8,power_law_1.01,0.1586291217803955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,4,8,power_law_1.01,0.1346009635925293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,4,8,power_law_1.01,0.17830400466918944
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,4,8,power_law_1.01,0.20280832290649414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,4,8,power_law_1.01,0.22714239120483398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,4,8,power_law_1.01,0.22289152145385743
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,4,8,power_law_1.01,0.23207935333251953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,4,8,power_law_1.01,0.25635839462280274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,4,8,power_law_1.01,0.26114816665649415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,4,8,power_law_1.01,0.26962432861328123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,4,8,power_law_1.01,0.2865100860595703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,4,8,power_law_1.01,0.29704063415527343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,4,8,power_law_1.01,0.32220672607421874
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,4,8,balanced,0.09052800178527833
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,4,8,power_law_1.01,0.3074777603149414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,4,8,power_law_1.01,0.32714752197265623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,4,8,power_law_1.01,0.3536652755737305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,4,8,power_law_1.01,0.41957118988037106
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,4,8,power_law_1.01,0.520654067993164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,4,8,power_law_1.01,0.6587289428710937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,4,8,power_law_1.01,0.8211251068115235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,4,8,power_law_1.01,1.1518950653076172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,4,8,power_law_1.01,1.473534698486328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,4,8,power_law_1.01,2.0907443237304686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,4,8,power_law_1.01,2.7184664916992185
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,4,8,power_law_1.01,4.058482055664062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,4,8,balanced,0.16727935791015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,4,8,balanced,0.17966079711914062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,4,8,balanced,0.25155839920043943
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,4,8,balanced,0.27572479248046877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,4,8,power_law_1.01,0.09379199981689454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,4,8,power_law_1.01,0.10060159683227539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,4,8,power_law_1.01,0.10364671707153321
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,4,8,power_law_1.01,0.11699456214904784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,4,8,power_law_1.01,0.06291327953338624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,4,8,balanced,0.4207014465332032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,4,8,power_law_1.01,0.12262144088745117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,4,8,power_law_1.01,0.13585280418395995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,4,8,power_law_1.01,0.16257280349731446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,4,8,power_law_1.01,0.16713855743408204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,4,8,power_law_1.01,0.17088640213012696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,4,8,power_law_1.01,0.1860428810119629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,4,8,power_law_1.01,0.19562496185302733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,4,8,power_law_1.01,0.21261951446533206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,4,8,power_law_1.01,0.19107839584350586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,4,8,power_law_1.01,0.20146175384521486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,4,8,power_law_1.01,0.22178432464599612
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,4,8,power_law_1.01,5.334337158203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,4,8,power_law_1.01,0.2651238441467285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,4,8,power_law_1.01,0.3127462387084961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,4,8,power_law_1.01,0.13513216018676757
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,4,8,power_law_1.01,0.40007167816162104
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,4,8,power_law_1.01,0.4892595291137695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,4,8,power_law_1.01,0.6578534698486328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,4,8,power_law_1.01,0.05106688022613526
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,4,8,balanced,0.5201702499389649
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,4,8,power_law_1.01,0.8421529388427734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,4,8,power_law_1.01,0.07334271907806397
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,4,8,power_law_1.01,0.06486144065856933
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,4,8,power_law_1.01,1.1891238403320312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,4,8,power_law_1.01,0.08172800064086913
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,4,8,power_law_1.01,0.08201472282409668
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,4,8,power_law_1.01,0.08337663650512696
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,4,8,power_law_1.01,0.08162176132202148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,4,8,power_law_1.01,1.5265498352050781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,4,8,power_law_1.01,0.08358400344848634
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,4,8,power_law_1.01,0.08288895606994628
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,4,8,power_law_1.01,0.08867584228515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,4,8,power_law_1.01,0.0880396842956543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,4,8,power_law_1.01,0.09230976104736328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,4,8,power_law_1.01,0.09036031723022461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,4,8,power_law_1.01,0.09692416191101075
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,4,8,power_law_1.01,0.13383423805236816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,4,8,power_law_1.01,2.256203460693359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,4,8,power_law_1.01,2.90948486328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,4,8,power_law_1.01,0.09965951919555664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,4,8,power_law_1.01,0.07657343864440917
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,4,8,power_law_1.01,0.10595199584960938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,4,8,power_law_1.01,0.16166143417358397
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,4,8,power_law_1.01,0.11813376426696778
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,4,8,power_law_1.01,0.19797504425048829
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,4,8,power_law_1.01,0.25272064208984374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,4,8,power_law_1.2,0.08829824447631836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,4,8,power_law_1.2,0.10838272094726562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,4,8,power_law_1.2,0.1329958438873291
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,4,8,power_law_1.01,0.32545406341552735
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,4,8,power_law_1.01,0.45537281036376953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,4,8,power_law_1.2,0.18308095932006835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,4,8,power_law_1.2,0.19860864639282227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,4,8,power_law_1.2,0.2578214454650879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,4,8,power_law_1.2,0.2520473670959473
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,4,8,power_law_1.2,0.27220863342285156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,4,8,power_law_1.2,0.29123071670532225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,4,8,power_law_1.2,0.29919872283935545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,4,8,power_law_1.2,0.32407039642333985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,4,8,power_law_1.2,0.22033664703369144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,4,8,power_law_1.2,0.3109491157531738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,4,8,power_law_1.2,0.22364032745361326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,4,8,power_law_1.2,0.22645120620727538
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,4,8,power_law_1.2,0.3290867233276367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,4,8,power_law_1.2,0.36000129699707034
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,4,8,power_law_1.2,0.4392537689208984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,4,8,power_law_1.2,0.5226419067382813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,4,8,power_law_1.2,0.6809394836425782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,4,8,power_law_1.2,0.8603212738037109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,4,8,power_law_1.2,1.1948480224609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,4,8,power_law_1.2,1.4986636352539062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,4,8,power_law_1.2,0.06464000225067139
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,4,8,power_law_1.2,0.08373632431030273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,4,8,power_law_1.2,0.09468159675598145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,4,8,power_law_1.2,2.7902041625976564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,4,8,power_law_1.2,0.10738304138183594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,4,8,power_law_1.2,2.1585740661621093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,4,8,power_law_1.2,0.12493439674377442
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,4,8,power_law_1.2,0.12296064376831055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,4,8,power_law_1.2,0.13627519607543945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,4,8,power_law_1.2,0.1411942386627197
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,4,8,power_law_1.2,0.16415359497070312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,4,8,power_law_1.2,0.17258880615234376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,4,8,power_law_1.2,0.16759424209594725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,4,8,power_law_1.2,0.18762752532958984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,4,8,power_law_1.2,4.262192687988281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,4,8,power_law_1.2,0.19610368728637695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,4,8,power_law_1.2,0.21569152832031252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,4,8,power_law_1.2,0.19233280181884765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,4,8,power_law_1.2,0.20658048629760742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,4,8,power_law_1.2,0.22873088836669925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,4,8,power_law_1.2,0.2755123138427734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,4,8,power_law_1.2,0.3159782409667969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,4,8,power_law_1.2,0.404398078918457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,4,8,power_law_1.2,0.5085516738891601
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,4,8,power_law_1.2,0.6796697235107422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,4,8,power_law_1.2,0.8639116668701172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,4,8,power_law_1.2,1.2182848358154297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,4,8,power_law_1.2,0.05130112171173096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,4,8,power_law_1.2,1.5830618286132814
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,4,8,power_law_1.2,0.06276480197906495
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,4,8,power_law_1.2,0.06468736171722413
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,4,8,power_law_1.2,0.07213056087493896
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,4,8,power_law_1.2,0.08475263595581055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,4,8,power_law_1.2,2.3173773193359373
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,4,8,power_law_1.2,0.07851391792297363
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,4,8,power_law_1.2,0.08024319648742675
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,4,8,power_law_1.2,0.08119296073913575
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,4,8,power_law_1.2,0.08355199813842773
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,4,8,power_law_1.2,0.08519935607910156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,4,8,power_law_1.2,0.08738304138183593
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,4,8,power_law_1.2,0.09026176452636718
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,4,8,power_law_1.2,0.0918489646911621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,4,8,power_law_1.2,3.027343444824219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,4,8,power_law_1.2,0.09712767601013184
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,4,8,power_law_1.2,0.10057600021362305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,4,8,power_law_1.2,5.684756469726563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,4,8,power_law_1.01,0.6710195159912109
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,4,8,power_law_1.01,0.9021094512939453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,4,8,power_law_1.01,1.2501888275146484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,4,8,power_law_1.2,0.10265600204467773
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,4,8,power_law_1.2,0.11508607864379883
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,4,8,power_law_1.2,0.1480460834503174
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,4,8,power_law_1.2,0.12975359916687013
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,4,8,power_law_1.2,0.19393407821655273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,8,4,balanced,0.028460800647735596
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,8,4,balanced,0.04742271900177002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,8,4,balanced,0.06573696136474609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,8,4,balanced,0.09635711669921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,8,4,balanced,0.17348480224609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,8,4,balanced,0.2388006401062012
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,8,4,balanced,0.24764160156249998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,8,4,balanced,0.25040895462036133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,8,4,balanced,0.2621772766113281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,8,4,balanced,0.26486911773681643
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,8,4,balanced,0.2684223937988281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,8,4,balanced,0.29830528259277345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,8,4,balanced,0.3038425636291504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,8,4,balanced,0.32357887268066404
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,8,4,balanced,0.3090137672424317
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,8,4,balanced,0.32278526306152344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,4,8,power_law_1.2,0.2428044891357422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,4,8,power_law_1.2,0.34116607666015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,8,4,balanced,0.3495999908447266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,8,4,balanced,0.4090444946289063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,8,4,balanced,0.464318733215332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,4,8,power_law_1.2,0.4522560119628906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,8,4,balanced,0.6390528106689454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,8,4,balanced,0.7420992279052735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,8,4,balanced,1.053729248046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,8,4,balanced,1.2723865509033203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,8,4,balanced,1.8967001342773435
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,8,4,balanced,0.03175551891326904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,8,4,balanced,0.0384934401512146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,8,4,balanced,2.426574096679688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,8,4,balanced,0.05797247886657715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,8,4,balanced,0.08666751861572267
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,8,4,balanced,0.12758015632629394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,8,4,balanced,0.17431039810180665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,8,4,balanced,0.16491519927978515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,8,4,balanced,0.17652992248535154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,8,4,balanced,0.18633472442626953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,8,4,balanced,0.1895449638366699
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,8,4,balanced,0.19287679672241212
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,8,4,balanced,0.2226572799682617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,8,4,balanced,0.22741119384765626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,8,4,balanced,3.582703247070312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,8,4,balanced,0.24680192947387694
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,8,4,balanced,0.2212633514404297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,8,4,balanced,0.2293619155883789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,8,4,balanced,0.2507648086547852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,8,4,balanced,0.28702335357666015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,8,4,balanced,0.3143564796447754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,8,4,balanced,0.3781452941894531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,8,4,balanced,0.4425446319580078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,8,4,balanced,4.650798034667969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,8,4,balanced,0.679210205078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,8,4,balanced,0.8017088317871094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,8,4,balanced,1.1603187561035155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,8,4,power_law_1.01,0.06241919994354248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,8,4,balanced,1.4126707458496095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,8,4,power_law_1.01,0.09126399993896485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,8,4,power_law_1.01,0.08941311836242676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,8,4,balanced,2.1340377807617186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,8,4,power_law_1.01,0.12141823768615723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,8,4,power_law_1.01,0.14596351623535156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,8,4,power_law_1.01,0.2091289520263672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,8,4,power_law_1.01,0.2288345527648926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,8,4,power_law_1.01,0.2368691253662109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,8,4,power_law_1.01,0.2530112075805664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,8,4,power_law_1.01,0.24425983428955078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,8,4,balanced,2.748834533691406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,8,4,power_law_1.01,0.25128063201904294
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,8,4,power_law_1.01,0.2867967987060547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,8,4,power_law_1.01,0.3205811309814453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,8,4,power_law_1.01,0.3026803207397461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,8,4,power_law_1.01,0.31701759338378904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,8,4,power_law_1.01,0.40342273712158205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,8,4,power_law_1.01,0.29441152572631835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,8,4,power_law_1.01,0.6280767822265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,8,4,power_law_1.01,0.7785075378417969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,8,4,power_law_1.01,1.0784268951416016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,4,8,power_law_1.2,0.6464947509765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,8,4,power_law_1.01,0.3459904098510742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,4,8,power_law_1.2,0.9547878265380859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,8,4,power_law_1.01,0.4945510482788086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,8,4,power_law_1.01,1.3710823059082031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,4,8,power_law_1.2,1.16714111328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,8,4,power_law_1.01,0.04243840217590332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,8,4,power_law_1.01,0.06823552131652832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,8,4,power_law_1.01,0.07491583824157715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,8,4,power_law_1.01,0.1160870361328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,8,4,power_law_1.01,1.9164544677734376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,8,4,power_law_1.01,0.1475609588623047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,8,4,power_law_1.01,0.15807616233825683
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,8,4,power_law_1.01,0.16775423049926758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,8,4,power_law_1.01,0.17341312408447265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,8,4,power_law_1.01,0.17808767318725588
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,8,4,power_law_1.01,0.10592896461486816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,8,4,power_law_1.01,2.548970184326172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,8,4,power_law_1.01,0.2142630386352539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,8,4,power_law_1.01,0.21949567794799804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,8,4,power_law_1.01,0.23492095947265623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,8,4,power_law_1.01,0.21051519393920898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,8,4,power_law_1.01,0.2215692710876465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,8,4,power_law_1.01,0.2378816032409668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,8,4,power_law_1.01,0.28731136322021483
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,8,4,power_law_1.01,0.32222335815429687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,8,4,power_law_1.01,0.4103488159179688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,8,4,power_law_1.01,0.18389888763427736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,8,4,power_law_1.01,0.5006771087646484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,8,4,power_law_1.01,0.6611865234375001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,8,4,power_law_1.01,0.8274317169189453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,8,4,power_law_1.01,3.665186462402344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,8,4,power_law_1.2,0.06263679981231689
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,8,4,power_law_1.01,1.1828096008300781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,8,4,power_law_1.01,4.917148132324219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,8,4,power_law_1.2,0.09191295623779297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,8,4,power_law_1.01,1.5096243286132813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,8,4,power_law_1.2,0.13703295707702637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,8,4,power_law_1.2,0.21468544006347656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,8,4,power_law_1.2,0.2231500816345215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,8,4,power_law_1.2,0.2315033531188965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,8,4,power_law_1.2,0.23777408599853517
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,4,8,power_law_1.2,1.913800964355469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,8,4,power_law_1.01,2.2196826171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,8,4,power_law_1.2,0.24527231216430664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,8,4,power_law_1.2,0.252097282409668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,8,4,power_law_1.2,0.28921472549438476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,8,4,power_law_1.2,0.2953945541381836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,8,4,power_law_1.2,0.07591040134429931
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,8,4,power_law_1.2,0.3192473602294922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,8,4,power_law_1.2,0.3057356834411621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,8,4,power_law_1.01,2.855433044433594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,8,4,power_law_1.2,0.1139571189880371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,8,4,power_law_1.2,0.32175487518310547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,8,4,power_law_1.2,0.35221118927001954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,8,4,power_law_1.2,0.4158233642578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,8,4,power_law_1.2,0.5082457733154297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,8,4,power_law_1.2,0.6396659088134766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,8,4,power_law_1.2,0.8224076843261718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,8,4,power_law_1.2,0.04370431900024414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,8,4,power_law_1.2,1.4018598937988282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,8,4,power_law_1.2,0.07673215866088867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,8,4,power_law_1.2,0.059782400131225585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,8,4,power_law_1.2,0.11363583564758302
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,8,4,power_law_1.2,0.10576383590698242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,8,4,power_law_1.2,0.14056320190429689
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,8,4,power_law_1.2,1.9646861267089843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,8,4,power_law_1.2,0.15606783866882323
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,8,4,power_law_1.2,0.16233856201171876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,8,4,power_law_1.2,0.17750783920288088
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,8,4,power_law_1.2,0.18321920394897462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,8,4,power_law_1.2,2.602259216308594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,8,4,power_law_1.2,0.21327232360839843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,8,4,power_law_1.2,0.21882623672485352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,8,4,power_law_1.2,0.2358143997192383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,8,4,power_law_1.2,0.21435775756835937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,8,4,power_law_1.2,0.22665727615356443
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,8,4,power_law_1.2,1.1053632354736327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,8,4,power_law_1.2,0.25204479217529296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,8,4,power_law_1.2,0.2889638328552246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,8,4,power_law_1.2,0.17385087966918947
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,8,4,power_law_1.2,0.3335168075561524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,8,4,power_law_1.2,0.41259777069091796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,8,4,power_law_1.2,3.740372619628906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,8,4,power_law_1.2,0.5109657669067383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,8,4,power_law_1.2,0.6792447662353516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,16,2,balanced,0.0678553581237793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,8,4,power_law_1.2,0.8461465454101562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,16,2,balanced,0.10550784111022948
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,8,4,power_law_1.2,1.211310043334961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,16,2,balanced,0.33327873229980465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,8,4,power_law_1.2,1.5504077148437498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,16,2,balanced,0.4676147079467773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,16,2,balanced,0.47824256896972656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,16,2,balanced,0.493963508605957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,16,2,balanced,0.4976076889038086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,8,4,power_law_1.2,2.287342071533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,16,2,balanced,0.5045580673217773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,16,2,balanced,0.5205977630615235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,16,2,balanced,0.5274303817749024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,16,2,balanced,0.17993087768554689
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,16,2,balanced,0.5503398513793945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,16,2,balanced,0.4831948852539062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,16,2,balanced,0.03872128009796143
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,16,2,balanced,0.5917606353759766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,16,2,balanced,0.6018521499633789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,16,2,balanced,0.6292633438110352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,16,2,balanced,0.6817088317871094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,16,2,balanced,0.7337267303466797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,16,2,balanced,1.2515007781982423
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,16,2,balanced,1.790771179199219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,8,4,power_law_1.2,5.114823608398437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,16,2,balanced,1.9961932373046873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,16,2,balanced,0.19783679962158202
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,16,2,balanced,0.34025600433349606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,16,2,balanced,3.040111389160156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,16,2,balanced,0.6216563034057617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,8,4,power_law_1.2,2.9573809814453123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,16,2,balanced,3.794150390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,16,2,balanced,1.3622579956054688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,16,2,balanced,1.6420262145996094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,16,2,balanced,1.1494246673583983
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,16,2,balanced,2.0973350524902346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,16,2,balanced,1.9929446411132812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,16,2,balanced,5.535848999023438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,16,2,balanced,1.6361331176757812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,16,2,balanced,1.9875506591796874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,16,2,balanced,2.151751708984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,16,2,balanced,1.6657958984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,16,2,balanced,2.0621183776855467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,16,2,balanced,6.957703857421874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,16,2,balanced,2.145964813232422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,16,2,balanced,2.7227493286132813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,16,2,balanced,2.2082598876953123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,16,2,balanced,2.2874009704589846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,16,2,balanced,2.4027993774414065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,16,2,balanced,2.7769561767578126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,16,2,balanced,2.3938829040527345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,16,2,balanced,4.683705749511718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,16,2,balanced,5.705043334960938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,16,2,balanced,6.822667236328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,16,2,power_law_1.01,0.06380032062530518
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,16,2,power_law_1.01,0.1006054401397705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,16,2,balanced,7.259009399414063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,16,2,power_law_1.01,0.11991040229797363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,16,2,power_law_1.01,0.24987775802612305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,16,2,power_law_1.01,0.37702144622802736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,16,2,power_law_1.01,0.41430912017822263
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,16,2,power_law_1.01,0.4349081420898438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,16,2,power_law_1.01,0.4578316879272461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,16,2,power_law_1.01,0.18725503921508788
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,16,2,power_law_1.01,0.4599628829956055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,16,2,power_law_1.01,0.4923571014404297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,16,2,power_law_1.01,0.520951042175293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,16,2,power_law_1.01,0.5575193786621094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,16,2,power_law_1.01,0.6415782165527344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,16,2,power_law_1.01,0.5756198501586913
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,16,2,power_law_1.01,0.5976780700683594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,16,2,power_law_1.01,0.6354662322998047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,16,2,power_law_1.01,0.7331903839111329
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,16,2,power_law_1.01,0.8698841857910157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,16,2,power_law_1.01,1.0733324432373048
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,16,2,power_law_1.01,1.3094117736816406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,16,2,power_law_1.01,1.7404147338867186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,16,2,power_law_1.01,2.144417266845703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,16,2,balanced,13.643211669921874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,16,2,balanced,14.558592529296874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,16,2,power_law_1.01,2.992010192871094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,16,2,balanced,0.14595968246459962
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,16,2,power_law_1.01,3.8872882080078126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,16,2,power_law_1.01,0.7365555572509765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,16,2,power_law_1.01,1.058419189453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,16,2,power_law_1.01,5.647053833007813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,16,2,power_law_1.01,1.6023526000976562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,16,2,power_law_1.01,1.6142233276367186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,16,2,power_law_1.01,1.4420697021484377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,16,2,power_law_1.01,1.6541734313964844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,16,2,power_law_1.01,7.2675366210937495
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,16,2,balanced,1.6303948974609377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,16,2,power_law_1.01,1.6975155639648438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,16,2,power_law_1.01,1.4826457214355468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,16,2,power_law_1.01,1.7485862731933595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,16,2,power_law_1.01,1.5631680297851562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,16,2,power_law_1.01,1.884064025878906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,16,2,power_law_1.01,2.5701376342773434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,16,2,power_law_1.01,2.0540594482421874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,16,2,power_law_1.01,2.0702796936035157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,16,2,power_law_1.01,2.528412170410156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,16,2,power_law_1.01,3.3861260986328126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,16,2,power_law_1.01,3.688445434570313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,16,2,power_law_1.01,4.166112060546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,16,2,power_law_1.01,4.962728881835938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,16,2,power_law_1.01,6.78552734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,16,2,power_law_1.2,0.06644351959228516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,16,2,power_law_1.2,0.093951997756958
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,16,2,power_law_1.2,0.12483200073242187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,16,2,power_law_1.01,10.559737548828124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,16,2,power_law_1.2,0.2296998405456543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,16,2,power_law_1.01,0.19647743225097655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,16,2,power_law_1.01,0.4434572982788086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,16,2,power_law_1.01,0.3834944152832031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,16,2,power_law_1.2,0.38592254638671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,16,2,power_law_1.2,0.396940803527832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,16,2,power_law_1.2,0.42718593597412113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,16,2,power_law_1.2,0.4431283187866211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,16,2,power_law_1.2,0.17931776046752929
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,16,2,power_law_1.2,0.4609344100952148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,16,2,power_law_1.2,0.48570880889892576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,16,2,power_law_1.01,11.8947412109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,16,2,power_law_1.2,0.5313702392578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,16,2,power_law_1.2,0.5732377624511719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,16,2,power_law_1.2,0.5816128158569336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,16,2,power_law_1.2,0.6631295776367188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,16,2,power_law_1.2,0.6069401550292969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,16,2,power_law_1.2,0.6628659057617188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,16,2,power_law_1.2,0.7630630493164062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,16,2,power_law_1.2,0.8932275390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,16,2,power_law_1.2,1.108573455810547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,16,2,power_law_1.2,1.340457000732422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,16,2,power_law_1.2,1.7846220397949217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,16,2,power_law_1.2,0.19267967224121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,16,2,power_law_1.2,2.184784698486328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,16,2,power_law_1.2,0.2894476890563965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,16,2,power_law_1.2,0.5886822509765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,16,2,power_law_1.2,3.0487078857421874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,16,2,power_law_1.01,14.768719482421876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,16,2,power_law_1.2,0.675462417602539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,16,2,power_law_1.2,3.9638003540039066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,16,2,power_law_1.2,1.0393126678466795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,16,2,power_law_1.2,1.3575999450683593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,16,2,power_law_1.2,1.4013786315917969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,16,2,power_law_1.2,1.6185139465332032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,16,2,power_law_1.2,1.4595558166503906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,16,2,power_law_1.2,5.775537719726563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,16,2,power_law_1.2,1.908294372558594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,16,2,power_law_1.2,1.5108262634277343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,16,2,power_law_1.2,1.5282931518554688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,16,2,power_law_1.2,1.9458828735351563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,16,2,power_law_1.2,1.6937600708007814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,16,2,power_law_1.2,2.085198059082031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,16,2,power_law_1.2,2.287619781494141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,16,2,power_law_1.2,2.5580006408691407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,16,2,power_law_1.01,2.928053894042969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,16,2,power_law_1.2,2.338656005859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,16,2,power_law_1.2,2.6116683959960936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,16,2,power_law_1.2,3.0429171752929687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,16,2,power_law_1.2,3.614726257324219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,16,2,power_law_1.2,7.400883178710937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,16,2,power_law_1.2,5.541303100585937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,32,1,balanced,0.1025932788848877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,32,1,balanced,0.06249087810516357
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,16,2,power_law_1.2,6.526688232421876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,16,2,power_law_1.2,6.872318725585937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,32,1,balanced,0.15440192222595214
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,32,1,balanced,0.25847295761108396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,16,2,power_law_1.2,8.330398559570312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,32,1,balanced,0.6828031921386719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,32,1,balanced,0.6895584106445313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,32,1,balanced,0.4795993423461914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,32,1,balanced,0.7058694458007813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,32,1,balanced,0.7127532958984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,32,1,balanced,0.7306784057617188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,32,1,balanced,0.7376115417480469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,32,1,balanced,0.7577753448486328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,32,1,balanced,0.6727046203613282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,32,1,balanced,0.7042598724365234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,32,1,balanced,0.7415808105468751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,32,1,balanced,1.0110157012939454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,32,1,balanced,1.3103155517578124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,32,1,balanced,0.9142649841308593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,32,1,balanced,0.96246337890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,16,2,power_law_1.2,15.97992431640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,32,1,balanced,1.5269151306152344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,32,1,balanced,1.6281759643554685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,16,2,power_law_1.2,15.38798583984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,32,1,balanced,2.3285504150390626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,32,1,balanced,2.5485292053222657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,32,1,balanced,3.9133843994140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,32,1,balanced,4.823639831542969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,32,1,balanced,0.15641599655151367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,32,1,balanced,0.3178688049316406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,32,1,balanced,0.4481331253051758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,32,1,balanced,0.8326099395751954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,32,1,balanced,7.06879638671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,32,1,balanced,8.874705200195312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,32,1,balanced,2.1657286071777344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,32,1,balanced,2.259396514892578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,32,1,balanced,2.9037799072265624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,32,1,balanced,2.710730895996094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,32,1,balanced,2.6188952636718748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,32,1,balanced,2.2977113342285156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,32,1,balanced,2.32472900390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,32,1,balanced,2.2776095581054685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,32,1,balanced,2.82860595703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,32,1,balanced,2.7333395385742185
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,32,1,balanced,2.6222573852539064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,32,1,balanced,2.298936309814453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,32,1,balanced,2.587666015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,32,1,balanced,2.8163348388671876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,32,1,balanced,2.9605035400390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,32,1,balanced,4.3603994750976565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,32,1,power_law_1.01,0.0613267183303833
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,32,1,power_law_1.01,0.09910207748413086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,32,1,balanced,4.742844848632812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,32,1,power_law_1.01,0.1540780830383301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,32,1,power_law_1.01,0.33220863342285156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,32,1,power_law_1.01,0.5257875061035155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,32,1,balanced,6.274201660156249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,32,1,power_law_1.01,0.5630073547363281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,32,1,power_law_1.01,0.5892095947265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,32,1,power_law_1.01,0.23866624832153321
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,32,1,power_law_1.01,0.6103174209594726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,32,1,power_law_1.01,0.6281491088867187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,32,1,power_law_1.01,0.6769344329833984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,32,1,power_law_1.01,0.7287673950195312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,32,1,power_law_1.01,0.7820812988281249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,32,1,balanced,9.329017333984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,32,1,power_law_1.01,0.900412826538086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,32,1,power_law_1.01,0.9720281219482422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,32,1,power_law_1.01,1.0790451049804688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,32,1,power_law_1.01,1.2834054565429687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,32,1,power_law_1.01,1.0220480346679688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,32,1,power_law_1.01,1.1242406463623047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,32,1,balanced,11.026005859375001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,32,1,power_law_1.01,1.381964874267578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,32,1,balanced,11.503823242187499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,32,1,power_law_1.01,1.6783084106445312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,32,1,power_law_1.01,2.187330627441406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,32,1,power_law_1.01,2.692778930664063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,32,1,power_law_1.01,0.284330883026123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,32,1,balanced,14.107662353515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,32,1,power_law_1.01,3.753116149902344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,32,1,power_law_1.01,0.49795776367187494
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,32,1,power_law_1.01,0.8426041412353517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,32,1,power_law_1.01,4.816618957519531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,32,1,power_law_1.01,1.4319007873535157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,32,1,power_law_1.01,1.7011769104003907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,32,1,power_law_1.01,1.7867686462402343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,32,1,power_law_1.01,6.935375366210937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,32,1,power_law_1.01,2.2238681030273435
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,32,1,power_law_1.01,1.9136480712890624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,32,1,power_law_1.01,1.9432838439941407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,32,1,power_law_1.01,1.9920697021484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,32,1,power_law_1.01,2.073768310546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,32,1,power_law_1.01,2.3829965209960937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,32,1,balanced,20.769475097656247
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,32,1,power_law_1.01,9.041307373046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,32,1,power_law_1.01,2.153852233886719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,32,1,power_law_1.01,2.2131622314453123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,32,1,power_law_1.01,2.328539581298828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,32,1,power_law_1.01,2.571697998046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,32,1,power_law_1.01,3.0929464721679687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,32,1,power_law_1.01,3.188452453613281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,32,1,power_law_1.01,5.746023559570313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,32,1,power_law_1.01,4.963511047363282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,32,1,power_law_1.2,0.06039616107940674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,32,1,power_law_1.01,6.1549615478515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,32,1,power_law_1.2,0.10177023887634278
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,32,1,power_law_1.2,0.1537452793121338
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,32,1,power_law_1.2,0.22070655822753907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,32,1,power_law_1.2,0.3013555145263672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,32,1,power_law_1.2,0.5178271865844726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,32,1,power_law_1.2,0.5633299255371094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,32,1,power_law_1.2,0.5829715347290039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,32,1,power_law_1.2,0.6139648056030274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,32,1,power_law_1.01,7.502623901367187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,32,1,power_law_1.2,0.6459212493896485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,32,1,power_law_1.2,0.691546859741211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,32,1,power_law_1.2,0.7575244903564453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,32,1,power_law_1.2,0.8125228881835938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,32,1,power_law_1.2,0.9267142486572265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,32,1,power_law_1.2,1.0107923126220704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,32,1,power_law_1.01,9.579827270507812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,32,1,power_law_1.2,1.1103616333007813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,32,1,power_law_1.2,1.0458937835693358
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,32,1,power_law_1.2,1.3145677185058593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,32,1,power_law_1.2,1.1775039672851562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,32,1,power_law_1.2,1.4393081665039062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,32,1,power_law_1.01,11.31703369140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,32,1,power_law_1.01,0.18543935775756837
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,32,1,power_law_1.2,1.6935192871093752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,32,1,power_law_1.2,2.2048486328125003
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,32,1,power_law_1.2,0.17628927230834962
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,32,1,power_law_1.2,2.7431475830078123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,32,1,power_law_1.2,0.2824511909484863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,32,1,power_law_1.2,0.6960179138183593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,32,1,power_law_1.01,15.009666748046874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,32,1,power_law_1.2,3.773471374511719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,32,1,power_law_1.2,0.9527168273925781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,32,1,power_law_1.2,1.6739077758789063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,32,1,power_law_1.2,1.7869754028320313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,32,1,power_law_1.2,4.8370336914062495
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,32,1,power_law_1.2,0.4373145675659179
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,32,1,power_law_1.2,1.836201629638672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,32,1,power_law_1.2,1.93197509765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,32,1,power_law_1.2,1.9750093078613282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,32,1,power_law_1.2,2.37728515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,32,1,power_law_1.2,2.2515565490722658
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,32,1,power_law_1.2,2.135148162841797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,32,1,power_law_1.2,2.281095733642578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,32,1,power_law_1.2,2.3509983825683594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,32,1,power_law_1.2,2.8075079345703124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,32,1,power_law_1.2,2.730940246582031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,32,1,power_law_1.2,4.456163940429688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,32,1,power_law_1.2,6.966876831054688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,32,1,power_law_1.2,3.678038940429688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,32,1,power_law_1.2,3.3301779174804684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,32,1,power_law_1.2,4.754157409667969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,32,1,power_law_1.2,5.5322418212890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,8,8,balanced,0.03127039909362793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,8,8,balanced,0.031315200328826905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,8,8,balanced,0.031261439323425295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,8,8,balanced,0.05398143768310547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,8,8,balanced,0.055822081565856926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,8,8,balanced,0.05630720138549805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,8,8,balanced,0.05712512016296387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,8,8,balanced,0.05712255954742431
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,8,8,balanced,0.058296318054199214
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,8,8,balanced,0.05916031837463379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,8,8,balanced,0.0595033597946167
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,8,8,balanced,0.06074880123138428
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,8,8,balanced,0.061227521896362304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,8,8,balanced,0.06247039794921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,8,8,balanced,0.06335103988647461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,8,8,balanced,0.0636198377609253
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,8,8,balanced,0.06550655841827394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,8,8,balanced,0.06986752033233642
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,8,8,balanced,0.07280767917633056
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,32,1,power_law_1.01,29.239716796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,8,8,balanced,0.09454719543457031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,8,8,balanced,0.08705792427062989
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,8,8,balanced,0.12573951721191406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,8,8,balanced,0.15025407791137696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,8,8,balanced,0.201397762298584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,8,8,balanced,0.2529587173461914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,8,8,balanced,0.36535552978515623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,8,8,balanced,0.0312716794013977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,8,8,balanced,0.4777113723754883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,8,8,balanced,0.0301145601272583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,8,8,balanced,0.031025919914245602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,8,8,balanced,0.03910399913787842
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,8,8,balanced,0.03991552114486695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,8,8,balanced,0.03962111949920654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,8,8,balanced,0.04008959770202637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,8,8,balanced,0.03970688104629517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,32,1,power_law_1.2,9.366533203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,8,8,balanced,0.040357117652893064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,8,8,balanced,0.03999615907669067
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,8,8,balanced,0.040332798957824705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,8,8,balanced,0.03992063999176025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,8,8,balanced,0.04118144035339356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,8,8,balanced,0.04109951972961426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,8,8,balanced,0.04159743785858154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,8,8,balanced,0.04324351787567139
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,8,8,balanced,0.04486400127410889
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,8,8,balanced,0.046430721282958984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,8,8,balanced,0.04025599956512451
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,8,8,balanced,0.05058559894561767
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,8,8,balanced,0.049199361801147465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,8,8,balanced,0.05845759868621826
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,8,8,balanced,0.06987264156341552
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,8,8,balanced,0.09851903915405273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,8,8,balanced,0.11388416290283203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,8,8,balanced,0.15784576416015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,8,8,balanced,0.21530368804931638
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,32,1,power_law_1.2,9.767086791992188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,32,1,power_law_1.2,13.147889404296876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,32,1,power_law_1.2,9.09158447265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,32,1,power_law_1.2,15.240904541015624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,8,8,balanced,0.016533759832382203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,8,8,balanced,0.016170239448547362
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,8,8,balanced,0.01659775972366333
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,8,8,balanced,0.016892160177230837
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,8,8,balanced,0.016119040250778198
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,8,8,balanced,0.016505600214004518
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,8,8,balanced,0.016743680238723753
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,8,8,balanced,0.016599040031433105
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,8,8,balanced,0.01693567991256714
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,8,8,balanced,0.017023999691009522
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,8,8,balanced,0.016517119407653807
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,8,8,balanced,0.01690879940986633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,8,8,balanced,0.01702272057533264
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,8,8,balanced,0.01654528021812439
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,8,8,balanced,0.017009919881820677
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,8,8,balanced,0.016641279458999632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,32,1,power_law_1.2,20.125360107421876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,8,8,balanced,0.016576000452041627
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,8,8,balanced,0.018383359909057616
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,8,8,balanced,0.01815935969352722
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,8,8,balanced,0.019435520172119143
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,8,8,balanced,0.021079039573669432
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,8,8,power_law_1.01,0.030600960254669192
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,8,8,power_law_1.01,0.05427711963653564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,8,8,power_law_1.01,0.05467391967773437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,8,8,power_law_1.01,0.05703936100006104
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,8,8,power_law_1.01,0.05835648059844971
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,8,8,power_law_1.01,0.06034175872802734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,8,8,power_law_1.01,0.04892928123474121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,8,8,power_law_1.01,0.06124288082122803
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,8,8,power_law_1.01,0.06247039794921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,8,8,power_law_1.01,0.06286464214324952
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,8,8,power_law_1.01,0.06425600051879883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,8,8,power_law_1.01,0.06870399951934815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,8,8,power_law_1.01,0.06859263896942139
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,8,8,power_law_1.01,0.07067903995513916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,8,8,power_law_1.01,0.09180800437927246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,8,8,power_law_1.01,0.09636096000671387
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,8,8,balanced,0.026981120109558103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,8,8,power_law_1.01,0.09607551574707032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,8,8,power_law_1.01,0.11541248321533204
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,8,8,balanced,0.028903679847717283
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,8,8,power_law_1.01,0.13484800338745118
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,8,8,power_law_1.01,0.20564992904663085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,8,8,power_law_1.01,0.23122432708740231
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,8,8,power_law_1.01,0.32475521087646486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,8,8,power_law_1.01,0.4633881759643555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,8,8,balanced,0.04768511772155762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,8,8,power_law_1.01,1.1561638641357423
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,8,8,balanced,0.03880320072174072
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,8,8,power_law_1.01,0.03826175928115845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,8,8,power_law_1.01,0.03952640056610107
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,8,8,power_law_1.01,0.03992063999176025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,8,8,power_law_1.01,0.039993600845336916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,8,8,power_law_1.01,0.04034560203552246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,8,8,power_law_1.01,0.030499839782714845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,8,8,power_law_1.01,0.040757761001586915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,8,8,power_law_1.01,0.04154623985290527
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,8,8,power_law_1.01,0.04113279819488526
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,8,8,power_law_1.01,0.03992448091506958
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,8,8,power_law_1.01,0.7028006744384766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,8,8,power_law_1.01,0.04236671924591064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,8,8,power_law_1.01,0.044257278442382815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,8,8,power_law_1.01,0.04564864158630371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,8,8,power_law_1.01,0.044556798934936526
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,8,8,power_law_1.01,0.0481331205368042
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,8,8,power_law_1.01,0.04745215892791748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,8,8,power_law_1.01,0.050958080291748045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,8,8,power_law_1.01,0.05582719802856445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,8,8,power_law_1.01,0.0675110387802124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,8,8,power_law_1.01,0.9063462066650392
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,8,8,power_law_1.01,0.14215807914733886
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,8,8,power_law_1.01,0.09091967582702637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,8,8,power_law_1.01,0.11172224044799806
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,8,8,balanced,0.06366335868835449
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,8,8,balanced,0.07967103958129883
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,8,8,power_law_1.01,0.016533759832382203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,8,8,power_law_1.01,0.01640959978103638
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,8,8,power_law_1.01,0.01658751964569092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,8,8,power_law_1.01,0.2846195220947266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,8,8,power_law_1.01,0.016131839752197265
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,8,8,power_law_1.01,0.016501760482788085
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,8,8,power_law_1.01,0.016637439727783206
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,8,8,power_law_1.01,0.016981760263442995
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,8,8,power_law_1.01,0.016939519643783568
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,8,8,power_law_1.01,0.0169267201423645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,8,8,power_law_1.01,0.20618112564086913
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,8,8,power_law_1.01,0.0170035195350647
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,8,8,power_law_1.01,0.017742079496383668
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,8,8,power_law_1.01,0.018165760040283203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,8,8,power_law_1.01,0.5399359893798829
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,8,8,power_law_1.01,1.7846937561035154
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,8,8,power_law_1.01,0.020247039794921876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,8,8,power_law_1.01,0.02402048110961914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,8,8,power_law_1.01,0.3787263870239258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,8,8,power_law_1.01,0.7623923492431641
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,8,8,power_law_1.01,0.01703935980796814
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,8,8,power_law_1.01,0.018725119829177857
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,8,8,power_law_1.01,0.018256640434265135
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,8,8,power_law_1.01,0.018644479513168336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,8,8,power_law_1.01,0.02726272106170654
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,8,8,power_law_1.01,0.03498752117156982
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,8,8,power_law_1.2,0.030924799442291262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,8,8,power_law_1.2,0.052610559463500975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,8,8,power_law_1.2,0.05385216236114502
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,8,8,power_law_1.2,0.05468416213989258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,8,8,power_law_1.2,0.057114877700805665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,8,8,power_law_1.2,0.05878143787384034
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,8,8,power_law_1.2,0.060140800476074216
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,8,8,power_law_1.2,0.061582078933715825
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,8,8,power_law_1.2,0.06285823822021484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,8,8,power_law_1.01,0.04543744087219238
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,8,8,power_law_1.2,0.06335103988647461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,8,8,power_law_1.2,0.06884352207183839
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,8,8,power_law_1.2,0.06524415969848632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,8,8,power_law_1.2,0.0689024019241333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,8,8,power_law_1.2,0.07367167949676515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,8,8,power_law_1.2,0.09864704132080078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,8,8,power_law_1.2,0.08907008171081543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,8,8,power_law_1.2,0.08735103607177734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,8,8,power_law_1.2,0.11598079681396485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,8,8,power_law_1.2,0.15024767875671388
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,8,8,power_law_1.2,0.20441471099853517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,8,8,power_law_1.2,0.24783103942871093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,8,8,power_law_1.01,0.05662335872650146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,8,8,power_law_1.2,0.3649203109741211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,8,8,power_law_1.2,0.4512704086303711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,8,8,power_law_1.01,0.07682816028594971
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,8,8,power_law_1.2,0.6619558715820313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,8,8,power_law_1.01,0.10182271957397462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,8,8,power_law_1.2,0.03051775932312012
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,8,8,power_law_1.2,0.039461119174957274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,8,8,power_law_1.2,0.039521279335021975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,8,8,power_law_1.2,0.039521279335021975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,8,8,power_law_1.2,0.039921920299530025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,8,8,power_law_1.2,0.040284161567687986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,8,8,power_law_1.2,0.04032256126403809
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,8,8,power_law_1.2,0.04132224082946777
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,8,8,power_law_1.2,0.04115583896636963
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,8,8,power_law_1.2,0.041958398818969726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,8,8,power_law_1.2,0.04302591800689697
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,8,8,power_law_1.2,0.04443520069122314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,8,8,power_law_1.2,0.044843521118164066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,8,8,power_law_1.2,0.04699007987976074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,8,8,power_law_1.2,0.048115200996398926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,8,8,power_law_1.2,0.05040128231048584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,8,8,power_law_1.01,0.13531904220581054
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,8,8,power_law_1.2,0.05188223838806152
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,8,8,power_law_1.2,0.057242879867553706
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,8,8,power_law_1.2,0.06590720176696777
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,8,8,power_law_1.2,0.09725055694580079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,8,8,power_law_1.2,0.11379712104797364
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,8,8,power_law_1.2,0.1516646385192871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,8,8,power_law_1.2,0.20327423095703123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,8,8,power_law_1.2,0.9258470153808593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,8,8,power_law_1.2,0.27985664367675783
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,8,8,power_law_1.2,0.016202239990234374
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,8,8,power_law_1.2,0.016592639684677123
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,8,8,power_law_1.2,0.016577279567718504
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,8,8,power_law_1.2,0.016436480283737183
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,8,8,power_law_1.2,0.01659775972366333
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,8,8,power_law_1.2,0.016536320447921755
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,8,8,power_law_1.2,0.016899839639663697
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,8,8,power_law_1.2,0.016596479415893556
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,8,8,power_law_1.2,0.01690879940986633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,8,8,power_law_1.2,0.016591360569000242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,8,8,power_law_1.2,0.016962560415267943
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,8,8,power_law_1.2,0.01780992031097412
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,8,8,power_law_1.2,0.018054399490356445
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,8,8,power_law_1.01,0.25902591705322264
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,8,8,power_law_1.2,0.018211840391159057
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,8,8,power_law_1.2,0.0186188805103302
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,8,8,power_law_1.2,0.01906048059463501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,8,8,power_law_1.2,0.505557746887207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,8,8,power_law_1.2,0.40020992279052736
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,8,8,power_law_1.2,0.028555519580841064
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,8,8,power_law_1.01,0.19910911560058592
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,8,8,power_law_1.2,1.5450764465332032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,8,8,power_law_1.2,0.6924237060546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,8,8,power_law_1.2,1.3499020385742189
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,8,8,power_law_1.2,0.0761894416809082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,8,8,power_law_1.2,0.020200960636138916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,8,8,power_law_1.2,0.02514559984207153
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,16,4,balanced,0.02634239912033081
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,16,4,balanced,0.026068480014801027
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,16,4,balanced,0.02898560047149658
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,16,4,balanced,0.04771584033966064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,16,4,balanced,0.049501438140869145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,16,4,balanced,0.049838080406188964
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,16,4,balanced,0.05082111835479737
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,16,4,balanced,0.051363840103149414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,16,4,balanced,0.05209727764129639
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,16,4,balanced,0.052203521728515626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,16,4,balanced,0.053186559677124025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,16,4,balanced,0.05506048202514648
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,8,8,power_law_1.2,0.038329598903656004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,16,4,balanced,0.055866880416870116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,16,4,balanced,0.057123842239379885
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,16,4,balanced,0.060803837776184075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,16,4,balanced,0.06259456157684326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,16,4,balanced,0.06284800052642822
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,16,4,balanced,0.08707839965820312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,16,4,balanced,0.07286528110504151
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,16,4,balanced,0.09973631858825684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,16,4,balanced,0.10894720077514648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,16,4,balanced,0.1292249584197998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,16,4,balanced,0.16316287994384765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,16,4,balanced,0.22640127182006836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,16,4,balanced,0.29495935440063475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,16,4,balanced,0.4298432159423828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,16,4,balanced,0.028221440315246583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,16,4,balanced,0.5640537643432617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,16,4,balanced,0.0280678391456604
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,16,4,balanced,0.030150399208068845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,16,4,balanced,0.036275200843811035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,16,4,balanced,0.03670399904251099
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,16,4,balanced,0.03678208112716675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,16,4,balanced,0.037544960975646975
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,8,8,power_law_1.2,0.04762112140655518
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,16,4,balanced,0.037150719165802
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,16,4,balanced,0.03746304035186768
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,16,4,balanced,0.03759104013442993
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,16,4,balanced,0.038119680881500244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,16,4,balanced,0.03871615886688232
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,16,4,balanced,0.0391923189163208
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,16,4,balanced,0.039961600303649904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,16,4,balanced,0.04078464031219482
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,16,4,balanced,0.04359039783477783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,16,4,balanced,0.04721280097961426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,16,4,balanced,0.04525568008422852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,16,4,balanced,0.05932672023773193
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,16,4,balanced,0.07506688117980957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,16,4,balanced,0.03709183931350708
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,16,4,balanced,0.08781567573547364
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,16,4,balanced,0.12667136192321776
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,16,4,balanced,0.15940735816955567
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,16,4,balanced,0.23004543304443362
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,16,4,balanced,0.05068799972534179
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,16,4,balanced,0.3032550430297852
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,8,8,power_law_1.2,0.061320958137512205
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,8,8,power_law_1.2,0.10345727920532226
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,8,8,power_law_1.2,0.17053823471069335
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,16,4,balanced,0.016784640550613402
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,16,4,balanced,0.01709568023681641
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,16,4,balanced,0.01697535991668701
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,16,4,balanced,0.016939519643783568
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,16,4,balanced,0.017379839420318604
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,16,4,balanced,0.0169868803024292
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,16,4,balanced,0.01742591977119446
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,16,4,balanced,0.01746816039085388
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,16,4,balanced,0.017840640544891356
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,16,4,balanced,0.01780608057975769
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,16,4,balanced,0.017825280427932737
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,8,8,power_law_1.2,0.13988479614257812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,16,4,balanced,0.017802239656448365
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,16,4,balanced,0.01817471981048584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,16,4,balanced,0.017640960216522217
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,16,4,balanced,0.017172479629516603
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,16,4,balanced,0.017240320444107053
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,16,4,balanced,0.017111040353775024
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,16,4,balanced,0.02022655963897705
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,16,4,balanced,0.0219596791267395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,16,4,balanced,0.019453439712524414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,16,4,balanced,0.02377216100692749
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,16,4,power_law_1.01,0.026438400745391843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,16,4,power_law_1.01,0.02803584098815918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,16,4,power_law_1.01,0.048126721382141115
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,16,4,power_law_1.01,0.04687488079071045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,16,4,power_law_1.01,0.04976255893707275
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,16,4,balanced,0.032104959487915044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,16,4,power_law_1.01,0.05143551826477051
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,16,4,power_law_1.01,0.05258368015289307
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,16,4,power_law_1.01,0.053606400489807124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,16,4,power_law_1.01,0.05558144092559815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,16,4,power_law_1.01,0.05588736057281494
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,16,4,power_law_1.01,0.05714687824249267
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,16,4,power_law_1.01,0.059932160377502444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,16,4,power_law_1.01,0.06168064117431641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,16,4,power_law_1.01,0.07095680236816407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,16,4,power_law_1.01,0.0807795238494873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,16,4,power_law_1.01,0.0813145637512207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,16,4,power_law_1.01,0.08821760177612305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,16,4,power_law_1.01,0.10665472030639649
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,16,4,power_law_1.01,0.09875840187072754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,16,4,power_law_1.01,0.1496448040008545
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,16,4,balanced,0.035903999805450434
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,16,4,balanced,0.05101568222045898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,16,4,power_law_1.01,0.19422208786010742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,16,4,power_law_1.01,0.2535500717163086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,16,4,power_law_1.01,0.41406337738037113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,16,4,power_law_1.01,0.33151615142822266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,16,4,power_law_1.01,0.02886399984359741
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,16,4,power_law_1.01,0.030154240131378175
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,16,4,power_law_1.01,0.03710592031478882
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,16,4,power_law_1.01,0.037084159851074217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,16,4,power_law_1.01,0.0369868803024292
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,16,4,power_law_1.01,0.03748608112335205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,16,4,power_law_1.01,0.0378982400894165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,16,4,power_law_1.01,0.0378982400894165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,16,4,power_law_1.01,0.03868544101715088
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,16,4,power_law_1.01,0.03870464086532593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,16,4,power_law_1.01,0.5715456008911133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,16,4,power_law_1.01,0.039930880069732666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,16,4,power_law_1.01,0.04027775764465332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,16,4,power_law_1.01,0.04164095878601074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,16,4,power_law_1.01,0.04308479785919189
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,16,4,balanced,0.08233599662780762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,16,4,power_law_1.01,0.04611839771270752
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,16,4,balanced,0.06035840034484863
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,16,4,balanced,0.10472960472106933
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,16,4,power_law_1.01,0.05255167961120606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,16,4,power_law_1.01,0.046428160667419435
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,16,4,power_law_1.01,0.058928642272949215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,16,4,power_law_1.01,0.0452620792388916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,16,4,power_law_1.01,0.12173952102661131
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,16,4,power_law_1.01,0.07328767776489258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,16,4,power_law_1.01,0.08831359863281249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,16,4,power_law_1.01,0.22794240951538086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,16,4,power_law_1.01,0.016576000452041627
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,16,4,power_law_1.01,0.01738368034362793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,16,4,power_law_1.01,0.017038079500198363
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,16,4,power_law_1.01,0.01706496000289917
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,16,4,power_law_1.01,0.017431039810180664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,16,4,power_law_1.01,0.017223680019378663
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,16,4,power_law_1.01,0.40447105407714845
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,16,4,power_law_1.01,0.01739392042160034
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,16,4,power_law_1.01,0.01739135980606079
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,8,8,power_law_1.2,0.26110591888427737
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,16,4,power_law_1.01,0.01742591977119446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,16,4,power_law_1.01,0.16732927322387697
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,16,4,power_law_1.01,0.01902591943740845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,16,4,power_law_1.01,0.3067110443115234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,16,4,power_law_1.01,0.01902591943740845
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,16,4,power_law_1.01,0.01943935990333557
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,16,4,power_law_1.01,0.8295091247558594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,16,4,power_law_1.01,1.0898675537109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,16,4,power_law_1.01,0.5972224044799804
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,16,4,power_law_1.01,0.017420799732208253
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,16,4,power_law_1.01,0.0176256000995636
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,16,4,power_law_1.01,0.01861376047134399
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,16,4,power_law_1.01,0.02026623964309692
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,16,4,power_law_1.01,0.021559040546417236
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,16,4,power_law_1.01,0.022767360210418704
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,16,4,power_law_1.01,0.025607678890228268
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,16,4,power_law_1.01,0.038297600746154785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,16,4,power_law_1.2,0.025968639850616454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,16,4,power_law_1.2,0.027955200672149662
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,16,4,power_law_1.2,0.04742656230926513
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,16,4,power_law_1.2,0.04852992057800293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,16,4,power_law_1.2,0.04969215869903564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,16,4,power_law_1.2,0.05134848117828369
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,16,4,power_law_1.2,0.052980480194091795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,16,4,power_law_1.2,0.05387648105621338
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,16,4,power_law_1.2,0.05712895870208741
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,16,4,power_law_1.2,0.05746816158294678
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,16,4,power_law_1.2,0.060275201797485356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,16,4,power_law_1.2,0.06225152015686035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,16,4,power_law_1.2,0.06221183776855469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,16,4,power_law_1.2,0.07048448085784911
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,16,4,power_law_1.2,0.07697792053222656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,16,4,power_law_1.2,0.08499839782714844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,16,4,power_law_1.01,0.044432640075683594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,16,4,power_law_1.2,0.09069952011108398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,16,4,power_law_1.2,0.0989900779724121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,16,4,power_law_1.2,0.1115135955810547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,16,4,power_law_1.01,0.07382016181945801
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,16,4,power_law_1.01,0.06072703838348389
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,16,4,power_law_1.2,0.14834943771362305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,16,4,power_law_1.01,0.09182080268859863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,16,4,power_law_1.2,0.19682559967041016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,16,4,power_law_1.2,0.26450944900512696
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,16,4,power_law_1.01,0.12129535675048828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,16,4,power_law_1.2,0.028064000606536865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,16,4,power_law_1.2,0.030104320049285888
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,16,4,power_law_1.2,0.31508096694946286
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,16,4,power_law_1.2,0.03641983985900879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,16,4,power_law_1.2,0.036942079067230224
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,16,4,power_law_1.2,0.037079041004180906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,16,4,power_law_1.2,0.03733247995376587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,16,4,power_law_1.2,0.03747328042984009
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,16,4,power_law_1.2,0.03829247951507568
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,16,4,power_law_1.2,0.03837183952331543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,16,4,power_law_1.2,0.039083518981933595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,16,4,power_law_1.2,0.03995520114898682
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,16,4,power_law_1.2,0.04108416080474854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,16,4,power_law_1.2,0.041625599861145016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,16,4,power_law_1.2,0.0449241590499878
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,16,4,power_law_1.2,0.043543038368225095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,16,4,power_law_1.2,0.046881279945373534
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,16,4,power_law_1.2,0.04774144172668457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,16,4,power_law_1.2,0.05269120216369629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,16,4,power_law_1.2,0.06115839958190918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,16,4,power_law_1.2,0.0767513608932495
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,16,4,power_law_1.2,0.0932249641418457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,16,4,power_law_1.2,0.12886528015136717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,16,4,power_law_1.2,0.44947456359863286
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,16,4,power_law_1.2,0.5857113647460938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,16,4,power_law_1.2,0.1708403205871582
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,16,4,power_law_1.01,0.16878976821899414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,16,4,power_law_1.2,0.01698559999465942
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,16,4,power_law_1.2,0.016993279457092288
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,16,4,power_law_1.2,0.016972800493240355
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,16,4,power_law_1.2,0.016989439725875854
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,16,4,power_law_1.2,0.017379839420318604
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,16,4,power_law_1.2,0.016999679803848266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,16,4,power_law_1.2,0.01740671992301941
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,16,4,power_law_1.2,0.017388800382614138
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,16,4,power_law_1.2,0.017448960542678832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,16,4,power_law_1.2,0.017408000230789183
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,16,4,power_law_1.2,0.01771775960922241
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,16,4,power_law_1.2,0.23876352310180665
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,16,4,power_law_1.01,0.25268480300903323
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,16,4,power_law_1.2,0.31147647857666017
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,16,4,power_law_1.2,0.8811122894287109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,16,4,power_law_1.2,0.5690687942504883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,16,4,power_law_1.2,1.173091812133789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,16,4,power_law_1.2,0.018592000007629395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,16,4,power_law_1.2,0.0186572802066803
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,16,4,power_law_1.2,0.4467814254760742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,16,4,power_law_1.2,0.019397120475769043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,16,4,power_law_1.2,0.043055357933044436
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,16,4,power_law_1.2,0.019423999786376954
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,16,4,power_law_1.2,0.020108799934387207
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,16,4,power_law_1.2,0.02182528018951416
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,16,4,power_law_1.2,0.023119359016418456
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,16,4,power_law_1.2,0.02899199962615967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,32,2,balanced,0.024463360309600828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,32,2,balanced,0.026835200786590574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,32,2,balanced,0.028102400302886965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,32,2,balanced,0.047528958320617674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,32,2,balanced,0.04769152164459228
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,32,2,balanced,0.04851840019226074
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,16,4,power_law_1.2,0.03869951963424682
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,32,2,balanced,0.049340162277221676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,32,2,balanced,0.04972671985626221
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,32,2,balanced,0.0518720006942749
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,32,2,balanced,0.052643837928771975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,32,2,balanced,0.05348864078521729
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,32,2,balanced,0.054675202369689944
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,32,2,balanced,0.05547776222229004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,32,2,balanced,0.05834623813629151
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,32,2,balanced,0.0698367977142334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,32,2,balanced,0.06827648162841797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,32,2,balanced,0.06459136009216308
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,32,2,balanced,0.08727552413940429
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,32,2,balanced,0.09814271926879883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,32,2,balanced,0.11443327903747558
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,32,2,balanced,0.13024127960205079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,32,2,balanced,0.16354047775268554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,32,2,balanced,0.20658559799194337
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,32,2,balanced,0.31196672439575196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,32,2,balanced,0.38851840972900387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,32,2,balanced,0.5693056106567382
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,32,2,balanced,0.02676991939544678
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,32,2,balanced,0.028894720077514646
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,32,2,balanced,0.74893310546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,32,2,balanced,0.035829761028289796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,32,2,balanced,0.035723519325256345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,32,2,balanced,0.03623680114746094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,32,2,balanced,0.03667072057723999
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,32,2,balanced,0.0367193603515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,32,2,balanced,0.03704832077026367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,32,2,balanced,0.03745023965835571
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,32,2,balanced,0.03787264108657837
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,32,2,balanced,0.03873791933059693
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,32,2,balanced,0.03951616048812866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,32,2,balanced,0.041976318359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,32,2,balanced,0.04445312023162842
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,32,2,balanced,0.04610176086425781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,32,2,balanced,0.04366335868835449
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,32,2,balanced,0.03053056001663208
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,16,4,power_law_1.2,0.06313087940216064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,32,2,balanced,0.052236800193786626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,32,2,balanced,0.05997439861297608
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,32,2,balanced,0.06976255893707276
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,32,2,balanced,0.0825216007232666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,32,2,balanced,0.10758144378662109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,32,2,balanced,0.13394175529479982
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,32,2,balanced,0.2022092819213867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,32,2,balanced,0.25610368728637695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,32,2,balanced,0.38082046508789064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,32,2,balanced,0.4979980850219727
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,16,4,power_law_1.2,0.06869120121002197
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,16,4,power_law_1.2,0.09421567916870117
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,16,4,power_law_1.2,0.12528512001037598
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,32,2,balanced,0.017436159849166872
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,32,2,balanced,0.017826559543609618
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,32,2,balanced,0.01814272046089172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,32,2,balanced,0.018314239978790284
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,32,2,balanced,0.01812351942062378
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,32,2,balanced,0.017418240308761594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,32,2,balanced,0.0183513605594635
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,32,2,balanced,0.01864959955215454
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,32,2,balanced,0.01860095977783203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,32,2,balanced,0.019052799940109253
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,32,2,balanced,0.018636800050735473
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,32,2,balanced,0.019418879747390747
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,32,2,balanced,0.01911679983139038
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,32,2,balanced,0.019633920192718507
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,32,2,balanced,0.018202879428863526
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,16,4,power_law_1.2,0.1818956756591797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,32,2,balanced,0.017812479734420777
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,32,2,balanced,0.01815808057785034
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,32,2,balanced,0.022723839282989503
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,32,2,balanced,0.02594815969467163
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,32,2,balanced,0.02230655908584595
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,32,2,balanced,0.028849918842315674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,32,2,power_law_1.01,0.024665598869323728
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,32,2,power_law_1.01,0.02687488079071045
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,32,2,balanced,0.04114687919616699
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,32,2,power_law_1.01,0.028025600910186764
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,32,2,power_law_1.01,0.04731008052825928
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,32,2,power_law_1.01,0.04818816184997558
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,32,2,power_law_1.01,0.04899968147277832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,32,2,power_law_1.01,0.05061376094818115
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,32,2,power_law_1.01,0.05149184226989746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,32,2,power_law_1.01,0.05328000068664551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,32,2,power_law_1.01,0.054133758544921876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,32,2,power_law_1.01,0.05467904090881348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,32,2,power_law_1.01,0.05713151931762696
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,32,2,balanced,0.04857215881347656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,32,2,power_law_1.01,0.057578239440917965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,32,2,power_law_1.01,0.06777728080749512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,32,2,power_law_1.01,0.07364863872528075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,32,2,power_law_1.01,0.08289664268493653
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,32,2,power_law_1.01,0.09054719924926757
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,32,2,power_law_1.01,0.10276608467102051
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,32,2,power_law_1.01,0.06457856178283691
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,32,2,power_law_1.01,0.12435456275939942
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,32,2,power_law_1.01,0.1478105640411377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,32,2,power_law_1.01,0.19125888824462892
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,32,2,power_law_1.01,0.2546316719055176
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,32,2,balanced,0.067292160987854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,32,2,power_law_1.01,0.3486617660522461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,32,2,power_law_1.01,0.6626316833496093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,32,2,power_law_1.01,0.028935680389404295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,32,2,power_law_1.01,0.029245440959930417
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,32,2,power_law_1.01,0.4477222442626953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,16,4,power_law_1.2,0.22007423400878906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,32,2,balanced,0.08156671524047851
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,32,2,balanced,0.11108608245849609
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,32,2,balanced,0.14338303565979005
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,32,2,power_law_1.01,0.02721280097961426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,32,2,power_law_1.01,0.03580159902572631
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,32,2,power_law_1.01,0.038684160709381105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,32,2,power_law_1.01,0.03991935968399048
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,32,2,power_law_1.01,0.03618815898895263
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,32,2,power_law_1.01,0.036307199001312254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,32,2,power_law_1.01,0.04240896224975586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,32,2,power_law_1.01,0.04525951862335205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,32,2,power_law_1.01,0.037095680236816406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,32,2,power_law_1.01,0.047642879486083985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,32,2,power_law_1.01,0.046291198730468754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,32,2,power_law_1.01,0.05568511962890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,32,2,power_law_1.01,0.059828481674194335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,32,2,power_law_1.01,0.040320000648498534
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,32,2,power_law_1.01,0.03659264087677002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,32,2,power_law_1.01,0.037194240093231204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,32,2,power_law_1.01,0.03790848016738892
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,32,2,power_law_1.01,0.07440383911132813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,32,2,power_law_1.01,0.08604288101196289
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,32,2,power_law_1.01,0.017418240308761594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,32,2,power_law_1.01,0.017776639461517335
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,32,2,power_law_1.01,0.017815040349960325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,32,2,power_law_1.01,0.11936896324157716
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,32,2,power_law_1.01,0.018179839849472045
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,32,2,power_law_1.01,0.01781888008117676
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,32,2,power_law_1.01,0.018242559432983398
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,32,2,power_law_1.01,0.018209279775619508
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,32,2,power_law_1.01,0.018576639890670776
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,32,2,power_law_1.01,0.01860607981681824
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,32,2,power_law_1.01,0.019029760360717775
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,32,2,power_law_1.01,0.16277759552001952
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,32,2,power_law_1.01,0.22604415893554686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,32,2,power_law_1.01,0.2890777587890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,32,2,power_law_1.01,0.5606975936889649
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,32,2,power_law_1.01,0.8127820587158203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,32,2,power_law_1.01,0.4106201553344727
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,32,2,power_law_1.01,0.01814272046089172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,32,2,power_law_1.01,0.019880959987640383
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,32,2,power_law_1.01,0.020273919105529788
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,32,2,power_law_1.01,0.02064255952835083
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,32,2,power_law_1.01,0.021359360218048094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,32,2,power_law_1.01,0.021524479389190672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,32,2,power_law_1.01,0.07096831798553467
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,32,2,power_law_1.01,0.02277888059616089
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,32,2,power_law_1.01,0.025278720855712887
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,32,2,power_law_1.01,0.030617599487304685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,32,2,power_law_1.2,0.024148480892181395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,32,2,power_law_1.2,0.026414079666137697
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,32,2,power_law_1.2,0.02805248022079468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,32,2,power_law_1.2,0.047306241989135744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,32,2,power_law_1.2,0.048126721382141115
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,32,2,power_law_1.2,0.04935808181762695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,32,2,power_law_1.2,0.050220799446105954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,32,2,power_law_1.2,0.05140351772308349
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,32,2,power_law_1.2,0.05341440200805664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,32,2,power_law_1.2,0.054214401245117186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,32,2,power_law_1.2,0.05507711887359619
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,32,2,power_law_1.2,0.05624959945678711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,32,2,power_law_1.2,0.057912321090698236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,32,2,power_law_1.2,0.06705535888671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,32,2,power_law_1.2,0.07035007953643799
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,32,2,power_law_1.2,0.07748608112335205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,32,2,power_law_1.2,0.08032256126403808
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,32,2,power_law_1.2,0.09346943855285644
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,32,2,power_law_1.01,0.039772160053253174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,32,2,power_law_1.2,0.10233087539672851
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,32,2,power_law_1.01,0.04493440151214599
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,32,2,power_law_1.2,0.12034048080444334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,32,2,power_law_1.2,0.1853388786315918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,32,2,power_law_1.01,0.059676160812377935
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,32,2,power_law_1.2,0.14926719665527344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,32,2,power_law_1.2,0.24265087127685545
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,32,2,power_law_1.01,0.12091520309448242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,32,2,power_law_1.2,0.02720383882522583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,32,2,power_law_1.2,0.029273600578308107
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,32,2,power_law_1.2,0.02924288034439087
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,32,2,power_law_1.2,0.03598592042922973
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,32,2,power_law_1.2,0.03582463979721069
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,32,2,power_law_1.2,0.03634943962097168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,32,2,power_law_1.2,0.03665663957595825
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,32,2,power_law_1.01,0.10003840446472168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,32,2,power_law_1.2,0.03688832044601441
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,32,2,power_law_1.2,0.47071617126464843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,32,2,power_law_1.2,0.0374835205078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,32,2,power_law_1.2,0.03784832000732422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,32,2,power_law_1.2,0.03870208024978637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,32,2,power_law_1.2,0.03960063934326172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,32,2,power_law_1.2,0.0407206392288208
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,32,2,power_law_1.2,0.04280447959899902
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,32,2,power_law_1.2,0.0454310417175293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,32,2,power_law_1.2,0.047298560142517085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,32,2,power_law_1.2,0.046534399986267086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,32,2,power_law_1.2,0.053498878479003906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,32,2,power_law_1.2,0.06020736217498779
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,32,2,power_law_1.2,0.07474048137664795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,32,2,power_law_1.2,0.08696831703186035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,32,2,power_law_1.2,0.33598209381103517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,32,2,power_law_1.2,0.11310848236083984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,32,2,power_law_1.2,0.6647360229492187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,32,2,power_law_1.2,0.1597439956665039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,32,2,power_law_1.01,0.1883340835571289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,32,2,power_law_1.2,0.22353023529052735
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,32,2,power_law_1.2,0.01734655976295471
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,32,2,power_law_1.2,0.017876479625701904
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,32,2,power_law_1.2,0.018151680231094362
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,32,2,power_law_1.2,0.018237440586090087
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,32,2,power_law_1.2,0.017757439613342287
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,32,2,power_law_1.2,0.0182259202003479
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,32,2,power_law_1.2,0.018402559757232668
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,32,2,power_law_1.2,0.018602240085601806
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,32,2,power_law_1.2,0.018252799510955812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,32,2,power_law_1.2,0.018648320436477663
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,32,2,power_law_1.2,0.02024960041046143
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,32,2,power_law_1.2,0.02059135913848877
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,32,2,power_law_1.2,0.021079039573669432
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,32,2,power_law_1.2,0.02109951972961426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,32,2,power_law_1.2,0.27948415756225586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,32,2,power_law_1.2,0.9058560180664064
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,32,2,power_law_1.01,0.2124736022949219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,32,2,power_law_1.2,0.017369600534439086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,32,2,power_law_1.2,0.40496257781982425
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,32,2,power_law_1.2,0.02020479917526245
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,32,2,power_law_1.2,0.022750720977783204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,32,2,power_law_1.2,0.5588185501098633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,32,2,power_law_1.2,0.025800960063934324
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,32,2,power_law_1.2,0.02920959949493408
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,32,2,power_law_1.2,0.0354150390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,8,8,balanced,0.050346240997314454
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,32,2,power_law_1.2,0.10942208290100097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,8,8,balanced,0.03753727912902832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,8,8,balanced,0.03713920116424561
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,8,8,balanced,0.06659200191497802
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,8,8,balanced,0.07711872100830078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,8,8,balanced,0.08220671653747559
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,8,8,balanced,0.07319295883178711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,8,8,balanced,0.08576895713806151
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,8,8,balanced,0.07809408187866211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,8,8,balanced,0.08047360420227051
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,32,2,power_law_1.2,0.04567552089691162
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,8,8,balanced,0.09170047760009765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,8,8,balanced,0.07890431880950928
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,8,8,balanced,0.07966976165771485
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,32,2,power_law_1.2,0.058768639564514155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,8,8,balanced,0.09155967712402344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,8,8,balanced,0.11157888412475585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,8,8,balanced,0.09733759880065918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,8,8,balanced,0.11427712440490723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,8,8,balanced,0.13084799766540528
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,8,8,balanced,0.13735039710998537
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,8,8,balanced,0.1607244873046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,8,8,balanced,0.14731648445129394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,8,8,balanced,0.18978687286376955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,8,8,balanced,0.2356518363952637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,8,8,balanced,0.31964544296264646
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,8,8,balanced,0.3955507278442383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,8,8,balanced,0.03543936014175415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,8,8,balanced,0.035092480182647705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,8,8,balanced,0.0354419207572937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,8,8,balanced,0.045690879821777344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,8,8,balanced,0.582817268371582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,8,8,balanced,0.04606207847595215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,8,8,balanced,0.04612095832824707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,8,8,balanced,0.046496000289916996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,8,8,balanced,0.7641548919677734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,8,8,balanced,0.04631552219390869
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,8,8,balanced,0.04640511989593506
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,8,8,balanced,0.04643328189849853
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,8,8,balanced,0.046487040519714355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,8,8,balanced,0.04688000202178955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,8,8,balanced,0.04644224166870117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,8,8,balanced,0.047294721603393555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,8,8,balanced,0.04739840030670166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,8,8,balanced,0.04970240116119385
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,8,8,balanced,0.052250881195068356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,8,8,balanced,0.05522304058074952
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,8,8,balanced,0.06869887828826904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,8,8,balanced,0.06223616123199462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,8,8,balanced,0.08097663879394532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,8,8,balanced,0.0937177562713623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,8,8,balanced,0.1329049587249756
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,8,8,balanced,0.15645824432373048
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,8,8,balanced,0.22497791290283203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,8,8,balanced,0.3020569610595703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,8,8,balanced,0.04768383979797363
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,32,2,power_law_1.2,0.08128512382507325
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,32,2,power_law_1.2,0.12919551849365235
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,32,2,power_law_1.2,0.19691648483276367
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,8,8,balanced,0.01880704045295715
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,8,8,balanced,0.01850111961364746
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,8,8,balanced,0.0186188805103302
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,8,8,balanced,0.01866752028465271
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,8,8,balanced,0.018615039587020873
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,8,8,balanced,0.019068160057067872
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,8,8,balanced,0.01862015962600708
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,8,8,balanced,0.01860479950904846
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,8,8,balanced,0.01860607981681824
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,8,8,balanced,0.019028480052948
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,8,8,balanced,0.019032319784164427
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,8,8,balanced,0.01908095955848694
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,8,8,balanced,0.018250240087509157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,8,8,balanced,0.01859071969985962
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,8,8,balanced,0.018584320545196535
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,8,8,balanced,0.018570239543914797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,32,2,power_law_1.2,0.17417856216430666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,8,8,balanced,0.018663680553436278
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,8,8,balanced,0.020684800148010253
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,8,8,balanced,0.021077759265899658
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,8,8,balanced,0.022623999118804933
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,8,8,power_law_1.01,0.03821055889129639
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,8,8,power_law_1.01,0.09403903961181641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,8,8,power_law_1.01,0.06689151763916015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,8,8,balanced,0.025579519271850586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,8,8,power_law_1.01,0.06849279880523682
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,8,8,power_law_1.01,0.08268159866333008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,8,8,power_law_1.01,0.08879232406616211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,8,8,power_law_1.01,0.10417152404785157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,8,8,power_law_1.01,0.1006668758392334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,8,8,power_law_1.01,0.10957183837890624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,8,8,power_law_1.01,0.11186431884765624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,8,8,power_law_1.01,0.11525631904602052
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,8,8,power_law_1.01,0.12185855865478516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,8,8,power_law_1.01,0.12402048110961914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,8,8,power_law_1.01,0.1289356803894043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,8,8,power_law_1.01,0.1390771198272705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,8,8,power_law_1.01,0.1475443172454834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,8,8,power_law_1.01,0.17228927612304687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,8,8,power_law_1.01,0.14915072441101074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,8,8,power_law_1.01,0.2079884719848633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,8,8,power_law_1.01,0.3083558464050293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,8,8,balanced,0.032576000690460204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,8,8,power_law_1.01,0.49521537780761715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,8,8,power_law_1.01,0.38235008239746093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,8,8,balanced,0.036988160610198974
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,8,8,power_law_1.01,1.2689421081542969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,8,8,power_law_1.01,0.7394649505615234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,8,8,power_law_1.01,0.035050239562988285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,8,8,balanced,0.0520627212524414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,8,8,power_law_1.01,0.045204482078552245
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,8,8,power_law_1.01,0.04563839912414551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,8,8,power_law_1.01,0.045630722045898436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,8,8,power_law_1.01,0.0464793586730957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,8,8,power_law_1.01,0.04646016120910644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,8,8,power_law_1.01,0.04687104225158691
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,8,8,power_law_1.01,0.046487040519714355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,8,8,power_law_1.01,0.04687488079071045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,8,8,power_law_1.01,0.04726655960083008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,8,8,power_law_1.01,0.04816768169403076
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,8,8,power_law_1.01,0.04971648216247558
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,8,8,power_law_1.01,0.0507212781906128
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,8,8,power_law_1.01,0.053840641975402836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,8,8,power_law_1.01,1.0964864349365233
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,8,8,power_law_1.01,0.05647999763488769
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,8,8,power_law_1.01,0.06000895977020264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,8,8,power_law_1.01,0.06265727996826172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,8,8,power_law_1.01,0.07783423900604249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,8,8,power_law_1.01,0.08808447837829589
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,8,8,power_law_1.01,0.1953727912902832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,8,8,power_law_1.01,0.12623231887817382
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,8,8,power_law_1.01,2.1172390747070313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,8,8,power_law_1.01,0.4841587066650391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,8,8,power_law_1.01,0.15091584205627442
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,8,8,power_law_1.01,0.3878412628173828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,8,8,power_law_1.01,0.018186240196228026
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,8,8,balanced,0.06731008052825928
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,8,8,power_law_1.01,0.01857792019844055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,8,8,power_law_1.01,0.28206592559814453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,8,8,power_law_1.01,0.018677760362625122
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,8,8,power_law_1.01,0.018247679471969605
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,8,8,power_law_1.01,0.018551039695739745
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,8,8,power_law_1.01,0.0189683198928833
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,8,8,power_law_1.01,0.0187007999420166
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,8,8,power_law_1.01,0.01857408046722412
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,8,8,power_law_1.01,0.018969600200653074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,8,8,power_law_1.01,2.5670874023437498
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,8,8,power_law_1.01,0.01869824051856995
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,8,8,power_law_1.01,0.0189900803565979
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,8,8,balanced,0.09302528381347656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,8,8,balanced,0.11394304275512694
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,8,8,power_law_1.01,0.77351806640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,8,8,power_law_1.01,1.0822118377685546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,8,8,power_law_1.01,0.019536639451980593
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,8,8,power_law_1.01,0.020334079265594482
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,8,8,power_law_1.01,0.020725760459899902
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,8,8,power_law_1.01,0.02150144100189209
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,8,8,power_law_1.01,0.02196991920471191
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,8,8,power_law_1.01,0.02355072021484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,8,8,power_law_1.01,0.06101632118225098
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,8,8,power_law_1.01,0.03224191904067993
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,8,8,power_law_1.01,0.034990079402923584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,8,8,power_law_1.2,0.037920000553131106
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,8,8,power_law_1.2,0.07162752151489257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,8,8,power_law_1.2,0.06777344226837159
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,8,8,power_law_1.2,0.0682419204711914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,8,8,power_law_1.2,0.07065343856811523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,8,8,power_law_1.2,0.07515647888183594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,8,8,power_law_1.2,0.09602304458618163
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,8,8,power_law_1.2,0.10013695716857911
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,8,8,power_law_1.2,0.10805888175964355
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,8,8,power_law_1.01,0.044120321273803706
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,8,8,power_law_1.2,0.1092262363433838
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,8,8,power_law_1.2,0.12294400215148926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,8,8,power_law_1.2,0.11750783920288085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,8,8,power_law_1.2,0.1258790397644043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,8,8,power_law_1.2,0.12931712150573732
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,8,8,power_law_1.2,0.14241024017333986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,8,8,power_law_1.2,0.15164159774780273
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,8,8,power_law_1.01,0.07315072059631347
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,8,8,power_law_1.2,0.14926591873168946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,8,8,power_law_1.2,0.1863871955871582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,8,8,power_law_1.2,0.21810688018798827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,8,8,power_law_1.2,0.32105857849121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,8,8,power_law_1.2,0.35286911010742184
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,8,8,power_law_1.01,0.1017574405670166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,8,8,power_law_1.2,0.5519308853149414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,8,8,power_law_1.2,0.6713254547119141
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,8,8,power_law_1.01,0.1355519962310791
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,8,8,power_law_1.2,0.0359116792678833
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,8,8,power_law_1.2,0.04606207847595215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,8,8,power_law_1.2,0.046474242210388185
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,8,8,power_law_1.2,0.04649472236633301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,8,8,power_law_1.2,0.9978150177001954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,8,8,power_law_1.2,0.04648320198059082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,8,8,power_law_1.2,0.0465715217590332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,8,8,power_law_1.2,0.046741762161254884
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,8,8,power_law_1.2,0.04688384056091309
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,8,8,power_law_1.2,0.047467517852783206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,8,8,power_law_1.2,0.047527680397033686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,8,8,power_law_1.2,0.04903808116912842
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,8,8,power_law_1.2,0.05014783859252929
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,8,8,power_law_1.2,0.05182591915130615
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,8,8,power_law_1.2,0.05423359870910645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,8,8,power_law_1.2,0.056272640228271484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,8,8,power_law_1.2,0.060787200927734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,8,8,power_law_1.01,0.29332223892211917
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,8,8,power_law_1.2,0.06276991844177246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,8,8,power_law_1.2,0.0783129596710205
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,8,8,power_law_1.01,0.18961664199829104
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,8,8,power_law_1.2,0.08837120056152344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,8,8,power_law_1.2,0.13373056411743164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,8,8,power_law_1.2,0.13241087913513183
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,8,8,power_law_1.2,0.21519487380981445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,8,8,power_law_1.2,1.4874879455566405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,8,8,power_law_1.2,0.288145923614502
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,8,8,power_law_1.2,0.3877811050415039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,8,8,power_law_1.2,0.0184115195274353
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,8,8,power_law_1.2,0.018220800161361694
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,8,8,power_law_1.2,0.01864575982093811
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,8,8,power_law_1.2,0.018622720241546632
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,8,8,power_law_1.2,0.01860479950904846
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,8,8,power_law_1.2,0.018856960535049438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,8,8,power_law_1.2,0.018867199420928953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,8,8,power_law_1.2,0.019048960208892823
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,8,8,power_law_1.2,0.01875712037086487
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,8,8,power_law_1.2,0.01906048059463501
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,8,8,power_law_1.2,0.01990272045135498
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,8,8,power_law_1.2,0.020290560722351074
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,8,8,power_law_1.2,0.020648961067199705
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,8,8,power_law_1.01,0.39262847900390624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,8,8,power_law_1.2,0.5306611251831055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,8,8,power_law_1.2,0.7627366638183594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,8,8,power_law_1.2,2.2264256286621094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,8,8,power_law_1.2,2.9884698486328127
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,8,8,power_law_1.2,0.0513651180267334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,8,8,power_law_1.2,0.018201600313186645
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,8,8,power_law_1.2,0.021235198974609376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,8,8,power_law_1.2,1.1034381103515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,8,8,power_law_1.2,0.02193536043167114
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,8,8,power_law_1.2,0.10675456047058104
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,8,8,power_law_1.2,0.025053439140319822
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,8,8,power_law_1.2,0.03305984020233154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,16,4,balanced,0.03175168037414551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,16,4,balanced,0.031390719413757324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,16,4,balanced,0.03575295925140381
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,16,4,balanced,0.0826803207397461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,16,4,balanced,0.08261631965637208
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,16,4,balanced,0.083438081741333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,16,4,balanced,0.08584575653076172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,8,8,power_law_1.2,0.03625344038009644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,16,4,balanced,0.08530048370361329
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,16,4,balanced,0.07112703800201417
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,16,4,balanced,0.08372735977172852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,16,4,balanced,0.08773759841918946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,16,4,balanced,0.0866483211517334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,16,4,balanced,0.10064895629882813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,16,4,balanced,0.10042880058288575
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,16,4,balanced,0.10790528297424315
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,16,4,balanced,0.10795519828796388
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,16,4,balanced,0.11884672164916991
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,16,4,balanced,0.138787841796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,16,4,balanced,0.1289971160888672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,16,4,balanced,0.1460313606262207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,16,4,balanced,0.1616089630126953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,16,4,balanced,0.18916351318359376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,16,4,balanced,0.2514316749572754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,16,4,balanced,0.3593254470825195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,16,4,balanced,0.4542208099365235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,16,4,balanced,0.6677209472656249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,16,4,balanced,0.03338367938995361
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,16,4,balanced,0.03337471961975098
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,16,4,balanced,0.035427839756011964
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,16,4,balanced,0.875832290649414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,16,4,balanced,0.04284671783447266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,16,4,balanced,0.042940158843994145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,16,4,balanced,0.04321407794952393
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,16,4,balanced,0.043600640296936034
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,16,4,balanced,0.043531517982482906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,16,4,balanced,0.04360576152801514
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,16,4,balanced,0.04402431964874268
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,16,4,balanced,0.044014081954956055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,16,4,balanced,0.04417535781860352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,16,4,balanced,0.04448383808135986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,16,4,balanced,0.045678081512451174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,16,4,balanced,0.046530561447143556
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,16,4,balanced,0.04738304138183594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,16,4,balanced,0.05185535907745361
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,16,4,balanced,0.06064511775970459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,16,4,balanced,0.05875584125518799
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,16,4,balanced,0.0703936004638672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,16,4,balanced,0.07974016189575195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,16,4,balanced,0.10088576316833495
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,16,4,balanced,0.12023296356201171
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,16,4,balanced,0.17840896606445314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,16,4,balanced,0.22562047958374026
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,16,4,balanced,0.32483711242675783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,16,4,balanced,0.4279206466674805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,8,8,power_law_1.2,0.06395008087158202
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,8,8,power_law_1.2,0.08968704223632813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,8,8,power_law_1.2,0.17847936630249023
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,8,8,power_law_1.2,0.14537216186523438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,16,4,balanced,0.018650879859924314
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,16,4,balanced,0.019018239974975586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,16,4,balanced,0.01862015962600708
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,16,4,balanced,0.018684159517288208
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,16,4,balanced,0.01861376047134399
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,16,4,balanced,0.019037439823150634
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,16,4,balanced,0.019022079706192015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,16,4,balanced,0.01904512047767639
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,16,4,balanced,0.019088640213012695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,16,4,balanced,0.0195251202583313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,16,4,balanced,0.019850239753723145
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,16,4,balanced,0.019691519737243653
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,16,4,balanced,0.019824639558792115
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,16,4,balanced,0.018617600202560425
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,16,4,balanced,0.018622720241546632
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,8,8,power_law_1.2,0.2953433609008789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,16,4,balanced,0.01863039970397949
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,16,4,balanced,0.018659839630126955
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,8,8,power_law_1.2,0.39394302368164064
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,16,4,balanced,0.023393280506134033
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,16,4,balanced,0.022325758934020994
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,16,4,balanced,0.024796159267425538
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,16,4,power_law_1.01,0.036290559768676754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,16,4,power_law_1.01,0.04510975837707519
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,16,4,power_law_1.01,0.061642241477966306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,16,4,power_law_1.01,0.06249983787536621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,16,4,power_law_1.01,0.07064447879791261
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,16,4,power_law_1.01,0.0719859218597412
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,16,4,power_law_1.01,0.09134079933166503
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,16,4,power_law_1.01,0.08398847579956055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,16,4,power_law_1.01,0.09618304252624513
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,16,4,power_law_1.01,0.10118016242980957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,16,4,power_law_1.01,0.10700544357299804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,16,4,power_law_1.01,0.11066752433776857
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,16,4,power_law_1.01,0.1112665557861328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,16,4,power_law_1.01,0.12249728202819825
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,16,4,power_law_1.01,0.13000063896179198
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,16,4,power_law_1.01,0.13443072319030763
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,16,4,power_law_1.01,0.13558143615722656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,16,4,power_law_1.01,0.1446003246307373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,16,4,power_law_1.01,0.17134592056274414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,16,4,balanced,0.02803071975708008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,16,4,power_law_1.01,0.18728832244873045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,16,4,power_law_1.01,0.2748313522338867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,16,4,balanced,0.04073855876922607
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,16,4,power_law_1.01,0.4101900863647461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,16,4,power_law_1.01,0.4705535888671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,16,4,power_law_1.01,0.8945305633544922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,16,4,power_law_1.01,0.8062834930419921
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,16,4,power_law_1.01,0.032505600452423095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,16,4,power_law_1.01,0.03386111974716187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,16,4,power_law_1.01,0.04278143882751465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,16,4,power_law_1.01,0.0432371187210083
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,16,4,power_law_1.01,0.04320127964019775
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,16,4,power_law_1.01,0.04360960006713867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,16,4,power_law_1.01,0.04367231845855713
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,16,4,balanced,0.04806528091430664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,16,4,power_law_1.01,0.04371327877044678
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,16,4,power_law_1.01,0.04457087993621826
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,16,4,power_law_1.01,0.04448639869689942
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,16,4,power_law_1.01,0.04584703922271728
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,16,4,power_law_1.01,0.04691199779510498
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,16,4,balanced,0.06509439945220948
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,16,4,power_law_1.01,0.048346881866455076
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,16,4,power_law_1.01,0.05122687816619873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,16,4,power_law_1.01,0.05811327934265137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,16,4,power_law_1.01,0.053544960021972655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,16,4,power_law_1.01,0.060729598999023436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,16,4,power_law_1.01,0.0682035207748413
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,16,4,power_law_1.01,0.07960063934326171
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,16,4,balanced,0.08065792083740234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,16,4,power_law_1.01,1.727431640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,16,4,power_law_1.01,0.22445823669433596
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,16,4,power_law_1.01,0.10574463844299316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,16,4,power_law_1.01,0.13164544105529785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,16,4,power_law_1.01,0.18372352600097658
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,16,4,power_law_1.01,0.018163199424743655
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,16,4,power_law_1.01,0.018581759929656983
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,16,4,balanced,0.11115008354187013
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,16,4,balanced,0.1377241611480713
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,16,4,power_law_1.01,0.01857151985168457
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,16,4,power_law_1.01,0.018958079814910888
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,16,4,power_law_1.01,0.01901311993598938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,16,4,power_law_1.01,0.018997759819030763
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,16,4,power_law_1.01,0.018721280097961427
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,16,4,power_law_1.01,0.01911296010017395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,16,4,power_law_1.01,0.01904255986213684
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,16,4,power_law_1.01,0.019443199634552003
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,16,4,power_law_1.01,0.019852800369262697
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,16,4,power_law_1.01,0.02076927900314331
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,16,4,power_law_1.01,0.021087999343872073
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,16,4,power_law_1.01,0.33383552551269535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,16,4,power_law_1.01,1.2164646148681642
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,16,4,power_law_1.01,0.41188095092773436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,16,4,power_law_1.01,0.5773401641845703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,16,4,power_law_1.01,0.894620132446289
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,16,4,power_law_1.01,0.020285439491271973
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,16,4,power_law_1.01,0.022712318897247313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,16,4,power_law_1.01,0.023549439907073973
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,16,4,power_law_1.01,0.024455680847167968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,16,4,power_law_1.01,0.027141120433807377
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,16,4,power_law_1.01,0.03382143974304199
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,16,4,power_law_1.01,0.0514086389541626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,16,4,power_law_1.2,0.040663042068481446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,16,4,power_law_1.2,0.04199808120727539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,16,4,power_law_1.2,0.061043200492858884
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,16,4,power_law_1.2,0.0640934419631958
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,16,4,power_law_1.2,0.06615295886993408
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,16,4,power_law_1.2,0.06875264167785644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,16,4,power_law_1.2,0.089617919921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,16,4,power_law_1.2,0.08846336364746094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,16,4,power_law_1.2,0.0954355239868164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,16,4,power_law_1.2,0.10076031684875489
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,16,4,power_law_1.2,0.1087987232208252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,16,4,power_law_1.2,0.11145343780517578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,16,4,power_law_1.2,0.1118886375427246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,16,4,power_law_1.2,0.12261631965637207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,16,4,power_law_1.2,0.12999168395996094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,16,4,power_law_1.2,0.13517184257507325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,16,4,power_law_1.2,0.13605759620666505
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,16,4,power_law_1.2,0.15274751663208008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,16,4,power_law_1.2,0.16902784347534178
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,16,4,power_law_1.2,0.23227008819580078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,16,4,power_law_1.01,0.05861248016357422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,16,4,power_law_1.2,0.2795302391052246
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,16,4,power_law_1.01,0.07241087913513183
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,16,4,power_law_1.2,0.3803046417236328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,16,4,power_law_1.01,0.09353856086730958
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,16,4,power_law_1.2,0.6706841278076172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,16,4,power_law_1.01,0.12223103523254394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,16,4,power_law_1.2,0.032592639923095704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,16,4,power_law_1.2,0.034206719398498536
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,16,4,power_law_1.2,0.04279935836791992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,16,4,power_law_1.2,0.04318719863891601
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,16,4,power_law_1.2,0.04280831813812256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,16,4,power_law_1.2,0.04358399868011474
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,16,4,power_law_1.2,0.04377471923828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,16,4,power_law_1.2,0.5290982437133789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,16,4,power_law_1.2,0.043819518089294435
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,16,4,power_law_1.2,0.04441088199615478
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,16,4,power_law_1.2,0.044834561347961426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,16,4,power_law_1.2,0.04578944206237793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,16,4,power_law_1.2,0.04689023971557617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,16,4,power_law_1.2,0.04816127777099609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,16,4,power_law_1.2,0.05277567863464355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,16,4,power_law_1.2,0.993776626586914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,16,4,power_law_1.2,0.054351358413696295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,16,4,power_law_1.2,0.0630079984664917
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,16,4,power_law_1.2,0.07058176040649414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,16,4,power_law_1.2,0.060270080566406245
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,16,4,power_law_1.2,0.08054143905639648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,16,4,power_law_1.2,0.10921343803405761
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,16,4,power_law_1.2,0.13317760467529297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,16,4,power_law_1.01,0.24630783081054686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,16,4,power_law_1.2,0.18662656784057616
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,16,4,power_law_1.01,0.15718015670776367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,16,4,power_law_1.2,0.23309568405151365
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,16,4,power_law_1.2,0.5896396636962891
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,16,4,power_law_1.2,0.018650879859924314
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,16,4,power_law_1.2,0.018655999898910522
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,16,4,power_law_1.2,0.3378009414672852
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,16,4,power_law_1.2,0.018622720241546632
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,16,4,power_law_1.2,0.018727680444717405
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,16,4,power_law_1.2,0.018668800592422485
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,16,4,power_law_1.2,0.01897600054740906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,16,4,power_law_1.2,0.01904255986213684
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,16,4,power_law_1.2,0.01929471969604492
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,16,4,power_law_1.2,0.01938688039779663
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,16,4,power_law_1.2,0.019456000328063966
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,16,4,power_law_1.2,0.4667379379272461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,16,4,power_law_1.2,1.3820544433593749
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,16,4,power_law_1.2,0.022476799488067627
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,16,4,power_law_1.2,0.02279423952102661
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,16,4,power_law_1.2,1.7804287719726564
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,16,4,power_law_1.2,0.018625279664993284
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,16,4,power_law_1.01,0.31852415084838864
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,16,4,power_law_1.2,0.7897714996337891
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,16,4,power_law_1.2,0.02025343894958496
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,16,4,power_law_1.2,0.02025984048843384
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,16,4,power_law_1.2,0.021523199081420898
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,16,4,power_law_1.2,0.024586238861083985
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,16,4,power_law_1.2,0.030515201091766357
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,16,4,power_law_1.2,0.034508800506591795
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,16,4,power_law_1.2,0.05234687805175782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,32,2,balanced,0.030478079319000245
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,32,2,balanced,0.03290112018585205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,32,2,balanced,0.035842559337615966
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,32,2,balanced,0.06351232051849365
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,32,2,balanced,0.0653542423248291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,32,2,balanced,0.07157631874084472
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,32,2,balanced,0.0793446397781372
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,32,2,balanced,0.07860991954803467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,32,2,balanced,0.08640512466430664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,32,2,balanced,0.08759039878845215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,32,2,balanced,0.09449728012084961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,32,2,balanced,0.09564800262451172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,32,2,balanced,0.09830911636352539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,32,2,balanced,0.10522751808166504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,32,2,balanced,0.12222975730895995
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,16,4,power_law_1.2,0.06021759986877442
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,32,2,balanced,0.12486528396606444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,32,2,balanced,0.12008831977844239
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,32,2,balanced,0.1331430435180664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,32,2,balanced,0.14508543968200685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,32,2,balanced,0.1674380874633789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,32,2,balanced,0.19140224456787108
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,32,2,balanced,0.2407206344604492
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,16,4,power_law_1.2,0.07466368198394775
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,32,2,balanced,0.3172800064086914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,32,2,balanced,0.4574259185791016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,32,2,balanced,0.5890188980102539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,32,2,balanced,0.03131135940551758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,32,2,balanced,0.03379456043243408
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,32,2,balanced,0.03500927925109863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,32,2,balanced,0.8621810913085938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,32,2,balanced,0.041553921699523926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,32,2,balanced,0.0415667200088501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,32,2,balanced,0.04187520027160645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,32,2,balanced,0.04231167793273925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,32,2,balanced,1.1327347564697265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,32,2,balanced,0.04282112121582031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,32,2,balanced,0.04363903999328613
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,32,2,balanced,0.04397439956665039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,32,2,balanced,0.044441599845886234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,32,2,balanced,0.04649472236633301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,32,2,balanced,0.047674880027771
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,32,2,balanced,0.050579199790954585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,32,2,balanced,0.05421055793762207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,32,2,balanced,0.05500544071197509
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,32,2,balanced,0.05545599937438965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,32,2,balanced,0.06819583892822265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,32,2,balanced,0.07567999839782716
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,32,2,balanced,0.0932249641418457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,32,2,balanced,0.109619197845459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,32,2,balanced,0.1472550392150879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,32,2,balanced,0.1879654312133789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,32,2,balanced,0.2787775993347168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,32,2,balanced,0.3541056060791016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,32,2,balanced,0.5182694244384766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,32,2,balanced,0.6819993591308593
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,16,4,power_law_1.2,0.1019212818145752
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,16,4,power_law_1.2,0.12924799919128419
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,16,4,power_law_1.2,0.173306884765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,32,2,balanced,0.018598400354385376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,32,2,balanced,0.0215065598487854
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,32,2,balanced,0.01900928020477295
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,32,2,balanced,0.021361920833587646
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,32,2,balanced,0.021546239852905276
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,32,2,balanced,0.021483519077301026
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,32,2,balanced,0.021064960956573488
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,32,2,balanced,0.02147968053817749
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,32,2,balanced,0.021505279541015627
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,32,2,balanced,0.02104192018508911
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,32,2,balanced,0.0214899206161499
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,32,2,balanced,0.021551361083984373
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,32,2,balanced,0.021454079151153566
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,32,2,balanced,0.02310271978378296
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,32,2,balanced,0.0209881591796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,32,2,balanced,0.0215116810798645
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,32,2,balanced,0.021495039463043215
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,16,4,power_law_1.2,0.2467276763916016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,32,2,balanced,0.02637696027755737
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,32,2,balanced,0.03148799896240234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,32,2,balanced,0.024808959960937498
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,32,2,balanced,0.05485439777374268
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,32,2,balanced,0.038609919548034666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,32,2,power_law_1.01,0.03049216032028198
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,32,2,power_law_1.01,0.03380608081817627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,32,2,power_law_1.01,0.04157567977905273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,32,2,power_law_1.01,0.06083968162536622
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,32,2,power_law_1.01,0.06368768215179443
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,32,2,power_law_1.01,0.0701145601272583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,32,2,power_law_1.01,0.08267007827758789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,32,2,power_law_1.01,0.0793881607055664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,32,2,power_law_1.01,0.08554495811462402
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,32,2,power_law_1.01,0.08811136245727538
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,32,2,power_law_1.01,0.09806464195251466
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,32,2,power_law_1.01,0.09837696075439453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,32,2,power_law_1.01,0.10105728149414062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,32,2,power_law_1.01,0.11614591598510741
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,32,2,power_law_1.01,0.12361599922180175
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,32,2,power_law_1.01,0.12917247772216797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,32,2,power_law_1.01,0.1269209575653076
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,32,2,power_law_1.01,0.13865983963012696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,32,2,power_law_1.01,0.1507084846496582
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,32,2,balanced,0.06498047828674317
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,32,2,power_law_1.01,0.17731840133666993
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,32,2,balanced,0.08554752349853516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,32,2,power_law_1.01,0.22712703704833986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,32,2,power_law_1.01,0.381267204284668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,32,2,power_law_1.01,0.3158937644958496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,32,2,power_law_1.01,0.5082009506225587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,32,2,power_law_1.01,0.030615038871765136
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,32,2,power_law_1.01,0.033327360153198246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,32,2,power_law_1.01,0.03379199981689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,32,2,power_law_1.01,0.04128511905670166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,32,2,power_law_1.01,0.041436161994934086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,32,2,power_law_1.01,0.041588478088378907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,32,2,power_law_1.01,0.04200831890106201
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,32,2,power_law_1.01,0.04276351928710938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,32,2,power_law_1.01,0.04354432106018066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,32,2,power_law_1.01,0.04407680034637451
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,32,2,power_law_1.01,0.04479616165161133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,32,2,power_law_1.01,0.704958724975586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,32,2,power_law_1.01,0.04684031963348388
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,32,2,power_law_1.01,0.04784768104553223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,32,2,power_law_1.01,0.05181824207305909
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,32,2,power_law_1.01,0.05712768077850342
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,32,2,power_law_1.01,0.05872511863708496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,32,2,power_law_1.01,0.0588595199584961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,32,2,power_law_1.01,0.06721920013427735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,32,2,power_law_1.01,0.07848703861236572
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,32,2,power_law_1.01,0.09970687866210938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,16,4,power_law_1.2,0.31619455337524416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,32,2,power_law_1.01,0.11836799621582031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,32,2,power_law_1.01,0.1957439994812012
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,32,2,power_law_1.01,0.1687539291381836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,32,2,power_law_1.01,1.0623129272460938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,32,2,balanced,0.106178560256958
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,32,2,power_law_1.01,0.01901695966720581
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,32,2,power_law_1.01,0.01938048005104065
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,32,2,balanced,0.15215231895446776
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,32,2,balanced,0.19127424240112306
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,32,2,power_law_1.01,0.01978495955467224
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,32,2,power_law_1.01,0.02082047939300537
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,32,2,power_law_1.01,0.020701439380645753
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,32,2,power_law_1.01,0.020660479068756104
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,32,2,power_law_1.01,0.02108288049697876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,32,2,power_law_1.01,0.02110464096069336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,32,2,power_law_1.01,0.021063680648803713
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,32,2,power_law_1.01,0.02110208034515381
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,32,2,power_law_1.01,0.021164801120758057
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,32,2,power_law_1.01,0.02147200107574463
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,32,2,power_law_1.01,0.022709760665893555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,32,2,power_law_1.01,0.31160064697265627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,32,2,power_law_1.01,1.3497625732421876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,32,2,power_law_1.01,0.38234878540039063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,32,2,power_law_1.01,0.5801472091674805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,32,2,power_law_1.01,0.05177984237670898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,32,2,power_law_1.01,0.7536921691894531
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,32,2,power_law_1.01,0.022261760234832763
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,32,2,power_law_1.01,0.023915519714355467
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,32,2,power_law_1.01,0.024350719451904295
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,32,2,power_law_1.01,0.026668798923492432
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,32,2,power_law_1.01,0.03318399906158447
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,32,2,power_law_1.01,0.035502080917358396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,32,2,power_law_1.2,0.03501183986663818
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,32,2,power_law_1.2,0.03583231925964355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,32,2,power_law_1.2,0.04362112045288086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,32,2,power_law_1.2,0.07932799816131592
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,32,2,power_law_1.2,0.07586944103240967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,32,2,power_law_1.2,0.0786841583251953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,32,2,power_law_1.2,0.0927244758605957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,32,2,power_law_1.2,0.09321599960327148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,32,2,power_law_1.2,0.09641983985900879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,32,2,power_law_1.2,0.0921945571899414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,32,2,power_law_1.2,0.09707776069641114
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,32,2,power_law_1.2,0.10259200096130372
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,32,2,power_law_1.2,0.1083788776397705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,32,2,power_law_1.2,0.11762816429138183
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,32,2,power_law_1.2,0.12382080078124999
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,32,2,power_law_1.2,0.12851200103759766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,32,2,power_law_1.2,0.1246566390991211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,32,2,power_law_1.2,0.14031999588012695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,32,2,power_law_1.2,0.15206527709960938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,32,2,power_law_1.2,0.17985408782958984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,32,2,power_law_1.01,0.05994112014770507
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,32,2,power_law_1.2,0.2213644790649414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,32,2,power_law_1.2,0.2611647987365723
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,32,2,power_law_1.01,0.08249728202819824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,32,2,power_law_1.2,0.36733566284179686
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,32,2,power_law_1.01,0.08797696113586426
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,32,2,power_law_1.01,0.17330175399780273
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,32,2,power_law_1.01,0.1493401622772217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,32,2,power_law_1.2,0.03210752010345459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,32,2,power_law_1.2,0.03293823957443237
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,32,2,power_law_1.2,0.03418751955032349
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,32,2,power_law_1.2,0.041587200164794925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,32,2,power_law_1.2,0.041667838096618656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,32,2,power_law_1.2,0.042351360321044924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,32,2,power_law_1.2,0.04210048198699951
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,32,2,power_law_1.2,0.04257919788360596
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,32,2,power_law_1.2,0.043681278228759765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,32,2,power_law_1.2,0.044778242111206054
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,32,2,power_law_1.2,0.045263361930847165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,32,2,power_law_1.2,0.04730368137359619
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,32,2,power_law_1.2,0.5448793411254883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,32,2,power_law_1.2,0.04865407943725586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,32,2,power_law_1.2,0.05215104103088379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,32,2,power_law_1.2,0.057346558570861815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,32,2,power_law_1.2,0.05885695934295654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,32,2,power_law_1.2,0.06846335887908936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,32,2,power_law_1.2,0.06059648036956787
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,32,2,power_law_1.2,0.07908736228942871
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,32,2,power_law_1.01,0.22546560287475587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,32,2,power_law_1.2,0.7257984161376954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,32,2,power_law_1.2,0.10020735740661621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,32,2,power_law_1.2,0.11804672241210938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,32,2,power_law_1.2,0.16426752090454103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,32,2,power_law_1.2,0.22282623291015624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,32,2,power_law_1.2,0.018859519958496093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,32,2,power_law_1.2,0.019444479942321777
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,32,2,power_law_1.2,0.019471360445022585
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,32,2,power_law_1.2,0.021103360652923585
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,32,2,power_law_1.2,0.020725760459899902
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,32,2,power_law_1.2,0.021112320423126222
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,32,2,power_law_1.2,0.021094400882720948
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,32,2,power_law_1.2,0.021073920726776125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,32,2,power_law_1.2,0.021107199192047118
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,32,2,power_law_1.2,0.0210752010345459
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,32,2,power_law_1.2,0.021249279975891114
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,32,2,power_law_1.2,0.30103679656982424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,32,2,power_law_1.2,0.3764339065551758
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,32,2,power_law_1.01,0.3179251289367676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,32,2,power_law_1.2,1.0438771057128906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,32,2,power_law_1.2,0.5913177490234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,32,2,power_law_1.2,1.312762908935547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,32,2,power_law_1.2,0.021907200813293455
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,32,2,power_law_1.2,0.02188800096511841
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,32,2,power_law_1.2,0.022735359668731688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,32,2,power_law_1.2,0.057937917709350584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,32,2,power_law_1.2,0.7485017395019531
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,32,2,power_law_1.2,0.024353280067443847
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,32,2,power_law_1.2,0.024802560806274417
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,32,2,power_law_1.2,0.09570560455322266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,32,2,power_law_1.2,0.02593024015426636
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,32,2,power_law_1.2,0.0374566388130188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,32,2,power_law_1.2,0.03298432111740112
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,64,balanced,0.06615551948547363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,64,balanced,0.06918272018432617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,64,balanced,0.06598656177520752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,64,balanced,0.0690124797821045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,64,balanced,0.0782489585876465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,64,balanced,0.09526783943176269
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,64,balanced,0.1509913635253906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,64,balanced,0.1500441646575928
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,64,balanced,0.15432064056396483
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,64,balanced,0.15275903701782229
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,64,balanced,0.15599360466003417
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,64,balanced,0.15678463935852052
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,64,balanced,0.16149248123168944
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,32,2,power_law_1.2,0.048244481086730955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,64,balanced,0.15912063598632814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,64,balanced,0.16312959671020508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,64,balanced,0.16737152099609376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,64,balanced,0.17068288803100587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,64,balanced,0.17853055953979494
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,64,balanced,0.17666303634643554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,64,balanced,0.1904742431640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,64,balanced,0.19850624084472657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,64,balanced,0.23687295913696288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,64,balanced,0.2439193534851074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,64,balanced,0.3322572708129883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,64,balanced,0.3997324752807617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,64,balanced,0.03826816082000732
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,64,balanced,0.5560550308227539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,64,balanced,0.03829504013061523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,64,balanced,0.04116352081298828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,64,balanced,0.6901862335205078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,64,balanced,0.03835520029067993
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,64,balanced,0.046209278106689455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,64,balanced,0.060823040008544924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,64,balanced,0.06326784133911133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,64,balanced,0.0645081615447998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,64,balanced,0.0667187213897705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,64,balanced,0.06456575870513917
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,64,balanced,0.0671782398223877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,64,balanced,0.07102975845336915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,64,balanced,0.06833280086517333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,64,balanced,0.07127935886383056
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,64,balanced,0.0712998390197754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,64,balanced,0.07275008201599122
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,64,balanced,0.07634560108184815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,64,balanced,0.0848742389678955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,64,balanced,0.0785420799255371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,64,balanced,0.08773632049560547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,64,balanced,0.09470591545104981
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,64,balanced,0.10454015731811524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,64,balanced,0.1166207981109619
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,64,balanced,0.1541977596282959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,64,balanced,0.17949312210083007
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,64,balanced,0.2536294364929199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,64,balanced,0.31060224533081054
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,32,2,power_law_1.2,0.0702566385269165
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,32,2,power_law_1.2,0.17408639907836915
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,32,2,power_law_1.2,0.146496000289917
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,1,64,balanced,0.020460801124572752
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,1,64,balanced,0.021048319339752198
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,1,64,balanced,0.034237439632415774
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,1,64,balanced,0.03457535982131958
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,1,64,balanced,0.020595200061798095
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,1,64,balanced,0.03399296045303345
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,1,64,balanced,0.03483776092529297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,1,64,balanced,0.03523967981338501
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,1,64,balanced,0.03456640005111694
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,1,64,balanced,0.03549823999404907
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,1,64,balanced,0.035301120281219484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,1,64,balanced,0.03492991924285889
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,1,64,balanced,0.0356441593170166
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,1,64,balanced,0.035983359813690184
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,1,64,balanced,0.035732479095458985
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,1,64,balanced,0.0359769606590271
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,1,64,balanced,0.035572481155395505
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,1,64,balanced,0.036556799411773686
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,1,64,balanced,0.03753983974456787
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,1,64,balanced,0.03860863924026489
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,1,64,balanced,0.02060287952423096
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,32,2,power_law_1.2,0.23207679748535157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,64,power_law_1.01,0.09598208427429199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,64,power_law_1.01,0.09574784278869629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,64,power_law_1.01,0.09676287651062013
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,64,power_law_1.01,0.09676416397094727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,64,power_law_1.01,0.1496025562286377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,64,power_law_1.01,0.14940799713134764
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,64,power_law_1.01,0.15532416343688965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,64,power_law_1.01,0.1572697639465332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,64,power_law_1.01,0.1608332824707031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,64,power_law_1.01,0.16216320037841797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,64,power_law_1.01,0.16337919235229492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,64,power_law_1.01,0.1670400047302246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,64,power_law_1.01,0.16880767822265624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,64,power_law_1.01,0.17412736892700195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,64,power_law_1.01,0.17757823944091797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,64,power_law_1.01,0.18568191528320313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,64,power_law_1.01,0.1883148765563965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,64,power_law_1.01,0.21230335235595704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,64,power_law_1.01,0.23330303192138674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,64,power_law_1.01,0.30443391799926756
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,64,power_law_1.01,0.3792755126953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,64,power_law_1.01,0.5264499282836914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,1,64,balanced,0.021736960411071777
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,64,power_law_1.01,0.5703398513793945
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,1,64,balanced,0.03458944082260132
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,64,power_law_1.01,0.8647411346435547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,64,power_law_1.01,1.124752655029297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,64,power_law_1.01,0.05936639785766602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,64,power_law_1.01,0.06040063858032226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,64,power_law_1.01,1.6836058044433593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,64,power_law_1.01,0.06095871925354004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,64,power_law_1.01,0.059425277709960936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,64,power_law_1.01,0.06516992092132569
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,64,power_law_1.01,0.06204160213470459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,64,power_law_1.01,0.06831232070922852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,64,power_law_1.01,0.06751999855041504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,64,power_law_1.01,0.07312511920928955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,64,power_law_1.01,0.07054975986480713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,64,power_law_1.01,0.07335680007934571
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,64,power_law_1.01,2.562696838378906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,64,power_law_1.01,0.07321983814239502
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,64,power_law_1.01,0.07422848224639893
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,64,power_law_1.01,0.08120832443237305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,64,power_law_1.01,0.08584192276000976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,64,power_law_1.01,0.08752896308898925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,64,power_law_1.01,0.08460800170898437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,64,power_law_1.01,0.09606783866882325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,64,power_law_1.01,0.10519680023193359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,64,power_law_1.01,0.13334783554077148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,64,power_law_1.01,0.1656537628173828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,64,power_law_1.01,0.19866367340087893
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,64,power_law_1.01,0.25337215423583986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,64,power_law_1.01,0.3666432189941406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,64,power_law_1.01,0.4643622589111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,64,power_law_1.01,0.6938214111328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,1,64,power_law_1.01,0.035475199222564695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,64,power_law_1.01,0.9029106903076173
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,1,64,power_law_1.01,0.03507711887359619
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,1,64,power_law_1.01,0.03500416040420532
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,1,64,power_law_1.01,0.03531775951385498
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,1,64,power_law_1.01,0.0303654408454895
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,1,64,power_law_1.01,0.035162880420684814
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,1,64,power_law_1.01,0.034910719394683834
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,1,64,power_law_1.01,0.03582079887390137
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,1,64,power_law_1.01,0.03520895957946778
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,1,64,power_law_1.01,0.035563519001007085
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,1,64,power_law_1.01,0.035228159427642826
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,1,64,power_law_1.01,0.035905280113220216
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,1,64,power_law_1.01,0.035897600650787356
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,1,64,power_law_1.01,0.03542655944824219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,1,64,power_law_1.01,0.03612672090530396
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,1,64,power_law_1.01,0.035600640773773194
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,1,64,balanced,0.05207551956176758
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,1,64,balanced,0.05065728187561035
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,1,64,balanced,0.06644480228424073
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,1,64,balanced,0.08023039817810058
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,32,2,power_law_1.2,0.3216115188598633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,1,64,power_law_1.01,0.03456768035888672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,1,64,power_law_1.01,0.04292223930358886
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,1,64,power_law_1.01,0.04516863822937012
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,1,64,power_law_1.01,0.05026815891265869
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,1,64,power_law_1.01,0.0590399980545044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,64,power_law_1.2,0.0951091194152832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,64,power_law_1.2,0.09542016029357911
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,64,power_law_1.2,0.0961228847503662
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,64,power_law_1.2,0.0931980800628662
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,64,power_law_1.2,0.14705408096313477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,64,power_law_1.2,0.1491379165649414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,64,power_law_1.2,0.15515904426574706
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,64,power_law_1.2,0.15573247909545898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,64,power_law_1.2,0.1613363265991211
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,1,64,power_law_1.01,0.06773759841918944
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,64,power_law_1.2,0.1573299217224121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,64,power_law_1.2,0.16562944412231445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,64,power_law_1.2,0.16666751861572265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,64,power_law_1.2,0.17080703735351563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,64,power_law_1.2,0.1768742370605469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,64,power_law_1.2,0.18663936614990234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,64,power_law_1.2,0.18351999282836914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,64,power_law_1.2,0.1930521583557129
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,1,64,power_law_1.01,0.08097663879394532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,64,power_law_1.2,0.2799807929992676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,64,power_law_1.2,0.3945382308959961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,64,power_law_1.2,0.5002803039550782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,64,power_law_1.2,0.24596223831176758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,64,power_law_1.2,0.6608576202392579
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,64,power_law_1.2,0.9539520263671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,64,power_law_1.2,1.3043661499023438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,64,power_law_1.2,0.059171838760375975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,64,power_law_1.2,1.795036163330078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,64,power_law_1.2,0.05920000076293945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,64,power_law_1.2,0.05996032238006592
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,64,power_law_1.2,0.05793280124664306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,64,power_law_1.2,0.057762560844421384
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,64,power_law_1.2,2.6368179321289062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,64,power_law_1.2,0.06119679927825927
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,64,power_law_1.2,0.06503424167633057
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,64,power_law_1.2,0.07057407855987549
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,64,power_law_1.2,0.06976768016815185
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,64,power_law_1.2,0.07037568092346191
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,64,power_law_1.2,0.07188992023468017
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,64,power_law_1.2,0.07359360218048096
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,1,64,power_law_1.01,0.11536895751953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,64,power_law_1.2,0.07833600044250488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,64,power_law_1.2,0.08300928115844727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,64,power_law_1.2,0.08643072128295899
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,64,power_law_1.2,0.0878758430480957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,64,power_law_1.2,0.08376447677612305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,64,power_law_1.2,0.10579968452453614
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,64,power_law_1.2,0.12411007881164551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,64,power_law_1.2,0.14775936126708983
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,64,power_law_1.2,0.2087808036804199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,64,power_law_1.2,0.29502975463867187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,64,power_law_1.2,0.3287500762939453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,64,power_law_1.2,0.5437670516967773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,64,power_law_1.2,4.3876416015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,64,power_law_1.2,0.7128486633300781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,1,64,power_law_1.2,0.034251520633697505
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,1,64,power_law_1.2,0.034910719394683834
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,1,64,power_law_1.2,0.031784958839416504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,64,power_law_1.2,1.106524124145508
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,1,64,power_law_1.2,0.02688256025314331
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,1,64,power_law_1.2,0.027292160987854003
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,1,64,power_law_1.2,0.03319040060043335
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,1,64,power_law_1.2,0.032412159442901614
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,64,power_law_1.2,1.4917721557617187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,1,64,power_law_1.2,0.03498879909515381
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,1,64,power_law_1.2,0.03447295904159546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,1,64,power_law_1.2,0.03461503982543945
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,1,64,power_law_1.2,0.03556479930877686
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,1,64,power_law_1.2,0.03225600004196167
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,1,64,power_law_1.2,0.03565824031829834
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,1,64,power_law_1.2,0.034703359603881836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,1,64,power_law_1.2,0.03602047920227051
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,1,64,power_law_1.2,0.045182719230651855
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,1,64,power_law_1.2,0.05315199851989746
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,1,64,power_law_1.2,0.035270400047302246
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,1,64,power_law_1.01,0.13273856163024902
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,1,64,power_law_1.01,0.28245887756347654
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,1,64,power_law_1.01,0.20164735794067382
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,1,64,power_law_1.2,0.034790399074554446
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,1,64,power_law_1.2,0.049247999191284184
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,32,balanced,0.04179200172424317
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,32,balanced,0.04083072185516358
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,1,64,power_law_1.2,0.09980928421020507
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,32,balanced,0.04076416015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,1,64,power_law_1.2,0.0631321620941162
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,1,64,power_law_1.2,0.08886143684387207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,32,balanced,0.07427072048187255
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,32,balanced,0.09848959922790528
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,32,balanced,0.14400511741638183
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,32,balanced,0.14481280326843263
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,32,balanced,0.1542527961730957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,32,balanced,0.15012096405029296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,32,balanced,0.1525887966156006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,32,balanced,0.1548736000061035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,32,balanced,0.15669376373291016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,32,balanced,0.16163583755493166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,32,balanced,0.16505599975585938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,32,balanced,0.1726963233947754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,32,balanced,0.16560768127441405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,32,balanced,0.17674879074096678
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,32,balanced,0.18696704864501953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,32,balanced,0.2050124740600586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,32,balanced,0.21362815856933595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,32,balanced,0.2637759971618652
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,32,balanced,0.29501184463500973
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,32,balanced,0.3780748748779297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,32,balanced,0.5005900955200195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,32,balanced,0.03557631969451904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,32,balanced,0.6989606475830078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,32,balanced,0.06308351993560792
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,32,balanced,0.035846400260925296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,32,balanced,0.03538304090499878
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,32,balanced,0.8927423858642578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,32,balanced,0.03749504089355469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,32,balanced,0.044442877769470215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,32,balanced,0.06270080089569093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,32,balanced,0.06898560047149657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,32,balanced,0.07123456001281739
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,32,balanced,0.07449600219726563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,32,balanced,0.07087744235992431
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,32,balanced,0.07640575885772705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,32,balanced,0.07296000003814698
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,32,balanced,0.07434112071990967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,32,balanced,0.0791756820678711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,32,balanced,0.08226431846618652
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,32,balanced,0.08356351852416992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,32,balanced,0.07680895805358887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,32,balanced,0.08531455993652344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,32,balanced,0.09042943954467773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,32,balanced,0.09791359901428223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,32,balanced,0.1087718391418457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,32,balanced,0.13192447662353515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,32,balanced,0.15191424369812012
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,32,balanced,0.20806655883789063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,32,balanced,0.24775808334350585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,32,balanced,0.360340461730957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,32,balanced,0.45637760162353513
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,1,64,power_law_1.2,0.16443904876708984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,1,64,power_law_1.2,0.19307647705078126
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,1,64,power_law_1.2,0.31647104263305664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,1,64,power_law_1.2,0.4084352111816406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,2,32,balanced,0.02027776002883911
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,2,32,balanced,0.03604991912841797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,2,32,balanced,0.020057599544525146
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,2,32,balanced,0.020751359462738036
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,2,32,balanced,0.03587199926376343
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,2,32,balanced,0.0359987211227417
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,2,32,balanced,0.03681792020797729
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,2,32,balanced,0.036532480716705326
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,2,32,balanced,0.020663039684295652
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,2,32,balanced,0.03581696033477783
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,2,32,balanced,0.03624703884124756
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,2,32,balanced,0.03606911897659302
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,2,32,balanced,0.036385281085968016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,2,32,balanced,0.03689343929290771
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,2,32,balanced,0.03703423976898194
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,2,32,balanced,0.03890304088592529
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,2,32,balanced,0.03936000108718872
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,2,32,balanced,0.040517120361328124
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,2,32,balanced,0.04219520092010498
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,2,32,balanced,0.02413696050643921
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,32,power_law_1.01,0.09650943756103517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,32,power_law_1.01,0.09683712005615235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,32,power_law_1.01,0.09757696151733398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,32,power_law_1.01,0.09923456192016603
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,32,power_law_1.01,0.14070015907287597
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,32,power_law_1.01,0.1423078441619873
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,2,32,balanced,0.03631360054016113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,32,power_law_1.01,0.1499903964996338
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,32,power_law_1.01,0.14895487785339356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,32,power_law_1.01,0.15222016334533692
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,32,power_law_1.01,0.1530662441253662
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,32,power_law_1.01,0.15541888236999513
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,32,power_law_1.01,0.158023681640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,32,power_law_1.01,0.16351615905761718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,32,power_law_1.01,0.17437311172485354
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,2,32,balanced,0.03605247974395752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,32,power_law_1.01,0.171331844329834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,32,power_law_1.01,0.17311231613159178
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,32,power_law_1.01,0.17902719497680664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,32,power_law_1.01,0.20118528366088867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,32,power_law_1.01,0.22487167358398436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,32,power_law_1.01,0.2718976020812988
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,32,power_law_1.01,0.31785728454589846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,32,power_law_1.01,0.47733249664306643
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,32,power_law_1.01,0.5545036697387695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,32,power_law_1.01,0.7321612548828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,32,power_law_1.01,0.9896153259277345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,32,power_law_1.01,0.060476160049438475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,32,power_law_1.01,1.4789414978027344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,32,power_law_1.01,0.060056319236755376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,32,power_law_1.01,0.06046720027923584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,32,power_law_1.01,0.06209152221679688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,32,power_law_1.01,0.06523007869720458
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,32,power_law_1.01,0.06318079948425293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,32,power_law_1.01,0.07004799842834472
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,32,power_law_1.01,0.07014656066894531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,32,power_law_1.01,0.07127679824829102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,32,power_law_1.01,1.9887283325195313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,32,power_law_1.01,0.07447423934936523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,32,power_law_1.01,0.07753087997436522
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,32,power_law_1.01,0.08277759552001954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,32,power_law_1.01,0.06925183773040772
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,32,power_law_1.01,0.08504575729370117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,32,power_law_1.01,0.07942143917083741
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,32,power_law_1.01,0.08293631553649902
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,32,power_law_1.01,0.08684288024902344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,32,power_law_1.01,0.09805312156677246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,32,power_law_1.01,0.1394585609436035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,32,power_law_1.01,0.1699648094177246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,32,power_law_1.01,0.23534463882446288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,32,power_law_1.01,0.2918732833862305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,32,power_law_1.01,0.10606847763061525
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,32,power_law_1.01,0.4058700942993164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,32,power_law_1.01,0.4836211013793945
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,2,32,power_law_1.01,0.03568896055221558
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,32,power_law_1.01,0.7270130920410156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,2,32,power_law_1.01,0.035848960876464844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,2,32,power_law_1.01,0.03581824064254761
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,32,power_law_1.01,1.0322137451171876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,2,32,power_law_1.01,0.03601279973983765
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,2,32,power_law_1.01,0.036046080589294434
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,2,32,power_law_1.01,0.030265600681304933
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,2,32,power_law_1.01,0.03517184019088745
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,2,32,power_law_1.01,0.034106879234313964
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,2,32,power_law_1.01,0.03561984062194824
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,2,32,power_law_1.01,0.03540607929229737
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,2,32,power_law_1.01,0.03574656009674072
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,2,32,power_law_1.01,0.03612416028976441
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,2,32,power_law_1.01,0.035828480720520015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,2,32,power_law_1.01,0.03593087911605835
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,2,32,power_law_1.01,0.03632512092590332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,2,32,power_law_1.01,0.03665535926818848
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,2,32,power_law_1.01,0.0366156792640686
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,2,32,balanced,0.0357696008682251
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,2,32,balanced,0.05670400142669678
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,2,32,balanced,0.09527168273925782
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,2,32,balanced,0.060604162216186516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,2,32,balanced,0.07876863956451416
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,2,32,power_law_1.01,0.04015359878540039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,2,32,power_law_1.01,0.05240320205688477
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,2,32,power_law_1.01,0.04521984100341797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,2,32,power_law_1.01,0.05891327857971192
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,32,power_law_1.2,0.09620736122131349
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,32,power_law_1.2,0.09664640426635743
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,32,power_law_1.2,0.0957862377166748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,32,power_law_1.2,0.09761407852172851
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,32,power_law_1.2,0.137957124710083
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,32,power_law_1.2,0.14518912315368654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,32,power_law_1.2,0.14870400428771974
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,32,power_law_1.2,0.1542848014831543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,32,power_law_1.2,0.15645695686340333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,32,power_law_1.2,0.15543423652648927
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,32,power_law_1.2,0.15959039688110352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,32,power_law_1.2,0.15975168228149414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,32,power_law_1.2,0.16703615188598633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,32,power_law_1.2,0.16915071487426758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,32,power_law_1.2,0.17184640884399413
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,32,power_law_1.2,0.175861759185791
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,32,power_law_1.2,0.18350847244262697
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,32,power_law_1.2,0.208353271484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,32,power_law_1.2,0.23392127990722655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,32,power_law_1.2,0.3168191909790039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,32,power_law_1.2,0.37074432373046873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,32,power_law_1.2,0.532058868408203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,32,power_law_1.2,0.6704678344726562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,32,power_law_1.2,1.1087347412109376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,32,power_law_1.2,1.4382067871093749
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,2,32,power_law_1.01,0.08536959648132324
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,2,32,power_law_1.01,0.11392767906188965
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,2,32,power_law_1.01,0.07099264144897462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,32,power_law_1.2,0.06051712036132813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,2,32,power_law_1.01,0.14578432083129883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,32,power_law_1.2,0.06063104152679444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,32,power_law_1.2,0.06047743797302246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,32,power_law_1.2,0.06148223876953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,32,power_law_1.2,0.06366335868835449
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,32,power_law_1.2,0.06892416000366211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,32,power_law_1.2,0.061308159828186035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,32,power_law_1.2,0.07382783889770508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,32,power_law_1.2,2.507098846435547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,32,power_law_1.2,0.07201536178588867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,32,power_law_1.2,0.073503999710083
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,32,power_law_1.2,0.07260287761688232
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,32,power_law_1.2,0.07544447898864745
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,32,power_law_1.2,0.08190976142883301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,32,power_law_1.2,0.0815334415435791
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,32,power_law_1.2,0.08493311882019043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,32,power_law_1.2,0.08860032081604004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,32,power_law_1.2,0.10094335556030273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,32,power_law_1.2,0.11635968208312988
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,32,power_law_1.2,0.14866944313049318
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,32,power_law_1.2,0.19393280029296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,32,power_law_1.2,0.27159551620483396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,32,power_law_1.2,3.001258239746094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,32,power_law_1.2,0.3636876678466797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,32,power_law_1.2,0.4468403244018555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,2,32,power_law_1.2,0.03584511995315552
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,32,power_law_1.2,0.086310396194458
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,2,32,power_law_1.2,0.03537919998168945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,32,power_law_1.2,0.6471513366699219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,2,32,power_law_1.2,0.03598464012145996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,2,32,power_law_1.2,0.031223039627075195
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,2,32,power_law_1.2,0.029020159244537352
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,2,32,power_law_1.2,0.036125440597534184
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,2,32,power_law_1.2,0.03417855978012085
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,2,32,power_law_1.2,0.035487999916076665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,32,power_law_1.2,1.0967820739746093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,2,32,power_law_1.2,0.035901439189910886
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,2,32,power_law_1.2,0.033251841068267826
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,2,32,power_law_1.2,0.031978240013122564
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,2,32,power_law_1.2,0.03201535940170288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,32,power_law_1.2,1.407132110595703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,2,32,power_law_1.2,0.03406975984573364
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,2,32,power_law_1.2,0.03581568002700806
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,2,32,power_law_1.2,0.03532543897628784
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,2,32,power_law_1.2,0.04386176109313965
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,2,32,power_law_1.2,0.03623807907104492
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,2,32,power_law_1.01,0.1947417640686035
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,2,32,power_law_1.01,0.27189632415771486
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,2,32,power_law_1.2,0.037931520938873295
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,2,32,power_law_1.2,0.050817279815673834
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,2,32,power_law_1.2,0.06603519916534424
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,2,32,power_law_1.2,0.06082560062408447
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,16,balanced,0.032958719730377194
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,16,balanced,0.03331968069076538
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,16,balanced,0.035479040145874025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,16,balanced,0.061975040435791016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,16,balanced,0.07557248115539551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,16,balanced,0.09923968315124512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,16,balanced,0.14693504333496094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,2,32,power_law_1.2,0.0943667221069336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,16,balanced,0.15095680236816406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,16,balanced,0.15172224044799804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,16,balanced,0.15454463958740233
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,16,balanced,0.16041215896606445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,16,balanced,0.16339199066162108
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,16,balanced,0.15825535774230956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,16,balanced,0.1637183952331543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,16,balanced,0.1669478416442871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,16,balanced,0.1762214469909668
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,2,32,power_law_1.2,0.12859904289245605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,16,balanced,0.19109247207641603
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,16,balanced,0.20383232116699218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,16,balanced,0.23630720138549804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,16,balanced,0.2650022315979004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,16,balanced,0.3353740692138672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,16,balanced,0.39733631134033204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,16,balanced,0.5332172775268555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,16,balanced,0.7108223724365235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,16,balanced,0.1510028839111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,16,balanced,0.034612479209899905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,16,balanced,0.03422719955444336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,16,balanced,0.03703423976898194
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,16,balanced,0.040317440032958986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,16,balanced,1.0162751770019531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,16,balanced,0.0470143985748291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,16,balanced,0.07095424175262452
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,16,balanced,0.08012031555175782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,16,balanced,0.08228480339050293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,16,balanced,0.07873280048370361
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,16,balanced,0.08309503555297851
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,16,balanced,1.3201394653320313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,16,balanced,0.08488448143005371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,16,balanced,0.08869248390197754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,16,balanced,0.0917414379119873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,16,balanced,0.08260352134704589
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,16,balanced,0.08673151969909668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,16,balanced,0.08953727722167969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,16,balanced,0.0948300838470459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,16,balanced,0.10141823768615724
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,16,balanced,0.11210880279541016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,16,balanced,0.13102720260620118
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,16,balanced,0.15050239562988282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,16,balanced,0.18926591873168946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,16,balanced,0.2246169662475586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,16,balanced,0.3197324752807617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,16,balanced,0.39600128173828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,16,balanced,0.5723315048217773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,16,balanced,0.7385702514648438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,2,32,power_law_1.2,0.20833408355712893
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,2,32,power_law_1.2,0.17337728500366212
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,2,32,power_law_1.2,0.29653631210327147
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,2,32,power_law_1.2,0.4620646286010742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,4,16,balanced,0.021095681190490722
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,4,16,balanced,0.041105918884277344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,4,16,balanced,0.0214899206161499
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,4,16,balanced,0.04164224147796631
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,4,16,balanced,0.04205823898315429
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,4,16,balanced,0.041776638031005855
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,4,16,balanced,0.041948161125183105
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,4,16,balanced,0.02372607946395874
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,4,16,balanced,0.04196608066558838
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,4,16,balanced,0.04197120189666748
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,4,16,balanced,0.04173056125640869
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,4,16,balanced,0.042131199836730956
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,4,16,balanced,0.04363391876220703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,4,16,balanced,0.04443903923034668
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,4,16,balanced,0.0459827184677124
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,4,16,balanced,0.020908799171447754
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,4,16,balanced,0.04741375923156738
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,4,16,balanced,0.04984447956085205
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,4,16,balanced,0.05294847965240479
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,4,16,balanced,0.02721407890319824
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,4,16,balanced,0.04165247917175293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,16,power_law_1.01,0.0743065595626831
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,16,power_law_1.01,0.09349504470825196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,16,power_law_1.01,0.09347711563110352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,16,power_law_1.01,0.09212544441223144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,16,power_law_1.01,0.09264896392822267
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,16,power_law_1.01,0.138090238571167
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,16,power_law_1.01,0.14291328430175781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,16,power_law_1.01,0.14330368041992186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,16,power_law_1.01,0.14858367919921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,16,power_law_1.01,0.14777983665466307
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,16,power_law_1.01,0.15455360412597657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,16,power_law_1.01,0.16089984893798828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,16,power_law_1.01,0.16466432571411133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,16,power_law_1.01,0.16148096084594726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,16,power_law_1.01,0.16808319091796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,16,power_law_1.01,0.17819007873535156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,16,power_law_1.01,0.20895872116088868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,16,power_law_1.01,0.22884992599487303
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,16,power_law_1.01,0.16390655517578126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,16,power_law_1.01,0.2852428817749023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,16,power_law_1.01,0.36039424896240235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,16,power_law_1.01,0.4371289443969727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,16,power_law_1.01,0.5954841613769531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,16,power_law_1.01,0.8066969299316407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,16,power_law_1.01,0.9861695861816406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,16,power_law_1.01,1.5441293334960937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,16,power_law_1.01,0.04263552188873291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,16,power_law_1.01,0.06589183807373047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,16,power_law_1.01,0.06544896125793456
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,16,power_law_1.01,2.1146829223632815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,16,power_law_1.01,0.0659878396987915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,16,power_law_1.01,0.06629504203796387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,16,power_law_1.01,0.0701478385925293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,16,power_law_1.01,0.07442048072814941
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,4,16,balanced,0.04140416145324707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,16,power_law_1.01,0.07416704177856445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,16,power_law_1.01,0.07772287845611572
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,16,power_law_1.01,0.07925248146057129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,16,power_law_1.01,0.08389504432678223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,16,power_law_1.01,0.08759167671203613
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,16,power_law_1.01,0.09282303810119628
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,4,16,balanced,0.04183807849884033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,16,power_law_1.01,0.08715904235839844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,16,power_law_1.01,0.08796544075012207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,16,power_law_1.01,0.08941823959350585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,16,power_law_1.01,0.0948953628540039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,16,power_law_1.01,0.1256704044342041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,16,power_law_1.01,0.15123456001281738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,16,power_law_1.01,0.17939840316772462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,16,power_law_1.01,0.25410560607910154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,16,power_law_1.01,0.3231526565551758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,16,power_law_1.01,0.46626560211181645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,16,power_law_1.01,0.11331583976745604
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,16,power_law_1.01,0.5623795318603515
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,4,16,power_law_1.01,0.02599936008453369
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,4,16,power_law_1.01,0.04137599945068359
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,4,16,power_law_1.01,0.04114431858062744
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,4,16,power_law_1.01,0.041785597801208496
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,4,16,power_law_1.01,0.040919041633605956
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,4,16,power_law_1.01,0.03801471948623657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,16,power_law_1.01,1.122266845703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,4,16,power_law_1.01,0.0385100793838501
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,4,16,power_law_1.01,0.03885184049606323
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,4,16,power_law_1.01,0.0378278398513794
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,4,16,power_law_1.01,0.03872512102127075
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,4,16,power_law_1.01,0.03940479993820191
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,4,16,power_law_1.01,0.040171518325805664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,16,power_law_1.01,0.7849382019042969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,4,16,power_law_1.01,0.04056960105895996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,4,16,power_law_1.01,0.04006271839141846
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,4,16,power_law_1.01,0.04075136184692383
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,4,16,power_law_1.01,0.041474561691284176
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,4,16,power_law_1.01,0.04309375762939453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,4,16,balanced,0.1280294418334961
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,4,16,balanced,0.07750016212463379
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,4,16,balanced,0.07193727970123291
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,4,16,balanced,0.10431488037109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,4,16,power_law_1.01,0.0657254409790039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,4,16,power_law_1.01,0.05506175994873047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,4,16,power_law_1.01,0.048533759117126464
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,4,16,power_law_1.01,0.07545983791351318
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,16,power_law_1.2,0.07405183792114259
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,16,power_law_1.2,0.09369600296020508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,16,power_law_1.2,0.09164544105529784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,16,power_law_1.2,0.09476223945617676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,16,power_law_1.2,0.09096960067749024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,16,power_law_1.2,0.13812095642089844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,16,power_law_1.2,0.1436044788360596
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,16,power_law_1.2,0.1410752010345459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,16,power_law_1.2,0.14530431747436523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,16,power_law_1.2,0.15118592262268066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,16,power_law_1.2,0.15693311691284179
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,16,power_law_1.2,0.16092031478881835
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,4,16,power_law_1.01,0.09333375930786134
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,16,power_law_1.2,0.16537471771240236
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,4,16,power_law_1.01,0.1117529582977295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,16,power_law_1.2,0.1623744010925293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,4,16,power_law_1.01,0.13940863609313964
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,16,power_law_1.2,0.16488576889038087
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,16,power_law_1.2,0.17055999755859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,16,power_law_1.2,0.18724863052368163
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,16,power_law_1.2,0.2159052848815918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,4,16,power_law_1.01,0.18590719223022462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,16,power_law_1.2,0.24144767761230468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,16,power_law_1.2,0.2968422317504883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,16,power_law_1.2,0.3656256103515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,16,power_law_1.2,0.5389913558959961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,16,power_law_1.2,0.6514214324951172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,16,power_law_1.2,0.045404157638549804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,16,power_law_1.2,0.9803609466552734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,16,power_law_1.2,0.06602240085601807
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,16,power_law_1.2,0.06420095920562743
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,16,power_law_1.2,1.2127129364013673
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,16,power_law_1.2,0.06613120079040527
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,16,power_law_1.2,0.06909311771392822
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,16,power_law_1.2,0.0701580810546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,16,power_law_1.2,0.07481215953826904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,16,power_law_1.2,0.07613056182861329
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,16,power_law_1.2,1.7779302978515623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,16,power_law_1.2,0.07989247798919677
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,16,power_law_1.2,0.08490495681762696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,16,power_law_1.2,0.08731648445129395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,16,power_law_1.2,0.08667008399963379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,16,power_law_1.2,0.08586239814758301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,16,power_law_1.2,0.08903039932250977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,16,power_law_1.2,0.0895206356048584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,16,power_law_1.2,0.09929216384887696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,16,power_law_1.2,2.3641459655761716
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,16,power_law_1.2,0.1179417610168457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,16,power_law_1.2,0.13017087936401367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,16,power_law_1.2,0.1648588752746582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,16,power_law_1.2,0.1929574394226074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,16,power_law_1.2,0.07683584213256836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,16,power_law_1.2,0.29605119705200195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,16,power_law_1.2,0.34118526458740234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,4,16,power_law_1.2,0.026049280166625978
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,4,16,power_law_1.2,0.041139202117919924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,16,power_law_1.2,0.6443520355224609
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,4,16,power_law_1.2,0.04092927932739258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,4,16,power_law_1.2,0.04117248058319092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,16,power_law_1.2,0.9908799743652343
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,4,16,power_law_1.2,0.034457600116729735
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,4,16,power_law_1.2,0.03784575939178467
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,4,16,power_law_1.2,0.03861504077911377
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,4,16,power_law_1.2,0.0374451208114624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,4,16,power_law_1.2,0.03851775884628296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,16,power_law_1.2,0.4724876785278321
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,16,power_law_1.2,1.4543142700195313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,4,16,power_law_1.2,0.04023551940917969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,4,16,power_law_1.2,0.04116352081298828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,4,16,power_law_1.2,0.04015359878540039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,4,16,power_law_1.2,0.03623424053192138
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,4,16,power_law_1.2,0.04187007904052734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,4,16,power_law_1.2,0.039441919326782225
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,4,16,power_law_1.2,0.04172416210174561
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,4,16,power_law_1.01,0.25456640243530276
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,4,16,power_law_1.01,0.3111104011535645
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,4,16,power_law_1.2,0.04639872074127197
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,4,16,power_law_1.2,0.07648896217346192
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,4,16,power_law_1.2,0.06321536064147949
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,4,16,power_law_1.2,0.054329600334167484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,8,8,balanced,0.032983040809631346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,8,8,balanced,0.03338624000549316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,8,8,balanced,0.03827199935913086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,8,8,balanced,0.06389503955841065
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,4,16,power_law_1.2,0.08465791702270507
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,4,16,power_law_1.2,0.11657728195190428
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,8,8,balanced,0.10852224349975585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,8,8,balanced,0.15920384407043459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,8,8,balanced,0.15670528411865234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,8,8,balanced,0.16500608444213866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,8,8,balanced,0.16917503356933594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,8,8,balanced,0.16109695434570312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,8,8,balanced,0.1663961601257324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,8,8,balanced,0.1703603172302246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,8,8,balanced,0.07889023780822754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,8,8,balanced,0.17816448211669922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,8,8,balanced,0.18169727325439453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,8,8,balanced,0.1911577606201172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,8,8,balanced,0.20168575286865234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,8,8,balanced,0.23156864166259766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,8,8,balanced,0.2619097518920898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,8,8,balanced,0.32051071166992184
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,8,8,balanced,0.3811417770385742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,8,8,balanced,0.5011955261230469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,8,8,balanced,0.6187968063354492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,8,8,balanced,0.8653107452392579
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,8,8,balanced,1.1548531341552735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,8,8,balanced,0.034781439304351805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,8,8,balanced,0.03747200012207032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,8,8,balanced,0.04450943946838379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,8,8,balanced,1.6814822387695312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,8,8,balanced,0.04764416217803955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,8,8,balanced,0.06094207763671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,8,8,balanced,0.08526975631713866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,8,8,balanced,0.09607680320739746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,8,8,balanced,0.0988748836517334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,8,8,balanced,0.102608642578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,8,8,balanced,0.10626303672790527
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,8,8,balanced,0.0960319995880127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,8,8,balanced,0.10092415809631347
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,8,8,balanced,0.10326016426086426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,8,8,balanced,0.10810879707336425
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,8,8,balanced,0.10819199562072754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,8,8,balanced,0.11361151695251466
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,8,8,balanced,0.1222719955444336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,8,8,balanced,0.14113408088684082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,8,8,balanced,0.16009471893310548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,8,8,balanced,0.19672447204589844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,8,8,balanced,0.2318604850769043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,8,8,balanced,0.3062899208068848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,8,8,balanced,2.2719949340820316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,8,8,balanced,0.37480960845947264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,8,8,balanced,0.542110710144043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,8,8,balanced,0.6815654754638671
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,8,8,balanced,1.008015365600586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,8,8,balanced,1.3094412231445314
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,4,16,power_law_1.2,0.13546367645263674
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,4,16,power_law_1.2,0.22231679916381836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,4,16,power_law_1.2,0.2533401679992676
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,4,16,power_law_1.2,0.42443008422851564
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,4,16,power_law_1.2,0.5599795150756836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,8,8,balanced,0.02350719928741455
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,8,8,balanced,0.050832638740539546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,8,8,balanced,0.05099008083343506
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,8,8,balanced,0.02409600019454956
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,8,8,balanced,0.05161983966827392
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,8,8,balanced,0.051746559143066403
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,8,8,balanced,0.05165952205657959
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,8,8,balanced,0.05170559883117676
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,8,8,balanced,0.0519923210144043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,8,8,balanced,0.025579519271850586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,8,8,balanced,0.05307007789611816
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,8,8,balanced,0.05358719825744629
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,8,8,balanced,0.05516287803649902
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,8,8,balanced,0.05678719997406005
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,8,8,balanced,0.06002175807952881
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,8,8,balanced,0.06242432117462158
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,8,8,balanced,0.06857215881347656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,8,8,balanced,0.07410687923431396
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,8,8,balanced,0.027258880138397217
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,8,8,balanced,0.03419647932052612
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,8,8,power_law_1.01,0.06209536075592041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,8,8,power_law_1.01,0.0762943983078003
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,8,8,power_law_1.01,0.08519807815551758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,8,8,power_law_1.01,0.08585984230041503
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,8,8,power_law_1.01,0.09676287651062013
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,8,8,power_law_1.01,0.10404224395751953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,8,8,power_law_1.01,0.15188223838806153
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,8,8,power_law_1.01,0.15268863677978514
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,8,8,power_law_1.01,0.1535206413269043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,8,8,power_law_1.01,0.1572812843322754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,8,8,power_law_1.01,0.16457855224609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,8,8,power_law_1.01,0.16415872573852539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,8,8,power_law_1.01,0.16106239318847657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,8,8,power_law_1.01,0.1646656036376953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,8,8,power_law_1.01,0.1702796745300293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,8,8,power_law_1.01,0.17439359664916992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,8,8,power_law_1.01,0.18976640701293945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,8,8,power_law_1.01,0.210516471862793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,8,8,power_law_1.01,0.2410931205749512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,8,8,power_law_1.01,0.3032383918762207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,8,8,power_law_1.01,0.33789184570312497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,8,8,power_law_1.01,0.4755596923828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,8,8,balanced,0.050348801612854
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,8,8,balanced,0.050865921974182125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,8,8,power_law_1.01,0.6258662414550782
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,8,8,balanced,0.050786561965942376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,8,8,power_law_1.01,0.8212595367431641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,8,8,power_law_1.01,0.04121600151062012
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,8,8,power_law_1.01,1.0624575805664063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,8,8,power_law_1.01,0.05069568157196045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,8,8,power_law_1.01,0.06514560222625733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,8,8,power_law_1.01,0.06957952022552491
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,8,8,power_law_1.01,1.5076173400878905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,8,8,power_law_1.01,0.07356287956237792
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,8,8,power_law_1.01,0.07333888053894043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,8,8,power_law_1.01,0.09083135604858399
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,8,8,power_law_1.01,0.090632963180542
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,8,8,power_law_1.01,0.09174528121948242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,8,8,power_law_1.01,0.10053631782531738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,8,8,power_law_1.01,0.09979647636413574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,8,8,power_law_1.01,0.09694720268249511
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,8,8,power_law_1.01,0.09960320472717285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,8,8,power_law_1.01,0.10253567695617676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,8,8,power_law_1.01,0.10329600334167481
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,8,8,power_law_1.01,0.11071231842041016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,8,8,power_law_1.01,0.1330457592010498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,8,8,power_law_1.01,0.09583744049072265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,8,8,power_law_1.01,0.14905983924865723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,8,8,power_law_1.01,0.18116479873657226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,8,8,power_law_1.01,0.21918079376220706
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,8,8,power_law_1.01,0.28928895950317385
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,8,8,power_law_1.01,0.33595649719238285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,8,8,power_law_1.01,0.5243584060668945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,8,8,power_law_1.01,2.094712371826172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,8,8,power_law_1.01,0.6435814666748046
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,8,8,power_law_1.01,0.02681727886199951
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,8,8,power_law_1.01,0.033720319271087644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,8,8,power_law_1.01,0.9256665802001953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,8,8,power_law_1.01,0.03948415994644165
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,8,8,power_law_1.01,0.04156928062438965
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,8,8,power_law_1.01,0.04525951862335205
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,8,8,power_law_1.01,0.04331647872924805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,8,8,power_law_1.01,0.046005759239196774
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,8,8,power_law_1.01,0.045139198303222654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,8,8,power_law_1.01,1.1787213134765626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,8,8,power_law_1.01,0.045885438919067385
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,8,8,power_law_1.01,0.047066879272460935
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,8,8,power_law_1.01,0.04955008029937744
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,8,8,power_law_1.01,0.04934271812438965
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,8,8,power_law_1.01,0.04934400081634521
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,8,8,power_law_1.01,0.05087615966796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,8,8,power_law_1.01,0.05181824207305909
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,8,8,power_law_1.01,0.05443327903747559
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,8,8,power_law_1.01,0.05795711994171142
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,8,8,balanced,0.10774656295776368
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,8,8,balanced,0.11950207710266114
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,8,8,balanced,0.16665088653564453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,8,8,balanced,0.21191551208496096
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,8,8,power_law_1.01,0.06633984088897706
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,8,8,power_law_1.01,0.07171711921691895
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,8,8,power_law_1.01,0.09207679748535155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,8,8,power_law_1.2,0.06329343795776368
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,8,8,power_law_1.2,0.07505663871765136
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,8,8,power_law_1.2,0.07757823944091798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,8,8,power_law_1.2,0.08340607643127441
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,8,8,power_law_1.01,0.10333567619323732
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,8,8,power_law_1.2,0.08847231864929199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,8,8,power_law_1.2,0.09248895645141601
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,8,8,power_law_1.2,0.15302911758422852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,8,8,power_law_1.2,0.1513113594055176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,8,8,power_law_1.2,0.15720191955566407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,8,8,power_law_1.2,0.16065919876098633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,8,8,power_law_1.2,0.16258687973022462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,8,8,power_law_1.2,0.16029823303222657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,8,8,power_law_1.2,0.1600499153137207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,8,8,power_law_1.2,0.1694655990600586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,8,8,power_law_1.2,0.17394559860229492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,8,8,power_law_1.2,0.1798643112182617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,8,8,power_law_1.2,0.19585535049438477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,8,8,power_law_1.2,0.22209152221679687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,8,8,power_law_1.2,0.24700672149658204
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,8,8,power_law_1.01,0.13521663665771483
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,8,8,power_law_1.2,0.3004070472717285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,8,8,power_law_1.01,0.16120704650878906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,8,8,power_law_1.2,0.36188926696777346
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,8,8,power_law_1.01,0.2194444847106934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,8,8,power_law_1.2,0.5139955139160156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,8,8,power_law_1.2,0.6701222229003906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,8,8,power_law_1.2,0.9301299285888671
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,8,8,power_law_1.2,0.04076416015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,8,8,power_law_1.2,1.1835238647460937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,8,8,power_law_1.2,0.058361601829528806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,8,8,power_law_1.2,0.05104383945465087
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,8,8,power_law_1.2,0.06557184219360351
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,8,8,power_law_1.2,0.07492735862731933
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,8,8,power_law_1.2,0.07750783920288086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,8,8,power_law_1.2,1.673192901611328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,8,8,power_law_1.2,0.08775551795959473
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,8,8,power_law_1.2,0.08955519676208497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,8,8,power_law_1.2,0.09545727729797363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,8,8,power_law_1.2,0.10019455909729005
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,8,8,power_law_1.2,0.09558015823364258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,8,8,power_law_1.2,0.09512191772460937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,8,8,power_law_1.2,0.10148608207702636
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,8,8,power_law_1.2,0.10352512359619141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,8,8,power_law_1.2,0.10553728103637697
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,8,8,power_law_1.2,0.11262720108032227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,8,8,power_law_1.2,0.13475456237792968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,8,8,power_law_1.2,2.479731140136719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,8,8,power_law_1.2,0.1505894374847412
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,8,8,power_law_1.2,0.18975231170654297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,8,8,power_law_1.2,0.09150079727172852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,8,8,power_law_1.2,0.21875839233398436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,8,8,power_law_1.2,0.3091814422607422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,8,8,power_law_1.2,0.37595264434814457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,8,8,power_law_1.2,0.5210508728027344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,8,8,power_law_1.2,0.02674560070037842
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,8,8,power_law_1.2,0.7051392364501953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,8,8,power_law_1.2,0.03367039918899536
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,8,8,power_law_1.2,0.04041855812072754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,8,8,power_law_1.2,1.007967987060547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,8,8,power_law_1.2,0.04350207805633545
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,8,8,power_law_1.2,0.041960959434509275
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,8,8,power_law_1.2,0.046443519592285154
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,8,8,power_law_1.2,0.044800000190734865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,8,8,power_law_1.2,1.3520486450195313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,8,8,power_law_1.2,0.045898241996765135
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,8,8,power_law_1.2,0.04603519916534424
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,8,8,power_law_1.2,0.04823808193206787
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,8,8,power_law_1.2,0.048198399543762205
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,8,8,power_law_1.2,0.05217408180236817
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,8,8,power_law_1.2,0.054535679817199714
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,8,8,power_law_1.2,0.03334144115447998
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,8,8,power_law_1.2,0.057815041542053226
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,8,8,power_law_1.2,0.06308864116668701
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,8,8,power_law_1.2,0.07975935935974121
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,8,8,power_law_1.2,0.12468480110168456
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,8,8,power_law_1.01,0.266878719329834
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,8,8,power_law_1.2,0.05072000026702881
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,8,8,power_law_1.01,0.5871744155883789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,8,8,power_law_1.01,0.4215961456298828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,8,8,power_law_1.2,0.07095680236816407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,8,8,power_law_1.2,0.20392576217651368
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,8,8,power_law_1.2,0.18572032928466797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,8,8,power_law_1.2,0.10386816024780274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,16,4,balanced,0.03097471952438354
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,16,4,balanced,0.034079999923706056
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,16,4,balanced,0.039106559753417966
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,8,8,power_law_1.2,0.33492992401123045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,16,4,balanced,0.06697087764739991
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,16,4,balanced,0.08382719993591309
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,16,4,balanced,0.10616064071655273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,16,4,balanced,0.10804479598999024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,16,4,balanced,0.10698111534118653
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,16,4,balanced,0.17265920639038085
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,8,8,power_law_1.2,0.4671295928955078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,16,4,balanced,0.17614591598510743
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,16,4,balanced,0.1801548767089844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,16,4,balanced,0.18283903121948242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,16,4,balanced,0.18904191970825196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,16,4,balanced,0.19420543670654297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,16,4,balanced,0.20652544021606448
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,16,4,balanced,0.2203660774230957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,16,4,balanced,0.2491110420227051
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,16,4,balanced,0.2771327972412109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,16,4,balanced,0.3337164688110351
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,16,4,balanced,0.39335807800292966
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,16,4,balanced,0.5155532836914063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,16,4,balanced,0.6294169616699219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,16,4,balanced,0.19854207992553713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,16,4,balanced,0.03447295904159546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,16,4,balanced,0.868963851928711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,16,4,balanced,0.03745791912078857
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,16,4,balanced,0.04565887928009033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,16,4,balanced,0.058847999572753905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,16,4,balanced,1.1531750488281252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,16,4,balanced,0.07548799991607666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,16,4,balanced,0.10759039878845215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,16,4,balanced,0.1039782428741455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,16,4,balanced,0.10608256340026856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,16,4,balanced,0.11870079994201661
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,16,4,balanced,0.12041983604431153
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,16,4,balanced,0.12261376380920411
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,16,4,balanced,0.1263539218902588
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,16,4,balanced,0.129683198928833
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,16,4,balanced,1.6800909423828123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,16,4,balanced,0.13457023620605468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,16,4,balanced,0.13350144386291504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,16,4,balanced,0.1389452838897705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,16,4,balanced,0.14644736289978028
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,16,4,balanced,0.16550016403198242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,16,4,balanced,0.18340736389160156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,16,4,balanced,0.21867136001586912
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,16,4,balanced,0.2513702392578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,16,4,balanced,0.32201343536376953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,16,4,balanced,0.3896563339233398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,16,4,balanced,2.2576217651367188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,16,4,balanced,0.5667609786987304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,16,4,balanced,0.7030553436279297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,16,4,balanced,1.0398258972167969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,16,4,balanced,1.3607257080078123
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,8,8,power_law_1.2,0.6331737518310547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,8,8,power_law_1.2,0.8589299011230469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,16,4,balanced,0.031371519565582276
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,16,4,balanced,0.03389695882797241
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,16,4,balanced,0.07188096046447753
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,16,4,balanced,0.03210880041122437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,16,4,balanced,0.07220223903656006
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,16,4,balanced,0.07186944007873534
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,16,4,balanced,0.03845119953155517
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,16,4,balanced,0.07261312007904053
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,16,4,balanced,0.07343103885650634
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,16,4,balanced,0.07374591827392578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,16,4,balanced,0.07399424076080321
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,16,4,balanced,0.07598591804504394
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,16,4,balanced,0.07669504165649414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,16,4,balanced,0.05179776191711426
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,16,4,balanced,0.08233087539672852
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,16,4,balanced,0.08847871780395508
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,16,4,balanced,0.09459199905395507
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,16,4,balanced,0.10517120361328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,16,4,balanced,0.11761919975280762
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,16,4,balanced,0.07090176105499267
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,16,4,power_law_1.01,0.03622015953063965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,16,4,power_law_1.01,0.06108672142028808
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,16,4,balanced,0.07991807937622071
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,16,4,power_law_1.01,0.07152512073516845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,16,4,power_law_1.01,0.08014719963073731
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,16,4,power_law_1.01,0.08475775718688965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,16,4,power_law_1.01,0.09880319595336914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,16,4,power_law_1.01,0.09723008155822754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,16,4,power_law_1.01,0.10199423789978028
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,16,4,power_law_1.01,0.16862592697143555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,16,4,power_law_1.01,0.16855808258056643
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,16,4,power_law_1.01,0.17508352279663086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,16,4,power_law_1.01,0.18008832931518554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,16,4,power_law_1.01,0.18311424255371095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,16,4,power_law_1.01,0.19041408538818358
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,16,4,power_law_1.01,0.19497983932495117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,16,4,power_law_1.01,0.20270208358764646
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,16,4,power_law_1.01,0.21697792053222656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,16,4,power_law_1.01,0.24888704299926756
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,16,4,power_law_1.01,0.27860992431640624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,16,4,power_law_1.01,0.3391936111450195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,16,4,power_law_1.01,0.394571533203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,16,4,power_law_1.01,0.507955207824707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,16,4,power_law_1.01,0.6265139389038086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,16,4,power_law_1.01,0.9185612487792969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,16,4,power_law_1.01,1.1767040252685548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,16,4,power_law_1.01,1.734544677734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,16,4,power_law_1.01,0.03909375905990601
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,16,4,power_law_1.01,0.0460646390914917
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,16,4,power_law_1.01,2.260367431640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,16,4,power_law_1.01,0.05695104122161865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,16,4,power_law_1.01,0.07392255783081055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,16,4,power_law_1.01,0.08072192192077636
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,16,4,power_law_1.01,0.09400320053100586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,16,4,power_law_1.01,0.10002431869506836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,16,4,power_law_1.01,0.09751551628112794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,16,4,power_law_1.01,0.104835844039917
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,16,4,balanced,0.07121407985687256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,16,4,power_law_1.01,0.11095040321350098
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,16,4,power_law_1.01,0.11073792457580567
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,16,4,power_law_1.01,0.11681023597717285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,16,4,power_law_1.01,0.11800319671630859
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,16,4,balanced,0.07193984031677246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,16,4,power_law_1.01,0.12590208053588867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,16,4,power_law_1.01,0.12795904159545898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,16,4,power_law_1.01,0.13058943748474122
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,16,4,power_law_1.01,0.14020352363586425
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,16,4,power_law_1.01,0.15767680168151857
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,16,4,power_law_1.01,0.18158464431762694
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,16,4,power_law_1.01,0.21965568542480468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,16,4,power_law_1.01,0.26353151321411133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,16,4,power_law_1.01,0.3414617538452148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,16,4,power_law_1.01,0.4308083343505859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,16,4,power_law_1.01,0.6101542282104492
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,16,4,power_law_1.01,0.0345356798171997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,16,4,power_law_1.01,0.7802304077148438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,16,4,power_law_1.01,0.03831295967102051
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,16,4,power_law_1.01,0.0435481595993042
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,16,4,power_law_1.01,1.1060070037841796
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,16,4,power_law_1.01,0.04935679912567138
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,16,4,power_law_1.01,0.056179199218749996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,16,4,power_law_1.01,0.058786559104919436
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,16,4,power_law_1.01,0.06844031810760498
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,16,4,power_law_1.01,0.06917632102966309
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,16,4,power_law_1.01,1.3930137634277344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,16,4,power_law_1.01,0.06897791862487793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,16,4,power_law_1.01,0.06860544204711913
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,16,4,power_law_1.01,0.0682700777053833
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,16,4,power_law_1.01,0.067325439453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,16,4,power_law_1.01,0.07259647846221924
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,16,4,power_law_1.01,0.06849152088165283
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,16,4,power_law_1.01,0.07382656097412109
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,16,4,power_law_1.01,0.07792511940002442
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,16,4,power_law_1.01,0.08226048469543457
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,16,4,balanced,0.20391679763793946
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,16,4,balanced,0.17912832260131836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,16,4,balanced,0.36785537719726563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,16,4,balanced,0.28549503326416015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,16,4,power_law_1.01,0.09581824302673339
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,16,4,power_law_1.01,0.10824192047119141
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,16,4,power_law_1.01,0.1313152027130127
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,16,4,power_law_1.01,0.16101503372192383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,16,4,power_law_1.2,0.03827584028244019
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,16,4,power_law_1.2,0.06123775959014892
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,16,4,power_law_1.2,0.06888319969177245
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,16,4,power_law_1.2,0.07875584125518799
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,16,4,power_law_1.2,0.08270336151123046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,16,4,power_law_1.2,0.10054656028747559
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,16,4,power_law_1.2,0.09903360366821289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,16,4,power_law_1.2,0.09934720039367675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,16,4,power_law_1.2,0.16818559646606446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,16,4,power_law_1.2,0.17018495559692384
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,16,4,power_law_1.2,0.17411327362060547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,16,4,power_law_1.2,0.17859840393066406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,16,4,power_law_1.2,0.18377344131469725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,16,4,power_law_1.2,0.19224191665649415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,16,4,power_law_1.2,0.19622655868530275
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,16,4,power_law_1.2,0.2044211196899414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,16,4,power_law_1.2,0.21926143646240231
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,16,4,power_law_1.2,0.25145343780517576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,16,4,power_law_1.2,0.2783488082885742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,16,4,power_law_1.2,0.342421760559082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,16,4,power_law_1.2,0.38962944030761715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,16,4,power_law_1.2,0.506728973388672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,16,4,power_law_1.2,0.6545676422119141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,16,4,power_law_1.2,0.9726105499267579
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,16,4,power_law_1.2,1.2219007873535157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,16,4,power_law_1.2,1.748209991455078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,16,4,power_law_1.2,0.039532799720764164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,16,4,power_law_1.2,0.04642687797546387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,16,4,power_law_1.2,0.05585023880004882
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,16,4,power_law_1.2,0.07281919956207275
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,16,4,power_law_1.2,2.2730111694335937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,16,4,power_law_1.2,0.08266752243041993
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,16,4,power_law_1.2,0.08789504051208497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,16,4,power_law_1.2,0.09488384246826172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,16,4,power_law_1.2,0.09948543548583985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,16,4,power_law_1.2,0.105730562210083
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,16,4,power_law_1.2,0.10502143859863282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,16,4,power_law_1.2,0.10887807846069335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,16,4,power_law_1.2,0.11231616020202637
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,16,4,power_law_1.01,0.22604032516479494
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,16,4,power_law_1.2,0.12164223670959473
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,16,4,power_law_1.2,0.12781184196472167
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,16,4,power_law_1.2,0.13020416259765627
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,16,4,power_law_1.01,0.27659263610839846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,16,4,power_law_1.2,0.13279232025146484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,16,4,power_law_1.01,0.4191487884521484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,16,4,power_law_1.2,0.1451968002319336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,16,4,power_law_1.2,0.1624025535583496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,16,4,power_law_1.2,0.188353271484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,16,4,power_law_1.2,0.2255833625793457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,16,4,power_law_1.2,0.35245567321777344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,16,4,power_law_1.2,0.43271041870117183
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,16,4,power_law_1.2,0.6143641662597656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,16,4,power_law_1.01,0.49283329010009763
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,16,4,power_law_1.2,0.034588160514831545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,16,4,power_law_1.2,0.2655001640319824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,16,4,power_law_1.2,0.7906559753417969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,16,4,power_law_1.2,0.04103040218353272
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,16,4,power_law_1.2,0.0472870397567749
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,16,4,power_law_1.2,0.05367936134338379
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,16,4,power_law_1.2,0.057528319358825686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,16,4,power_law_1.2,1.154882583618164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,16,4,power_law_1.2,0.06564735889434814
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,16,4,power_law_1.2,0.07015039920806884
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,16,4,power_law_1.2,0.06439680099487305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,16,4,power_law_1.2,0.06879104137420654
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,16,4,power_law_1.2,0.06909567832946777
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,16,4,power_law_1.2,0.07013376235961914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,16,4,power_law_1.2,0.07041791915893555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,16,4,power_law_1.2,0.07442048072814941
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,16,4,power_law_1.2,0.07628032207489013
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,16,4,power_law_1.2,0.08093440055847168
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,16,4,power_law_1.2,0.03861759901046753
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,16,4,power_law_1.2,0.0926425552368164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,16,4,power_law_1.2,0.11261568069458008
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,16,4,power_law_1.2,0.13018112182617186
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,16,4,power_law_1.2,0.3703142547607422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,16,4,power_law_1.2,1.4832357788085937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,16,4,power_law_1.01,0.760400619506836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,16,4,power_law_1.01,0.9941337585449219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,32,2,balanced,0.03293312072753906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,32,2,balanced,0.03584127902984619
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,32,2,balanced,0.054928641319274905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,32,2,balanced,0.07278848171234131
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,32,2,balanced,0.0976473617553711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,32,2,balanced,0.13278592109680176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,32,2,balanced,0.12615039825439453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,32,2,balanced,0.12772607803344727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,32,2,balanced,0.12903552055358886
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,32,2,balanced,0.13018367767333985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,32,2,balanced,0.13352959632873534
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,32,2,balanced,0.22502784729003905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,32,2,balanced,0.2289331245422363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,32,2,balanced,0.23638656616210935
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,32,2,balanced,0.23983743667602536
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,32,2,balanced,0.24767871856689455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,32,2,balanced,0.26132736206054685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,32,2,balanced,0.29026687622070313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,32,2,balanced,0.3170060729980469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,32,2,balanced,0.37182590484619144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,32,2,balanced,0.43117950439453123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,32,2,balanced,0.554595832824707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,32,2,balanced,0.6659455871582031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,32,2,balanced,0.9012569427490235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,32,2,balanced,1.1952166748046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,32,2,balanced,1.7190118408203126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,32,2,balanced,2.298138885498047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,16,4,power_law_1.2,0.15922816276550295
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,16,4,power_law_1.2,0.3206412887573242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,16,4,power_law_1.2,0.1994393539428711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,32,2,balanced,0.035786240100860595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,32,2,balanced,0.03947135925292969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,32,2,balanced,0.04869887828826904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,32,2,balanced,0.0661683177947998
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,16,4,power_law_1.2,0.5907519912719726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,32,2,balanced,0.11071871757507323
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,32,2,balanced,0.15136128425598144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,32,2,balanced,0.15307135581970216
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,32,2,balanced,0.1561689567565918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,32,2,balanced,0.1543628787994385
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,32,2,balanced,0.15912447929382326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,32,2,balanced,0.18656639099121092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,32,2,balanced,0.1891289520263672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,32,2,balanced,0.19370624542236328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,32,2,balanced,0.19280384063720704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,32,2,balanced,0.19789440155029298
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,16,4,power_law_1.2,0.7331314849853515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,32,2,balanced,0.20552064895629885
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,32,2,balanced,0.2227097511291504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,32,2,balanced,0.23861248016357423
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,32,2,balanced,0.2729715156555176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,32,2,balanced,0.3064780807495117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,32,2,balanced,0.37198848724365235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,32,2,balanced,0.15545599937438964
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,32,2,balanced,0.4367142486572265
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,16,4,power_law_1.2,1.2358195495605468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,32,2,balanced,0.6831501007080079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,32,2,balanced,0.8161971282958984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,32,2,balanced,1.1929625701904296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,32,2,balanced,1.5721830749511718
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,16,4,power_law_1.2,1.3904806518554689
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,32,2,balanced,0.04610432147979736
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,32,2,balanced,0.04524415969848633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,32,2,balanced,0.11174400329589844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,32,2,balanced,0.11256575584411621
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,32,2,balanced,0.11277055740356445
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,32,2,balanced,0.11503487586975099
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,32,2,balanced,0.05999872207641601
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,32,2,balanced,0.05100927829742432
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,32,2,balanced,0.1150054359436035
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,32,2,balanced,0.11684096336364747
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,32,2,balanced,0.11795455932617187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,32,2,balanced,0.11940735816955565
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,32,2,balanced,0.12280960083007812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,32,2,balanced,0.07850111961364746
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,32,2,balanced,0.1297651195526123
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,32,2,balanced,0.13547648429870604
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,32,2,balanced,0.14615936279296876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,32,2,balanced,0.1570521640777588
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,32,2,balanced,0.1822822380065918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,32,2,balanced,0.20695808410644528
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,32,2,balanced,0.11152640342712403
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,32,2,power_law_1.01,0.034955520629882816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,32,2,power_law_1.01,0.0390720009803772
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,32,2,power_law_1.01,0.06755712032318115
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,32,2,power_law_1.01,0.07787007808685302
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,32,2,power_law_1.01,0.09339008331298829
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,32,2,balanced,0.11075455665588378
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,32,2,power_law_1.01,0.10749567985534667
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,32,2,power_law_1.01,0.1136128044128418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,32,2,power_law_1.01,0.11577088356018066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,32,2,power_law_1.01,0.11652864456176756
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,32,2,power_law_1.01,0.11927552223205566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,32,2,power_law_1.01,0.12517888069152833
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,32,2,power_law_1.01,0.2214131164550781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,32,2,power_law_1.01,0.22526720046997067
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,32,2,power_law_1.01,0.23450239181518553
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,32,2,power_law_1.01,0.23698303222656253
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,32,2,power_law_1.01,0.24612607955932617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,32,2,power_law_1.01,0.2611520004272461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,32,2,power_law_1.01,0.2922368049621582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,32,2,power_law_1.01,0.3209804916381836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,32,2,power_law_1.01,0.3815987014770508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,32,2,power_law_1.01,0.4402956771850586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,32,2,power_law_1.01,0.5636313629150391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,32,2,power_law_1.01,0.6898406219482422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,32,2,power_law_1.01,0.9677503967285157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,32,2,power_law_1.01,1.2565670776367188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,32,2,power_law_1.01,0.038229761123657224
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,32,2,power_law_1.01,0.045698561668396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,32,2,power_law_1.01,1.7781709289550782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,32,2,power_law_1.01,0.054376959800720215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,32,2,power_law_1.01,0.07724671840667725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,32,2,power_law_1.01,0.10753279685974122
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,32,2,power_law_1.01,0.1214412784576416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,32,2,power_law_1.01,0.13370880126953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,32,2,power_law_1.01,2.3072050476074217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,32,2,power_law_1.01,0.1359615993499756
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,32,2,power_law_1.01,0.13771007537841798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,32,2,power_law_1.01,0.13882752418518068
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,32,2,power_law_1.01,0.1458803176879883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,32,2,power_law_1.01,0.16770048141479493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,32,2,power_law_1.01,0.1703615951538086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,32,2,power_law_1.01,0.17786495208740233
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,32,2,power_law_1.01,0.17898880004882814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,32,2,power_law_1.01,0.18102272033691408
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,32,2,power_law_1.01,0.19649408340454103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,32,2,power_law_1.01,0.2138649559020996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,32,2,power_law_1.01,0.2401817512512207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,32,2,power_law_1.01,0.29742080688476563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,32,2,power_law_1.01,0.3508403015136719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,32,2,power_law_1.01,0.44313983917236327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,32,2,power_law_1.01,0.5339814376831055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,32,2,power_law_1.01,0.7200819396972656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,32,2,power_law_1.01,0.9092262268066407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,32,2,power_law_1.01,0.047540478706359864
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,32,2,power_law_1.01,0.05060991764068603
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,32,2,balanced,0.11207807540893555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,32,2,power_law_1.01,1.3167462158203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,32,2,power_law_1.01,0.05630464076995849
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,32,2,power_law_1.01,0.06528512001037598
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,32,2,power_law_1.01,0.07469183921813964
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,32,2,power_law_1.01,0.08017151832580567
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,32,2,power_law_1.01,1.6812287902832033
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,32,2,power_law_1.01,0.09173503875732422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,32,2,power_law_1.01,0.0969753646850586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,32,2,power_law_1.01,0.09765376091003418
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,32,2,power_law_1.01,0.10293248176574707
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,32,2,power_law_1.01,0.10646783828735351
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,32,2,power_law_1.01,0.11362303733825682
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,32,2,power_law_1.01,0.11851648330688476
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,32,2,power_law_1.01,0.1255078411102295
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,32,2,power_law_1.01,0.12804736137390135
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,32,2,power_law_1.01,0.13523455619812014
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,32,2,power_law_1.01,0.10363007545471192
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,32,2,balanced,0.5363251113891602
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,32,2,balanced,0.3251468658447266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,32,2,balanced,0.6951436614990234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,32,2,balanced,0.374835205078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,32,2,power_law_1.01,0.1775168037414551
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,32,2,power_law_1.01,0.18818304061889649
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,32,2,power_law_1.01,0.23558271408081052
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,32,2,power_law_1.2,0.03458944082260132
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,32,2,power_law_1.2,0.038695681095123294
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,32,2,power_law_1.01,0.3004287910461426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,32,2,power_law_1.2,0.06638336181640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,32,2,power_law_1.2,0.07526527881622315
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,32,2,power_law_1.2,0.08848383903503418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,32,2,power_law_1.2,0.10234496116638184
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,32,2,power_law_1.2,0.11248512268066406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,32,2,power_law_1.2,0.11421824455261231
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,32,2,power_law_1.2,0.11645952224731446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,32,2,power_law_1.2,0.1194611167907715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,32,2,power_law_1.2,0.1267955207824707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,32,2,power_law_1.2,0.2218764877319336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,32,2,power_law_1.2,0.2270182418823242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,32,2,power_law_1.2,0.23458303451538084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,32,2,power_law_1.2,0.2398361587524414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,32,2,power_law_1.2,0.24698112487792967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,32,2,power_law_1.2,0.2623910331726074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,32,2,power_law_1.2,0.29369344711303713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,32,2,power_law_1.2,0.3241292953491211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,32,2,power_law_1.2,0.38445568084716797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,32,2,power_law_1.2,0.44745342254638676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,32,2,power_law_1.2,0.5673574447631836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,32,2,power_law_1.2,0.6942463684082031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,32,2,power_law_1.2,0.9808870697021485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,32,2,power_law_1.2,1.2668876647949219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,32,2,power_law_1.2,1.808503112792969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,32,2,power_law_1.2,0.038309121131896974
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,32,2,power_law_1.2,0.045646080970764155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,32,2,power_law_1.2,0.05428736209869385
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,32,2,power_law_1.2,2.3572108459472654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,32,2,power_law_1.2,0.07400320053100587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,32,2,power_law_1.2,0.10102272033691406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,32,2,power_law_1.2,0.11857664108276367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,32,2,power_law_1.2,0.13156736373901368
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,32,2,power_law_1.2,0.13507840156555176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,32,2,power_law_1.2,0.13864319801330566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,32,2,power_law_1.2,0.13912447929382324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,32,2,power_law_1.2,0.14676351547241212
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,32,2,power_law_1.2,0.16762367248535157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,32,2,power_law_1.2,0.17320192337036133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,32,2,power_law_1.2,0.18271871566772463
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,32,2,power_law_1.2,0.18642431259155273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,32,2,power_law_1.2,0.18324607849121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,32,2,power_law_1.2,0.2051968002319336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,32,2,power_law_1.2,0.22335744857788087
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,32,2,power_law_1.2,0.24814079284667967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,32,2,power_law_1.2,0.3186675262451172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,32,2,power_law_1.2,0.3654694366455078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,32,2,power_law_1.2,0.45490688323974615
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,32,2,power_law_1.2,0.5434022521972656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,32,2,power_law_1.2,0.7430528259277344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,32,2,power_law_1.01,0.4006003189086914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,32,2,power_law_1.01,0.5254848098754883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,32,2,power_law_1.2,0.927646713256836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,32,2,power_law_1.2,0.047356162071228024
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,32,2,power_law_1.2,0.05054336071014405
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,32,2,power_law_1.2,0.055287041664123536
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,32,2,power_law_1.2,0.06324351787567138
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,32,2,power_law_1.2,1.3481497192382812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,32,2,power_law_1.2,0.06873472213745117
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,32,2,power_law_1.2,0.07591423988342286
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,32,2,power_law_1.2,0.09391743659973144
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,32,2,power_law_1.01,0.8458649444580079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,32,2,power_law_1.2,1.6926899719238282
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,32,2,power_law_1.2,0.0951756763458252
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,32,2,power_law_1.2,0.09818495750427246
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,32,2,power_law_1.2,0.10095744132995606
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,32,2,power_law_1.2,0.10457344055175781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,32,2,power_law_1.2,0.11009407997131349
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,32,2,power_law_1.2,0.12006784439086915
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,32,2,power_law_1.2,0.12676095962524414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,32,2,power_law_1.2,0.13417728424072267
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,32,2,power_law_1.2,0.11278464317321776
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,32,2,power_law_1.01,0.8901222229003907
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,32,2,power_law_1.01,1.4777574157714843
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,32,2,power_law_1.01,1.825410614013672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,32,2,power_law_1.2,0.1590617561340332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,32,2,power_law_1.2,0.2017190361022949
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,32,2,power_law_1.2,0.3103167915344238
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,32,2,power_law_1.2,0.3735091018676758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,64,balanced,0.04076543807983398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,64,balanced,0.04041984081268311
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,32,2,power_law_1.2,0.6011417770385743
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,64,balanced,0.04004992008209228
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,64,balanced,0.038254079818725584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,64,balanced,0.07099008083343505
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,64,balanced,0.08504192352294922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,64,balanced,0.13373824119567873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,64,balanced,0.13414912223815917
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,64,balanced,0.13667967796325683
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,64,balanced,0.13694080352783203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,32,2,power_law_1.2,0.23281408309936524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,64,balanced,0.13779328346252442
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,64,balanced,0.14065792083740233
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,64,balanced,0.1439641571044922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,64,balanced,0.1413145637512207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,64,balanced,0.1486207962036133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,64,balanced,0.15046272277832032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,64,balanced,0.15361663818359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,64,balanced,0.1611302375793457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,64,balanced,0.15668224334716796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,64,balanced,0.17125631332397462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,64,balanced,0.17765632629394532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,64,balanced,0.20413312911987305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,64,balanced,0.21684223175048828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,64,balanced,0.2995968055725098
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,64,balanced,0.3590374374389648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,64,balanced,0.4949619293212891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,64,balanced,0.03850111961364746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,64,balanced,0.6132633590698242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,64,balanced,0.03878015995025635
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,64,balanced,0.03707135915756225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,64,balanced,0.041356801986694336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,64,balanced,0.05440383911132812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,64,balanced,0.05876351833343506
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,64,balanced,0.058824958801269533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,64,balanced,0.060245761871337886
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,64,balanced,0.06085247993469238
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,64,balanced,0.06098176002502441
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,64,balanced,0.0640934419631958
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,64,balanced,0.06286719799041748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,64,balanced,0.06790143966674804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,64,balanced,0.06540800094604492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,64,balanced,0.03583231925964355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,64,balanced,0.06724480152130127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,64,balanced,0.07128575801849366
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,64,balanced,0.07881855964660645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,64,balanced,0.07132031917572021
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,64,balanced,0.0790553617477417
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,64,balanced,0.08645888328552245
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,64,balanced,0.09486975669860839
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,64,balanced,0.10630528450012207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,64,balanced,0.14027520179748537
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,64,balanced,0.16428159713745116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,64,balanced,0.230883846282959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,64,balanced,0.28252544403076174
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,32,2,power_law_1.2,0.7403020477294922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,32,2,power_law_1.2,1.1420095825195313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,32,2,power_law_1.2,1.3786277770996094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,32,2,power_law_1.2,2.134316864013672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,32,2,power_law_1.2,3.1697894287109376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,1,64,balanced,0.020368640422821042
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,1,64,balanced,0.01985919952392578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,1,64,balanced,0.019879679679870605
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,1,64,balanced,0.029208319187164305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,1,64,balanced,0.028314878940582277
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,1,64,balanced,0.02900991916656494
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,1,64,balanced,0.029730560779571535
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,1,64,balanced,0.028725759983062742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,1,64,balanced,0.02935551881790161
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,1,64,balanced,0.028787200450897214
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,1,64,balanced,0.030154240131378175
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,1,64,balanced,0.029771521091461185
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,1,64,balanced,0.030557439327239987
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,1,64,balanced,0.03044095993041992
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,1,64,balanced,0.030982398986816408
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,1,64,balanced,0.031746559143066407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,1,64,balanced,0.03346688032150268
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,1,64,balanced,0.030035200119018557
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,1,64,balanced,0.01988479971885681
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,1,64,balanced,0.020341761112213134
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,64,power_law_1.01,0.08462592124938964
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,64,power_law_1.01,0.08458751678466797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,64,power_law_1.01,0.08605695724487304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,64,power_law_1.01,0.08575360298156738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,64,power_law_1.01,0.13217023849487303
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,64,power_law_1.01,0.1378124809265137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,64,power_law_1.01,0.13827712059020997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,64,power_law_1.01,0.1432140827178955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,64,power_law_1.01,0.14413056373596192
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,64,power_law_1.01,0.14524928092956543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,64,power_law_1.01,0.14354047775268555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,64,power_law_1.01,0.14969728469848634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,64,power_law_1.01,0.15416064262390136
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,64,power_law_1.01,0.15852031707763672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,64,power_law_1.01,0.16126207351684568
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,64,power_law_1.01,0.1668147277832031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,64,power_law_1.01,0.16603776931762695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,64,power_law_1.01,0.18879104614257813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,64,power_law_1.01,0.1971571159362793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,64,power_law_1.01,0.2712499237060547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,64,power_law_1.01,0.3361881637573242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,64,power_law_1.01,0.4849177551269531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,64,power_law_1.01,0.5873779296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,64,power_law_1.01,0.7969612884521484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,64,power_law_1.01,1.1269951629638673
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,64,power_law_1.01,1.3855845642089843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,64,power_law_1.01,1.9622784423828126
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,1,64,balanced,0.030867199897766113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,64,power_law_1.01,0.05478400230407715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,64,power_law_1.01,0.055057921409606934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,64,power_law_1.01,0.0557478380203247
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,64,power_law_1.01,0.055429120063781735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,64,power_law_1.01,0.05756288051605225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,64,power_law_1.01,0.060349440574645995
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,1,64,balanced,0.035462400913238525
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,64,power_law_1.01,0.06197120189666748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,64,power_law_1.01,0.06204288005828858
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,64,power_law_1.01,0.06763391971588135
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,64,power_law_1.01,0.06378623962402344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,64,power_law_1.01,0.06751999855041504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,64,power_law_1.01,0.06806655883789062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,64,power_law_1.01,0.06983039855957032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,64,power_law_1.01,0.07570943832397461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,1,64,balanced,0.03636480093002319
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,64,power_law_1.01,0.0754585599899292
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,64,power_law_1.01,0.07913856029510498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,64,power_law_1.01,0.08967935562133789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,64,power_law_1.01,0.09696640014648437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,64,power_law_1.01,0.1207590389251709
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,64,power_law_1.01,0.15884544372558593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,64,power_law_1.01,0.19186048507690429
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,64,power_law_1.01,0.08105216026306153
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,64,power_law_1.01,0.2632473564147949
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,64,power_law_1.01,0.3649190521240234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,64,power_law_1.01,0.4658355331420898
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,1,64,power_law_1.01,0.028776960372924803
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,64,power_law_1.01,0.6132454299926758
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,1,64,power_law_1.01,0.027243518829345705
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,1,64,power_law_1.01,0.028980479240417478
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,1,64,power_law_1.01,0.028860158920288086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,64,power_law_1.01,0.8894937896728516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,1,64,power_law_1.01,0.024062719345092774
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,1,64,power_law_1.01,0.02678656101226807
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,1,64,power_law_1.01,0.02739840030670166
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,1,64,power_law_1.01,0.026087679862976075
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,1,64,power_law_1.01,0.029239039421081543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,1,64,power_law_1.01,0.026801919937133788
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,1,64,power_law_1.01,0.027601919174194335
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,1,64,power_law_1.01,0.02934015989303589
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,1,64,power_law_1.01,0.030141439437866215
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,1,64,power_law_1.01,0.02927488088607788
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,1,64,power_law_1.01,0.028185598850250244
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,1,64,power_law_1.01,0.030416638851165773
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,1,64,power_law_1.01,0.028302080631256103
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,1,64,balanced,0.047397122383117676
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,1,64,balanced,0.04484992027282715
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,1,64,balanced,0.05933440208435059
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,1,64,balanced,0.07253759860992431
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,1,64,power_law_1.01,0.04076543807983398
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,1,64,power_law_1.01,0.045114879608154294
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,1,64,power_law_1.01,0.03696000099182129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,64,power_law_1.2,0.08527104377746582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,64,power_law_1.2,0.08511232376098633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,1,64,power_law_1.01,0.047837438583374026
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,64,power_law_1.2,0.0859443187713623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,64,power_law_1.2,0.1309286403656006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,64,power_law_1.2,0.13380352020263672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,64,power_law_1.2,0.13472000122070313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,64,power_law_1.2,0.14084480285644532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,64,power_law_1.2,0.13819647789001466
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,64,power_law_1.2,0.14389375686645506
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,64,power_law_1.2,0.1431603240966797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,64,power_law_1.2,0.14729344367980957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,64,power_law_1.2,0.15193471908569336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,64,power_law_1.2,0.1540480041503906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,64,power_law_1.2,0.16104959487915038
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,64,power_law_1.2,0.16705280303955078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,1,64,power_law_1.01,0.06998655796051026
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,64,power_law_1.2,0.17167104721069334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,1,64,power_law_1.01,0.07896192073822021
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,64,power_law_1.2,0.17231359481811523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,64,power_law_1.2,0.2137446403503418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,64,power_law_1.2,0.25271167755126955
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,1,64,power_law_1.01,0.10246015548706056
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,64,power_law_1.2,0.326058235168457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,64,power_law_1.2,0.41559806823730466
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,64,power_law_1.2,0.6342233657836914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,64,power_law_1.2,0.7898636627197265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,64,power_law_1.2,0.054585599899291994
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,1,64,power_law_1.01,0.12753791809082032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,64,power_law_1.2,1.149813766479492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,64,power_law_1.2,0.05441535949707031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,64,power_law_1.2,0.05516799926757813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,64,power_law_1.2,0.05764095783233643
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,64,power_law_1.2,1.5175398254394532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,64,power_law_1.2,0.05824512004852295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,64,power_law_1.2,0.05931263923645019
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,64,power_law_1.2,0.06018176078796387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,64,power_law_1.2,0.06271103858947755
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,64,power_law_1.2,0.06251264095306397
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,64,power_law_1.2,0.0664844799041748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,64,power_law_1.2,0.06498176097869873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,64,power_law_1.2,0.0681113576889038
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,64,power_law_1.2,0.07138815879821778
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,64,power_law_1.2,0.07840640068054198
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,64,power_law_1.2,0.07922560214996338
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,64,power_law_1.2,2.3406745910644533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,64,power_law_1.2,0.07769472122192382
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,64,power_law_1.2,0.08397695541381836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,64,power_law_1.2,0.0944934368133545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,64,power_law_1.2,0.10537216186523438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,64,power_law_1.2,0.14233216285705566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,64,power_law_1.2,0.17235071182250977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,64,power_law_1.2,0.2519910430908203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,64,power_law_1.2,0.3190566444396973
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,64,power_law_1.2,0.522750701904297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,1,64,power_law_1.2,0.027779839038848876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,64,power_law_1.2,0.7116339111328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,1,64,power_law_1.2,0.02789504051208496
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,1,64,power_law_1.2,0.029141759872436522
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,64,power_law_1.2,3.676252136230469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,1,64,power_law_1.2,0.02516223907470703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,64,power_law_1.2,1.092968978881836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,1,64,power_law_1.2,0.025100800991058346
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,1,64,power_law_1.2,0.026912000179290775
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,1,64,power_law_1.2,0.028034560680389405
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,1,64,power_law_1.2,0.02702591896057129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,64,power_law_1.2,1.4594163513183593
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,1,64,power_law_1.2,0.02665855884552002
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,1,64,power_law_1.2,0.028285439014434817
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,1,64,power_law_1.2,0.029081599712371824
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,1,64,power_law_1.2,0.029079039096832276
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,1,64,power_law_1.2,0.0294323205947876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,1,64,power_law_1.2,0.029771521091461185
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,1,64,power_law_1.2,0.04073984146118164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,1,64,power_law_1.2,0.024863998889923095
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,1,64,power_law_1.2,0.03029632091522217
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,1,64,power_law_1.01,0.21942016601562503
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,1,64,power_law_1.01,0.21448703765869143
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,1,64,power_law_1.2,0.03310847997665405
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,1,64,power_law_1.2,0.04444416046142578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,1,64,power_law_1.2,0.060117759704589845
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,1,64,power_law_1.2,0.05237504005432129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,2,32,balanced,0.03752703905105591
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,2,32,balanced,0.03246975898742675
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,1,64,power_law_1.2,0.07644927978515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,2,32,balanced,0.036652801036834715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,2,32,balanced,0.052330241203308106
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,2,32,balanced,0.06649983882904052
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,2,32,balanced,0.08351103782653808
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,2,32,balanced,0.1276364803314209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,2,32,balanced,0.1288755226135254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,2,32,balanced,0.12900351524353026
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,2,32,balanced,0.13104640007019044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,2,32,balanced,0.13350527763366699
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,1,64,power_law_1.2,0.08983679771423339
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,2,32,balanced,0.13574527740478515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,2,32,balanced,0.14051199913024903
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,2,32,balanced,0.14539135932922362
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,2,32,balanced,0.15014271736145018
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,2,32,balanced,0.14377087593078614
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,2,32,balanced,0.15421952247619627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,2,32,balanced,0.16400384902954102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,2,32,balanced,0.17933567047119142
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,2,32,balanced,0.1898406410217285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,2,32,balanced,0.23412864685058593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,2,32,balanced,0.2620774459838867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,2,32,balanced,0.33792385101318356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,2,32,balanced,0.13766783714294434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,2,32,balanced,0.033756160736083986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,2,32,balanced,0.4434534454345703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,2,32,balanced,0.03379712104797363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,2,32,balanced,0.03297280073165894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,2,32,balanced,0.6180249786376952
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,2,32,balanced,0.03430527925491333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,2,32,balanced,0.040332798957824705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,2,32,balanced,0.05697152137756347
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,2,32,balanced,0.7920806121826172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,2,32,balanced,0.06779647827148437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,2,32,balanced,0.06650879859924316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,2,32,balanced,0.06613632202148437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,2,32,balanced,0.06594816207885743
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,2,32,balanced,0.0678656005859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,2,32,balanced,0.06964735984802246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,2,32,balanced,0.07159167766571045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,2,32,balanced,0.07253888130187988
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,2,32,balanced,0.07964672088623047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,2,32,balanced,0.07013887882232665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,2,32,balanced,0.07663487911224365
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,2,32,balanced,0.08251775741577148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,2,32,balanced,0.0876748752593994
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,2,32,balanced,0.0986086368560791
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,2,32,balanced,0.11853183746337889
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,2,32,balanced,0.13779711723327637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,2,32,balanced,0.18850048065185548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,2,32,balanced,0.22500608444213865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,2,32,balanced,0.07453055858612061
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,2,32,balanced,0.41187072753906245
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,2,32,balanced,0.32513534545898437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,1,64,power_law_1.2,0.14427392005920409
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,1,64,power_law_1.2,0.18363136291503906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,1,64,power_law_1.2,0.3618368148803711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,1,64,power_law_1.2,0.41677055358886717
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,2,32,balanced,0.019874559640884398
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,2,32,balanced,0.03175551891326904
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,2,32,balanced,0.031825919151306156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,2,32,balanced,0.032048640251159666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,2,32,balanced,0.01938048005104065
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,2,32,balanced,0.019813120365142822
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,2,32,balanced,0.032316160202026364
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,2,32,balanced,0.03161600112915039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,2,32,balanced,0.03150207996368408
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,2,32,balanced,0.032459518909454345
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,2,32,balanced,0.03172096014022827
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,2,32,balanced,0.03161216020584107
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,2,32,balanced,0.03280384063720703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,2,32,balanced,0.031306240558624264
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,2,32,balanced,0.03252480030059814
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,2,32,balanced,0.034410240650177004
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,2,32,balanced,0.03625600099563599
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,2,32,balanced,0.0375385594367981
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,2,32,balanced,0.019900159835815428
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,2,32,balanced,0.03178112030029297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,2,32,balanced,0.022355198860168457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,2,32,power_law_1.01,0.08169343948364258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,2,32,power_law_1.01,0.08124287605285645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,2,32,power_law_1.01,0.08393343925476074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,2,32,power_law_1.01,0.0837491226196289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,2,32,power_law_1.01,0.12371711730957032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,2,32,power_law_1.01,0.12149248123168946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,2,32,power_law_1.01,0.12901503562927247
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,2,32,power_law_1.01,0.12728832244873048
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,2,32,power_law_1.01,0.13451775550842285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,2,32,power_law_1.01,0.13212672233581543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,2,32,power_law_1.01,0.13757951736450197
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,2,32,power_law_1.01,0.14006400108337402
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,2,32,power_law_1.01,0.14556544303894042
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,2,32,power_law_1.01,0.1533248043060303
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,2,32,power_law_1.01,0.14942079544067383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,2,32,power_law_1.01,0.1493990421295166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,2,32,power_law_1.01,0.1564812755584717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,2,32,power_law_1.01,0.18571136474609376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,2,32,power_law_1.01,0.19752704620361328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,2,32,power_law_1.01,0.2355251121520996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,2,32,power_law_1.01,0.3112128067016601
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,2,32,power_law_1.01,0.4270425415039062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,2,32,power_law_1.01,0.454469108581543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,2,32,power_law_1.01,0.6720358276367188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,2,32,power_law_1.01,0.8731455993652343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,2,32,power_law_1.01,1.3562905883789063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,2,32,power_law_1.01,0.05494783878326416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,2,32,power_law_1.01,0.055637760162353514
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,2,32,power_law_1.01,1.7844671630859374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,2,32,power_law_1.01,0.05646592140197754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,2,32,power_law_1.01,0.05842432022094727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,2,32,power_law_1.01,0.05972224235534668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,2,32,power_law_1.01,0.056894721984863283
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,2,32,power_law_1.01,0.06285056114196777
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,2,32,power_law_1.01,0.06606719970703126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,2,32,power_law_1.01,0.06408319950103761
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,2,32,power_law_1.01,0.06614528179168702
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,2,32,power_law_1.01,0.06595071792602539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,2,32,power_law_1.01,0.0704038381576538
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,2,32,balanced,0.030201599597930905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,2,32,power_law_1.01,0.0723583984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,2,32,power_law_1.01,0.07788159847259521
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,2,32,power_law_1.01,0.07330175876617431
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,2,32,power_law_1.01,0.07588096141815186
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,2,32,balanced,0.031063039302825928
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,2,32,power_law_1.01,0.07924352169036866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,2,32,power_law_1.01,0.08561535835266113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,2,32,power_law_1.01,0.10179967880249023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,2,32,power_law_1.01,0.12255359649658204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,2,32,power_law_1.01,0.14688896179199218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,2,32,power_law_1.01,0.19614591598510742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,2,32,power_law_1.01,0.25830528259277347
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,2,32,power_law_1.01,0.31298303604125977
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,2,32,power_law_1.01,0.03097471952438354
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,2,32,power_law_1.01,0.43696128845214843
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,2,32,power_law_1.01,0.031381759643554684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,2,32,power_law_1.01,0.6368204879760742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,2,32,power_law_1.01,0.030504961013793946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,2,32,power_law_1.01,0.8009497833251953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,2,32,power_law_1.01,0.031217920780181884
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,2,32,power_law_1.01,0.02723072052001953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,2,32,power_law_1.01,0.028587520122528076
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,2,32,power_law_1.01,0.02891263961791992
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,2,32,power_law_1.01,0.029000959396362304
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,2,32,power_law_1.01,0.02859519958496094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,2,32,power_law_1.01,0.030588159561157225
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,2,32,power_law_1.01,0.029509119987487793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,2,32,power_law_1.01,0.030529279708862305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,2,32,power_law_1.01,0.03097471952438354
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,2,32,power_law_1.01,0.03049344062805176
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,2,32,power_law_1.01,0.028350720405578612
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,2,32,power_law_1.01,0.02926208019256592
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,2,32,power_law_1.01,0.03099136114120483
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,2,32,balanced,0.05090559959411621
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,2,32,balanced,0.05502975940704345
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,2,32,balanced,0.07060991764068604
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,2,32,balanced,0.0851699161529541
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,2,32,power_law_1.01,0.03703039884567261
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,2,32,power_law_1.01,0.039696640968322754
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,2,32,power_law_1.01,0.04794879913330078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,2,32,power_law_1.01,0.05236480236053467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,2,32,power_law_1.2,0.08062848091125488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,2,32,power_law_1.2,0.08128767967224121
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,2,32,power_law_1.01,0.06725120067596435
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,2,32,power_law_1.2,0.08311424255371094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,2,32,power_law_1.2,0.12054528236389159
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,2,32,power_law_1.2,0.12271488189697266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,2,32,power_law_1.01,0.10025983810424806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,2,32,power_law_1.2,0.12776191711425783
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,2,32,power_law_1.01,0.07919104099273681
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,2,32,power_law_1.2,0.1276915168762207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,2,32,power_law_1.2,0.12950400352478025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,2,32,power_law_1.2,0.08221311569213867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,2,32,power_law_1.2,0.13251968383789064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,2,32,power_law_1.2,0.13813119888305664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,2,32,power_law_1.2,0.1420582389831543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,2,32,power_law_1.2,0.1473318386077881
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,2,32,power_law_1.2,0.15064064025878907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,2,32,power_law_1.2,0.15065216064453124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,2,32,power_law_1.2,0.15980416297912597
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,2,32,power_law_1.2,0.1615065574645996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,2,32,power_law_1.2,0.19103744506835937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,2,32,power_law_1.2,0.22709632873535157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,2,32,power_law_1.2,0.27785984039306644
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,2,32,power_law_1.01,0.12247424125671387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,2,32,power_law_1.2,0.3532223892211914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,2,32,power_law_1.2,0.5077324676513671
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,2,32,power_law_1.2,0.5744486236572266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,2,32,power_law_1.2,0.05615231990814209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,2,32,power_law_1.2,0.830350112915039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,2,32,power_law_1.2,0.05612287998199463
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,2,32,power_law_1.2,0.05605631828308105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,2,32,power_law_1.2,1.3070413208007812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,2,32,power_law_1.2,0.056619520187377925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,2,32,power_law_1.2,0.06361599922180175
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,2,32,power_law_1.2,0.05932544231414795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,2,32,power_law_1.2,0.06166783809661865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,2,32,power_law_1.2,0.0625484800338745
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,2,32,power_law_1.2,0.06407296180725097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,2,32,power_law_1.2,0.06720640182495116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,2,32,power_law_1.2,0.07088640213012695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,2,32,power_law_1.2,2.019390716552734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,2,32,power_law_1.2,0.07390719890594483
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,2,32,power_law_1.2,0.07598336219787598
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,2,32,power_law_1.2,0.07534848213195801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,2,32,power_law_1.2,0.078090238571167
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,2,32,power_law_1.2,0.08260095596313477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,2,32,power_law_1.2,0.09671039581298828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,2,32,power_law_1.2,0.10592000007629394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,2,32,power_law_1.2,2.4013247680664063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,2,32,power_law_1.2,0.13793536186218264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,2,32,power_law_1.2,0.16441984176635743
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,2,32,power_law_1.2,0.05948800086975098
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,2,32,power_law_1.2,0.2421004867553711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,2,32,power_law_1.2,0.26523391723632816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,2,32,power_law_1.2,0.39247745513916016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,2,32,power_law_1.2,0.03059200048446655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,2,32,power_law_1.2,0.560021743774414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,2,32,power_law_1.2,0.0315827202796936
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,2,32,power_law_1.2,0.030599679946899418
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,2,32,power_law_1.2,0.028679680824279786
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,2,32,power_law_1.2,0.8586521911621094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,2,32,power_law_1.2,0.027502079010009766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,2,32,power_law_1.2,0.02918272018432617
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,2,32,power_law_1.2,0.02877952098846436
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,2,32,power_law_1.2,0.029615359306335447
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,2,32,power_law_1.2,0.02605952024459839
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,2,32,power_law_1.2,0.02969599962234497
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,2,32,power_law_1.2,0.028040959835052493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,2,32,power_law_1.2,1.1447821044921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,2,32,power_law_1.2,0.03091200113296509
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,2,32,power_law_1.2,0.02949631929397583
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,2,32,power_law_1.2,0.030687999725341798
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,2,32,power_law_1.2,0.03085311889648438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,2,32,power_law_1.2,0.031217920780181884
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,2,32,power_law_1.2,0.041149439811706545
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,2,32,power_law_1.2,0.045698561668396
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,2,32,power_law_1.01,0.16671104431152345
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,2,32,power_law_1.01,0.2373593521118164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,2,32,power_law_1.2,0.03462271928787232
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,2,32,power_law_1.2,0.053212161064147945
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,2,32,power_law_1.2,0.08353280067443848
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,2,32,power_law_1.2,0.12549504280090332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,2,32,power_law_1.2,0.09700991630554198
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,4,16,balanced,0.030938880443572996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,4,16,balanced,0.030512640476226805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,4,16,balanced,0.03188096046447754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,4,16,balanced,0.06681983947753907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,4,16,balanced,0.08497023582458496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,4,16,balanced,0.12956543922424318
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,4,16,balanced,0.1331289577484131
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,4,16,balanced,0.13392383575439454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,4,16,balanced,0.13706624031066894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,4,16,balanced,0.1414579200744629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,4,16,balanced,0.14665984153747558
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,4,16,balanced,0.1393446445465088
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,4,16,balanced,0.1443712043762207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,4,16,balanced,0.14748031616210938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,4,16,balanced,0.15547648429870606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,4,16,balanced,0.053034238815307624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,4,16,balanced,0.1668339157104492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,4,16,balanced,0.18073984146118166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,4,16,balanced,0.133373441696167
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,4,16,balanced,0.20861440658569336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,4,16,balanced,0.23528064727783202
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,4,16,balanced,0.29648767471313475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,4,16,balanced,0.35137535095214845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,4,16,balanced,0.4707379150390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,4,16,balanced,0.032517120838165284
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,4,16,balanced,0.6280934524536133
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,2,32,power_law_1.2,0.06531455993652344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,4,16,balanced,0.03237119913101196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,4,16,balanced,0.033799679279327394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,4,16,balanced,0.895946273803711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,4,16,balanced,0.03625344038009644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,4,16,balanced,0.043786239624023435
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,4,16,balanced,0.06425983905792236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,4,16,balanced,0.07110784053802491
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,4,16,balanced,1.1608819580078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,4,16,balanced,0.07320576190948487
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,4,16,balanced,0.0731712007522583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,4,16,balanced,0.0767526388168335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,4,16,balanced,0.07786240100860595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,4,16,balanced,0.08185088157653808
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,4,16,balanced,0.07443712234497071
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,4,16,balanced,0.07793791770935059
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,4,16,balanced,0.08087936401367188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,4,16,balanced,0.08551168441772461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,4,16,balanced,0.09027584075927735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,4,16,balanced,0.10001664161682129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,4,16,balanced,0.11576191902160644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,4,16,balanced,0.1347430419921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,4,16,balanced,0.16986751556396484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,4,16,balanced,0.20153087615966797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,4,16,balanced,0.06982016086578369
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,4,16,balanced,0.28732416152954104
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,4,16,balanced,0.35575424194335936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,4,16,balanced,0.5127718353271484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,4,16,balanced,0.6608934020996093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,2,32,power_law_1.2,0.19942527770996094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,2,32,power_law_1.2,0.25587711334228513
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,2,32,power_law_1.2,0.3438681411743164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,4,16,balanced,0.01988479971885681
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,4,16,balanced,0.02025343894958496
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,4,16,balanced,0.03720704078674316
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,4,16,balanced,0.021067519187927247
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,4,16,balanced,0.03752448081970215
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,4,16,balanced,0.03715199947357178
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,4,16,balanced,0.03769471883773804
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,4,16,balanced,0.038200318813323975
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,4,16,balanced,0.037676799297332766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,4,16,balanced,0.037532160282135005
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,4,16,balanced,0.037904639244079594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,4,16,balanced,0.03782399892807007
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,4,16,balanced,0.03801599979400635
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,4,16,balanced,0.02532480001449585
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,4,16,balanced,0.03976703882217407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,4,16,balanced,0.04057983875274658
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,4,16,balanced,0.04268159866333008
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,4,16,balanced,0.044441599845886234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,4,16,balanced,0.0478656005859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,4,16,balanced,0.022693119049072265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,4,16,power_law_1.01,0.06443007946014404
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,4,16,power_law_1.01,0.08228863716125488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,4,16,power_law_1.01,0.0824563217163086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,4,16,power_law_1.01,0.08447999954223633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,4,16,power_law_1.01,0.08531071662902831
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,4,16,power_law_1.01,0.1263539218902588
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,4,16,power_law_1.01,0.12923647880554198
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,4,16,power_law_1.01,0.12885503768920897
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,4,16,power_law_1.01,0.13241727828979494
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,4,16,power_law_1.01,0.13337471961975098
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,4,16,power_law_1.01,0.13811200141906738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,4,16,power_law_1.01,0.14109184265136718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,4,16,power_law_1.01,0.14487808227539062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,4,16,power_law_1.01,0.14249088287353515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,4,16,power_law_1.01,0.14324864387512207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,4,16,power_law_1.01,0.14863231658935547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,4,16,power_law_1.01,0.15844351768493653
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,4,16,power_law_1.01,0.18143999099731445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,4,16,power_law_1.01,0.20570495605468747
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,4,16,power_law_1.01,0.2532083129882813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,4,16,power_law_1.01,0.2935577583312988
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,4,16,power_law_1.01,0.3954111862182617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,4,16,power_law_1.01,0.5094348907470703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,4,16,power_law_1.01,0.7135692596435547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,4,16,power_law_1.01,0.876953582763672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,4,16,balanced,0.0372326397895813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,4,16,balanced,0.03737087965011597
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,4,16,power_law_1.01,0.03877887964248657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,4,16,power_law_1.01,1.3015180969238282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,4,16,power_law_1.01,0.05975168228149415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,4,16,power_law_1.01,0.06084479808807373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,4,16,power_law_1.01,0.062048001289367674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,4,16,power_law_1.01,1.7786087036132812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,4,16,power_law_1.01,0.06101119995117188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,4,16,power_law_1.01,0.06296063899993896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,4,16,power_law_1.01,0.07053567886352538
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,4,16,power_law_1.01,0.06752511978149414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,4,16,power_law_1.01,0.06681344032287598
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,4,16,power_law_1.01,0.07226751804351807
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,4,16,power_law_1.01,0.07326464176177978
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,4,16,power_law_1.01,0.0773644781112671
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,4,16,power_law_1.01,0.07891583919525147
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,4,16,power_law_1.01,0.07484032154083252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,4,16,power_law_1.01,0.07789311885833741
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,4,16,power_law_1.01,0.0786956787109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,4,16,power_law_1.01,0.08472319602966308
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,4,16,power_law_1.01,0.09588352203369141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,4,16,power_law_1.01,0.10863103866577148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,4,16,power_law_1.01,0.13709823608398436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,4,16,power_law_1.01,0.1675059127807617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,4,16,power_law_1.01,0.2237273597717285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,4,16,power_law_1.01,0.2721241569519043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,4,16,power_law_1.01,0.39532161712646485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,4,16,power_law_1.01,0.5055667114257812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,4,16,power_law_1.01,0.7528896331787109
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,4,16,power_law_1.01,0.0240447998046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,4,16,power_law_1.01,0.037500159740448
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,4,16,power_law_1.01,0.03688191890716553
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,4,16,power_law_1.01,0.9229337310791015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,4,16,power_law_1.01,0.03606784105300903
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,4,16,power_law_1.01,0.033207039833068844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,4,16,power_law_1.01,0.02918528079986572
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,4,16,power_law_1.01,0.031028480529785157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,4,16,power_law_1.01,0.03165568113327026
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,4,16,power_law_1.01,0.032647678852081294
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,4,16,power_law_1.01,0.03247999906539917
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,4,16,power_law_1.01,0.03370239973068238
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,4,16,power_law_1.01,0.03435264110565185
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,4,16,power_law_1.01,0.03593728065490723
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,4,16,power_law_1.01,0.035258879661560064
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,4,16,power_law_1.01,0.033859839439392095
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,4,16,power_law_1.01,0.036161279678344725
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,4,16,power_law_1.01,0.03868799924850464
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,4,16,balanced,0.03771647930145264
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,4,16,balanced,0.06933119773864746
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,4,16,balanced,0.06433792114257812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,4,16,balanced,0.11245823860168458
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,4,16,balanced,0.09263615608215332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,4,16,power_law_1.01,0.057736959457397465
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,4,16,power_law_1.01,0.04338687896728516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,4,16,power_law_1.01,0.050022401809692384
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,4,16,power_law_1.2,0.06436480045318603
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,4,16,power_law_1.2,0.0822156810760498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,4,16,power_law_1.2,0.07890687942504883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,4,16,power_law_1.2,0.08366335868835449
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,4,16,power_law_1.01,0.06638591766357423
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,4,16,power_law_1.2,0.08569343566894531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,4,16,power_law_1.2,0.12067968368530273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,4,16,power_law_1.2,0.12874367713928223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,4,16,power_law_1.2,0.12781824111938475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,4,16,power_law_1.2,0.13065728187561035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,4,16,power_law_1.2,0.13441791534423828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,4,16,power_law_1.2,0.13667200088500978
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,4,16,power_law_1.2,0.14075008392333985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,4,16,power_law_1.2,0.14372096061706544
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,4,16,power_law_1.2,0.14444416046142577
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,4,16,power_law_1.2,0.14754176139831543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,4,16,power_law_1.2,0.1553228759765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,4,16,power_law_1.2,0.1625663948059082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,4,16,power_law_1.2,0.18755071640014648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,4,16,power_law_1.2,0.21690752029418947
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,4,16,power_law_1.2,0.25841407775878905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,4,16,power_law_1.2,0.3327743911743164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,4,16,power_law_1.2,0.4504601669311524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,4,16,power_law_1.2,0.5821721649169922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,4,16,power_law_1.2,0.8478336334228516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,4,16,power_law_1.2,1.1196364593505859
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,4,16,power_law_1.01,0.0765772819519043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,4,16,power_law_1.01,0.09232895851135255
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,4,16,power_law_1.2,0.03805952072143555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,4,16,power_law_1.01,0.1524569606781006
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,4,16,power_law_1.01,0.11945088386535645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,4,16,power_law_1.2,0.05834623813629151
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,4,16,power_law_1.2,0.056030721664428705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,4,16,power_law_1.2,0.058661122322082516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,4,16,power_law_1.2,0.06130047798156738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,4,16,power_law_1.2,0.06116223812103272
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,4,16,power_law_1.2,1.6076722717285157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,4,16,power_law_1.2,0.06258687973022461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,4,16,power_law_1.2,0.06400256156921387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,4,16,power_law_1.2,0.06688767910003662
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,4,16,power_law_1.2,0.06958208084106446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,4,16,power_law_1.2,0.07482751846313476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,4,16,power_law_1.2,0.0786841583251953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,4,16,power_law_1.2,0.07974527835845947
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,4,16,power_law_1.2,0.07844480037689208
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,4,16,power_law_1.2,0.08226431846618652
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,4,16,power_law_1.2,0.10320768356323243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,4,16,power_law_1.2,0.15143296241760254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,4,16,power_law_1.2,2.1797080993652345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,4,16,power_law_1.2,0.20006912231445312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,4,16,power_law_1.2,0.28335487365722656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,4,16,power_law_1.2,0.32009086608886717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,4,16,power_law_1.2,0.07513599872589112
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,4,16,power_law_1.2,0.08854144096374512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,4,16,power_law_1.2,0.44673919677734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,4,16,power_law_1.2,0.11659903526306152
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,4,16,power_law_1.2,0.02479487895965576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,4,16,power_law_1.2,0.5536844635009766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,4,16,power_law_1.2,0.03628416061401367
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,4,16,power_law_1.2,0.03250175952911377
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,4,16,power_law_1.2,0.035870718955993655
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,4,16,power_law_1.2,0.03681407928466797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,4,16,power_law_1.2,0.0278438401222229
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,4,16,power_law_1.2,0.03127808094024658
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,4,16,power_law_1.2,0.03137792110443115
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,4,16,power_law_1.2,0.032700159549713136
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,4,16,power_law_1.2,0.03413248062133789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,4,16,power_law_1.2,0.8441766357421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,4,16,power_law_1.2,0.03399296045303345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,4,16,power_law_1.2,1.046288604736328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,4,16,power_law_1.2,0.03540992021560669
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,4,16,power_law_1.2,0.03430399894714355
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,4,16,power_law_1.2,0.03669120073318481
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,4,16,power_law_1.2,0.038895359039306646
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,4,16,power_law_1.2,0.04609151840209961
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,4,16,power_law_1.2,0.03321727991104126
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,4,16,power_law_1.01,0.3273971176147461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,4,16,power_law_1.01,0.23156991958618164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,4,16,power_law_1.2,0.04183040142059326
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,4,16,power_law_1.2,0.053503999710083
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,4,16,power_law_1.2,0.06987264156341552
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,8,8,balanced,0.030508799552917482
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,8,8,balanced,0.030881280899047854
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,4,16,power_law_1.2,0.0770905590057373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,8,8,balanced,0.033841919898986814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,8,8,balanced,0.05823232173919678
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,8,8,balanced,0.07024256229400636
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,8,8,balanced,0.09833727836608887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,8,8,balanced,0.13762304306030274
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,4,16,power_law_1.2,0.10011648178100585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,8,8,balanced,0.14247936248779297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,8,8,balanced,0.14568832397460937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,8,8,balanced,0.15011072158813477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,8,8,balanced,0.14140288352966307
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,8,8,balanced,0.14628992080688477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,8,8,balanced,0.14987008094787596
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,8,8,balanced,0.1560537624359131
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,4,16,power_law_1.2,0.13026944160461426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,8,8,balanced,0.15990912437438964
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,8,8,balanced,0.16644607543945314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,8,8,balanced,0.17776512145996093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,8,8,balanced,0.2032908821105957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,8,8,balanced,0.23013120651245117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,8,8,balanced,0.28179967880249024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,8,8,balanced,0.33485313415527346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,8,8,balanced,0.4418342590332031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,8,8,balanced,0.5456051254272462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,8,8,balanced,0.7613581085205079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,8,8,balanced,0.03253247976303101
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,8,8,balanced,1.0111309051513673
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,8,8,balanced,0.03414783954620361
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,8,8,balanced,0.04151167869567871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,8,8,balanced,0.04436992168426514
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,8,8,balanced,0.055101442337036136
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,8,8,balanced,0.08091775894165039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,8,8,balanced,0.08726528167724609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,8,8,balanced,1.4782246398925782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,8,8,balanced,0.08900223731994629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,8,8,balanced,0.09365887641906738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,8,8,balanced,0.09633664131164552
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,8,8,balanced,0.08602879524230958
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,8,8,balanced,0.09039360046386719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,8,8,balanced,0.09256064414978027
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,8,8,balanced,0.09694592475891113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,8,8,balanced,1.937829132080078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,8,8,balanced,0.097324800491333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,8,8,balanced,0.1014668846130371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,8,8,balanced,0.10871935844421388
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,8,8,balanced,0.12486144065856934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,8,8,balanced,0.14212991714477538
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,8,8,balanced,0.1750156784057617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,8,8,balanced,0.20623104095458986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,8,8,balanced,0.27350784301757813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,8,8,balanced,0.33603328704833985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,8,8,balanced,0.4833536148071289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,8,8,balanced,0.6071385574340821
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,8,8,balanced,0.8943615722656251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,8,8,balanced,1.162228469848633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,4,16,power_law_1.2,0.19671680450439452
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,4,16,power_law_1.2,0.23204736709594725
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,4,16,power_law_1.2,0.3455718231201172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,4,16,power_law_1.2,0.49559680938720707
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,8,8,balanced,0.02311295986175537
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,8,8,balanced,0.023311359882354738
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,8,8,balanced,0.04596223831176758
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,8,8,balanced,0.046671361923217775
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,8,8,balanced,0.04676352024078369
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,8,8,balanced,0.02434943914413452
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,8,8,balanced,0.04663167953491211
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,8,8,balanced,0.046869759559631345
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,8,8,balanced,0.047006721496582034
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,8,8,balanced,0.04715904235839844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,8,8,balanced,0.04735487937927246
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,8,8,balanced,0.04850048065185547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,8,8,balanced,0.05154687881469726
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,8,8,balanced,0.025612800121307372
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,8,8,balanced,0.05425920009613037
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,8,8,balanced,0.05663487911224365
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,8,8,balanced,0.06223616123199462
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,8,8,balanced,0.06669951915740967
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,8,8,balanced,0.04977280139923096
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,8,8,balanced,0.031013119220733642
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,8,8,power_law_1.01,0.03581568002700806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,8,8,power_law_1.01,0.0671295976638794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,8,8,power_law_1.01,0.07318272113800048
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,8,8,power_law_1.01,0.07667712211608887
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,8,8,balanced,0.046151680946350096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,8,8,power_law_1.01,0.0845580768585205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,8,8,power_law_1.01,0.08914560317993164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,8,8,power_law_1.01,0.13246463775634765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,8,8,power_law_1.01,0.13612544059753418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,8,8,power_law_1.01,0.13992064476013183
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,8,8,power_law_1.01,0.14229503631591797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,8,8,power_law_1.01,0.14727807998657227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,8,8,power_law_1.01,0.1461235237121582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,8,8,power_law_1.01,0.14237695693969726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,8,8,power_law_1.01,0.14834303855895997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,8,8,power_law_1.01,0.15259519577026368
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,8,8,power_law_1.01,0.15833984375000001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,8,8,power_law_1.01,0.16964223861694336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,8,8,power_law_1.01,0.1936319923400879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,8,8,power_law_1.01,0.2220607948303223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,8,8,power_law_1.01,0.2628915214538574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,8,8,power_law_1.01,0.31007232666015627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,8,8,power_law_1.01,0.39651199340820314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,8,8,power_law_1.01,0.5075942230224609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,8,8,power_law_1.01,0.7323136138916015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,8,8,power_law_1.01,0.9946726226806641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,8,8,power_law_1.01,1.4161331176757812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,8,8,power_law_1.01,0.037785599231719966
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,8,8,power_law_1.01,0.04437632083892822
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,8,8,power_law_1.01,0.05678336143493652
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,8,8,power_law_1.01,1.9632588195800782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,8,8,power_law_1.01,0.059082241058349604
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,8,8,power_law_1.01,0.0735206413269043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,8,8,power_law_1.01,0.07891456127166749
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,8,8,power_law_1.01,0.07966976165771485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,8,8,power_law_1.01,0.08164095878601074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,8,8,power_law_1.01,0.08644351959228516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,8,8,power_law_1.01,0.0905356788635254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,8,8,power_law_1.01,0.06406015872955322
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,8,8,balanced,0.046209278106689455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,8,8,power_law_1.01,0.0943667221069336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,8,8,power_law_1.01,0.0853388786315918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,8,8,power_law_1.01,0.08895999908447265
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,8,8,balanced,0.046220798492431645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,8,8,power_law_1.01,0.09235199928283691
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,8,8,power_law_1.01,0.09478655815124512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,8,8,power_law_1.01,0.10285823822021485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,8,8,power_law_1.01,0.11135616302490234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,8,8,power_law_1.01,0.1299404811859131
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,8,8,power_law_1.01,0.16226688385009766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,8,8,power_law_1.01,0.19723007202148438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,8,8,power_law_1.01,0.25746175765991214
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,8,8,power_law_1.01,0.32736000061035153
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,8,8,power_law_1.01,0.43204097747802733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,8,8,power_law_1.01,0.5418393707275391
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,8,8,power_law_1.01,0.02558079957962036
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,8,8,power_law_1.01,0.030488319396972656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,8,8,power_law_1.01,0.03476351976394653
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,8,8,power_law_1.01,0.8335167694091797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,8,8,power_law_1.01,0.03802367925643921
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,8,8,power_law_1.01,0.04131711959838867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,8,8,power_law_1.01,0.03471872091293335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,8,8,power_law_1.01,1.1340518188476563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,8,8,power_law_1.01,0.0444927978515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,8,8,power_law_1.01,0.04176127910614014
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,8,8,power_law_1.01,0.04228608131408691
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,8,8,power_law_1.01,0.043680000305175784
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,8,8,power_law_1.01,0.043805441856384277
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,8,8,power_law_1.01,0.044430079460144045
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,8,8,power_law_1.01,0.04645887851715088
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,8,8,power_law_1.01,0.04532351970672607
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,8,8,power_law_1.01,0.04639103889465332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,8,8,power_law_1.01,0.04947199821472168
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,8,8,power_law_1.01,0.04227839946746827
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,8,8,balanced,0.09469440460205078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,8,8,balanced,0.1467123222351074
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,8,8,balanced,0.18536832809448242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,8,8,balanced,0.1052569580078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,8,8,power_law_1.01,0.05753983974456787
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,8,8,power_law_1.01,0.062991361618042
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,8,8,power_law_1.01,0.07843071937561034
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,8,8,power_law_1.2,0.04032256126403809
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,8,8,power_law_1.01,0.0939737606048584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,8,8,power_law_1.2,0.06652800083160401
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,8,8,power_law_1.2,0.06796800136566163
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,8,8,power_law_1.2,0.07416704177856445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,8,8,power_law_1.2,0.08781311988830567
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,8,8,power_law_1.2,0.13265024185180666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,8,8,power_law_1.2,0.13400959968566895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,8,8,power_law_1.2,0.1378636837005615
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,8,8,power_law_1.2,0.07878143787384033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,8,8,power_law_1.2,0.1434444808959961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,8,8,power_law_1.2,0.1471782398223877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,8,8,power_law_1.2,0.14582015991210937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,8,8,power_law_1.2,0.1409011173248291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,8,8,power_law_1.2,0.1453427219390869
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,8,8,power_law_1.01,0.11677824020385744
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,8,8,power_law_1.01,0.1483737564086914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,8,8,power_law_1.2,0.1547929573059082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,8,8,power_law_1.2,0.1600281524658203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,8,8,power_law_1.2,0.16946943283081056
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,8,8,power_law_1.2,0.2271718406677246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,8,8,power_law_1.2,0.1991116714477539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,8,8,power_law_1.2,0.2762918472290039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,8,8,power_law_1.2,0.3218739318847656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,8,8,power_law_1.01,0.18143999099731445
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,8,8,power_law_1.01,0.24769407272338867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,8,8,power_law_1.2,0.6000012969970703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,8,8,power_law_1.2,0.8056076812744142
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,8,8,power_law_1.2,1.0055500793457033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,8,8,power_law_1.2,0.03814656019210815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,8,8,power_law_1.2,0.04987904071807862
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,8,8,power_law_1.2,0.46479358673095705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,8,8,power_law_1.2,0.05712639808654785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,8,8,power_law_1.2,1.5356057739257811
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,8,8,power_law_1.2,0.062350077629089354
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,8,8,power_law_1.2,0.08005632400512695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,8,8,power_law_1.2,0.0809113597869873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,8,8,power_law_1.2,0.08113151550292969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,8,8,power_law_1.2,0.08561535835266113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,8,8,power_law_1.2,0.08970111846923828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,8,8,power_law_1.2,0.04356863975524902
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,8,8,power_law_1.2,0.0895206356048584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,8,8,power_law_1.2,2.1930880737304688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,8,8,power_law_1.2,0.08827263832092284
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,8,8,power_law_1.2,0.08921728134155274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,8,8,power_law_1.2,0.09355775833129883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,8,8,power_law_1.2,0.09658495903015137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,8,8,power_law_1.2,0.10053119659423829
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,8,8,power_law_1.2,0.06941440105438232
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,8,8,power_law_1.2,0.13650688171386718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,8,8,power_law_1.2,0.16509824752807617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,8,8,power_law_1.2,0.19231103897094726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,8,8,power_law_1.2,0.2797145652770996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,8,8,power_law_1.2,0.32260353088378907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,8,8,power_law_1.2,0.47074432373046876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,8,8,power_law_1.2,0.02551424026489258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,8,8,power_law_1.2,0.6423257446289062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,8,8,power_law_1.2,0.03074944019317627
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,8,8,power_law_1.2,0.031425280570983885
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,8,8,power_law_1.2,0.1186457633972168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,8,8,power_law_1.2,0.8213196563720704
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,8,8,power_law_1.2,0.03486079931259155
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,8,8,power_law_1.2,0.0439027214050293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,8,8,power_law_1.2,0.03904639959335327
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,8,8,power_law_1.2,0.04197120189666748
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,8,8,power_law_1.2,0.041242880821228026
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,8,8,power_law_1.2,0.04038015842437744
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,8,8,power_law_1.2,0.04225024223327637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,8,8,power_law_1.2,1.223114242553711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,8,8,power_law_1.2,0.043022079467773436
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,8,8,power_law_1.2,0.04382207870483398
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,8,8,power_law_1.2,0.04644480228424072
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,8,8,power_law_1.2,0.04807680130004883
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,8,8,power_law_1.2,0.05534207820892334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,8,8,power_law_1.2,0.05219456195831299
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,8,8,power_law_1.2,0.0718387222290039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,8,8,power_law_1.2,0.04011392116546631
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,8,8,power_law_1.01,0.33782913208007814
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,8,8,power_law_1.01,0.4851545715332032
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,8,8,power_law_1.2,0.06613503932952881
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,8,8,power_law_1.2,0.09102463722229004
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,8,8,power_law_1.2,0.1503872013092041
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,8,8,power_law_1.2,0.10643199920654296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,16,4,balanced,0.02934655904769897
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,16,4,balanced,0.030933759212493896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,16,4,balanced,0.03525759935379028
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,8,8,power_law_1.2,0.1885568046569824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,16,4,balanced,0.06027904033660889
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,16,4,balanced,0.07169280052185059
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,16,4,balanced,0.09324543952941895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,16,4,balanced,0.09440256118774414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,16,4,balanced,0.09545727729797363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,16,4,balanced,0.15117823600769043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,16,4,balanced,0.15353983879089356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,16,4,balanced,0.1574143981933594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,16,4,balanced,0.15997440338134766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,16,4,balanced,0.1657151985168457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,16,4,balanced,0.1694553565979004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,16,4,balanced,0.17321983337402344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,16,4,balanced,0.17876224517822265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,16,4,balanced,0.19351295471191404
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,16,4,balanced,0.21689855575561526
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,16,4,balanced,0.24147840499877932
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,16,4,balanced,0.2918963241577148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,16,4,balanced,0.34426494598388674
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,8,8,power_law_1.2,0.2712639999389649
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,16,4,balanced,0.45155582427978513
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,16,4,balanced,0.5506905746459961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,16,4,balanced,0.03461888074874878
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,16,4,balanced,0.7589516448974609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,16,4,balanced,0.03542272090911865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,16,4,balanced,0.04159872055053711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,16,4,balanced,0.05335807800292969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,16,4,balanced,1.0116864013671876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,16,4,balanced,0.06888448238372803
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,16,4,balanced,0.09338111877441406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,16,4,balanced,0.09209728240966797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,16,4,balanced,0.09450240135192871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,16,4,balanced,0.1051865577697754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,16,4,balanced,0.10660736083984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,16,4,balanced,0.10888575553894042
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,16,4,balanced,1.467157745361328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,16,4,balanced,0.11204095840454102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,16,4,balanced,0.1146675205230713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,16,4,balanced,0.11946880340576171
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,16,4,balanced,0.1193395233154297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,16,4,balanced,0.12261631965637207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,16,4,balanced,0.13092096328735353
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,16,4,balanced,0.14631296157836915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,16,4,balanced,1.9255859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,16,4,balanced,0.1615999984741211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,16,4,balanced,0.19250560760498048
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,16,4,balanced,0.22384639739990236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,16,4,balanced,0.28581119537353517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,16,4,balanced,0.34646400451660153
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,16,4,balanced,0.5027008056640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,16,4,balanced,0.6240832138061523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,16,4,balanced,0.9208409881591797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,16,4,balanced,1.201624298095703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,8,8,power_law_1.2,0.38190975189208987
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,8,8,power_law_1.2,0.48787841796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,8,8,power_law_1.2,0.7889049530029297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,16,4,balanced,0.03058687925338745
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,16,4,balanced,0.0309171199798584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,16,4,balanced,0.06470911979675294
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,16,4,balanced,0.06507008075714112
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,16,4,balanced,0.06526976108551026
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,16,4,balanced,0.06514431953430175
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,16,4,balanced,0.03256831884384155
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,16,4,balanced,0.06565120220184326
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,16,4,balanced,0.066494722366333
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,16,4,balanced,0.0675878381729126
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,16,4,balanced,0.06764287948608398
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,16,4,balanced,0.06947455883026124
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,16,4,balanced,0.07209216117858887
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,16,4,balanced,0.07443583965301513
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,16,4,balanced,0.07975552082061768
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,16,4,balanced,0.036669440269470215
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,16,4,balanced,0.09419903755187989
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,16,4,balanced,0.10355199813842772
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,16,4,balanced,0.046383361816406246
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,16,4,balanced,0.0843609619140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,16,4,power_law_1.01,0.03466495990753174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,16,4,power_law_1.01,0.042571520805358885
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,16,4,power_law_1.01,0.06250751972198486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,16,4,power_law_1.01,0.068504319190979
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,16,4,power_law_1.01,0.07668479919433593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,16,4,power_law_1.01,0.08008576393127441
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,16,4,power_law_1.01,0.08831616401672363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,16,4,power_law_1.01,0.09102592468261719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,16,4,power_law_1.01,0.1489472007751465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,16,4,power_law_1.01,0.14845312118530274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,16,4,power_law_1.01,0.151342077255249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,16,4,power_law_1.01,0.15712127685546876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,16,4,power_law_1.01,0.15975423812866213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,16,4,power_law_1.01,0.1678528022766113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,16,4,power_law_1.01,0.17084800720214843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,16,4,power_law_1.01,0.17725696563720703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,16,4,power_law_1.01,0.19124095916748046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,16,4,power_law_1.01,0.21719936370849607
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,16,4,power_law_1.01,0.2443328094482422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,16,4,power_law_1.01,0.2941695976257324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,16,4,power_law_1.01,0.35015937805175784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,16,4,power_law_1.01,0.446220817565918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,16,4,balanced,0.06458623886108399
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,16,4,power_law_1.01,0.5665356826782226
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,16,4,balanced,0.06441088199615479
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,16,4,power_law_1.01,0.80195068359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,16,4,power_law_1.01,1.0540377807617187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,16,4,power_law_1.01,0.035802879333496095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,16,4,power_law_1.01,0.041175041198730465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,16,4,power_law_1.01,1.5357965087890624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,16,4,power_law_1.01,0.050399999618530276
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,16,4,power_law_1.01,0.06402048110961914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,16,4,power_law_1.01,0.06825727939605712
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,16,4,power_law_1.01,0.0711359977722168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,16,4,power_law_1.01,2.007283172607422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,16,4,power_law_1.01,0.08332415580749511
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,16,4,power_law_1.01,0.08266112327575684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,16,4,power_law_1.01,0.09966336250305176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,16,4,power_law_1.01,0.09255935668945312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,16,4,power_law_1.01,0.09540608406066894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,16,4,power_law_1.01,0.1017369556427002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,16,4,power_law_1.01,0.10470911979675293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,16,4,power_law_1.01,0.1152409553527832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,16,4,power_law_1.01,0.11288576126098633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,16,4,power_law_1.01,0.11740287780761718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,16,4,power_law_1.01,0.12347135543823243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,16,4,power_law_1.01,0.1415385627746582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,16,4,power_law_1.01,0.15861375808715822
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,16,4,power_law_1.01,0.19521791458129883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,16,4,power_law_1.01,0.23041664123535158
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,16,4,power_law_1.01,0.29996160507202146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,16,4,power_law_1.01,0.36722175598144535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,16,4,power_law_1.01,0.5238169479370117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,16,4,power_law_1.01,0.6698162841796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,16,4,power_law_1.01,0.033011200428009035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,16,4,power_law_1.01,0.9434329223632811
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,16,4,power_law_1.01,0.03581952095031739
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,16,4,power_law_1.01,0.04113152027130127
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,16,4,power_law_1.01,0.04549248218536377
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,16,4,power_law_1.01,0.048162560462951656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,16,4,power_law_1.01,0.050647039413452145
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,16,4,power_law_1.01,0.05857535839080811
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,16,4,power_law_1.01,0.059390721321105955
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,16,4,power_law_1.01,0.0606220817565918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,16,4,power_law_1.01,0.061054720878601074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,16,4,power_law_1.01,1.2531609344482422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,16,4,power_law_1.01,0.06318336009979249
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,16,4,power_law_1.01,0.06229119777679444
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,16,4,power_law_1.01,0.06457344055175782
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,16,4,power_law_1.01,0.06812160015106201
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,16,4,power_law_1.01,0.06867072105407715
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,16,4,power_law_1.01,0.07170815944671631
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,16,4,power_law_1.01,0.07564671993255614
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,16,4,balanced,0.06486271858215333
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,16,4,balanced,0.17795072555541994
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,16,4,balanced,0.25004287719726564
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,16,4,balanced,0.15648639678955079
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,16,4,balanced,0.32038528442382813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,16,4,power_law_1.01,0.09037823677062988
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,16,4,power_law_1.01,0.09857791900634766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,16,4,power_law_1.01,0.12323328018188477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,16,4,power_law_1.2,0.035088639259338375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,16,4,power_law_1.01,0.14140543937683106
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,16,4,power_law_1.2,0.04454016208648682
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,16,4,power_law_1.2,0.05793280124664306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,16,4,power_law_1.2,0.06736639976501466
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,16,4,power_law_1.2,0.07736832141876221
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,16,4,power_law_1.2,0.07578368186950683
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,16,4,power_law_1.2,0.08902144432067871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,16,4,power_law_1.2,0.09069567680358886
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,16,4,power_law_1.2,0.14570879936218262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,16,4,power_law_1.2,0.14821760177612303
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,16,4,power_law_1.2,0.15213184356689452
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,16,4,power_law_1.2,0.15704192161560057
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,16,4,power_law_1.2,0.16040191650390626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,16,4,power_law_1.2,0.16737024307250975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,16,4,power_law_1.2,0.17147008895874022
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,16,4,power_law_1.2,0.1779852867126465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,16,4,power_law_1.2,0.1934694480895996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,16,4,power_law_1.2,0.21993215560913085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,16,4,power_law_1.2,0.24596736907958983
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,16,4,power_law_1.2,0.29821439743041994
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,16,4,power_law_1.2,0.34617473602294924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,16,4,power_law_1.2,0.4682406234741211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,16,4,power_law_1.2,0.5921228790283204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,16,4,power_law_1.2,0.8577279663085937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,16,4,power_law_1.01,0.2127462387084961
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,16,4,power_law_1.01,0.2424844741821289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,16,4,power_law_1.2,1.0991487884521485
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,16,4,power_law_1.01,0.4075660705566406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,16,4,power_law_1.2,0.03671551942825317
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,16,4,power_law_1.2,0.04159359931945801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,16,4,power_law_1.2,0.04904319763183594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,16,4,power_law_1.2,0.06329472064971924
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,16,4,power_law_1.01,0.48376319885253904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,16,4,power_law_1.2,0.06936575889587401
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,16,4,power_law_1.2,0.08291711807250976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,16,4,power_law_1.2,0.0697318410873413
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,16,4,power_law_1.2,0.09248640060424804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,16,4,power_law_1.2,0.0816102409362793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,16,4,power_law_1.2,1.5734938049316407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,16,4,power_law_1.2,0.09238783836364746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,16,4,power_law_1.2,0.09683967590332031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,16,4,power_law_1.2,0.10082816123962401
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,16,4,power_law_1.2,0.10631808280944824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,16,4,power_law_1.2,0.11464447975158691
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,16,4,power_law_1.2,0.1150284767150879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,16,4,power_law_1.2,0.11860992431640624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,16,4,power_law_1.2,0.1278822422027588
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,16,4,power_law_1.2,0.14698880195617675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,16,4,power_law_1.2,0.16552192687988282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,16,4,power_law_1.2,0.20116607666015626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,16,4,power_law_1.2,0.2346726417541504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,16,4,power_law_1.2,0.3078630447387695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,16,4,power_law_1.2,0.3845043182373047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,16,4,power_law_1.2,0.5493542480468749
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,16,4,power_law_1.2,0.03297024011611939
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,16,4,power_law_1.2,0.6963085174560547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,16,4,power_law_1.2,0.03601023912429809
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,16,4,power_law_1.2,0.038270719051361084
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,16,4,power_law_1.2,0.04269696235656738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,16,4,power_law_1.2,0.9725234985351563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,16,4,power_law_1.2,2.012497863769531
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,16,4,power_law_1.2,0.047295999526977536
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,16,4,power_law_1.2,0.051457281112670894
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,16,4,power_law_1.2,0.05818880081176758
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,16,4,power_law_1.2,0.05702271938323975
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,16,4,power_law_1.2,0.05730688095092773
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,16,4,power_law_1.2,0.06163328170776368
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,16,4,power_law_1.2,0.06236671924591064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,16,4,power_law_1.2,1.2857830810546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,16,4,power_law_1.2,0.06507520198822021
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,16,4,power_law_1.2,0.06655744075775147
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,16,4,power_law_1.2,0.06900991916656493
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,16,4,power_law_1.2,0.0767193603515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,16,4,power_law_1.2,0.10641023635864258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,16,4,power_law_1.2,0.07203455924987792
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,16,4,power_law_1.01,0.6716646575927735
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,16,4,power_law_1.01,1.043214111328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,16,4,power_law_1.2,0.08242176055908204
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,16,4,power_law_1.2,0.11234560012817382
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,16,4,power_law_1.2,0.14991871833801268
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,16,4,power_law_1.2,0.25124095916748046
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,16,4,power_law_1.2,0.19793535232543946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,32,2,balanced,0.03087104082107544
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,32,2,balanced,0.032711679935455325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,32,2,balanced,0.035676159858703614
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,32,2,balanced,0.0617139196395874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,32,2,balanced,0.08617088317871094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,32,2,balanced,0.10969344139099121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,32,2,balanced,0.11060992240905762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,32,2,balanced,0.1113088035583496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,32,2,balanced,0.11310208320617676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,32,2,balanced,0.11344511985778809
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,16,4,power_law_1.2,0.3213286590576172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,32,2,balanced,0.11699071884155274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,32,2,balanced,0.1967308807373047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,32,2,balanced,0.19969280242919923
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,32,2,balanced,0.206744327545166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,32,2,balanced,0.20923776626586915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,32,2,balanced,0.21568511962890624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,32,2,balanced,0.2287500762939453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,32,2,balanced,0.25144704818725583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,32,2,balanced,0.2760895919799805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,32,2,balanced,0.32583423614501955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,32,2,balanced,0.3763071823120117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,32,2,balanced,0.48414337158203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,32,2,balanced,0.5851110458374024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,32,2,balanced,0.7910054779052734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,32,2,balanced,0.03299839973449707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,32,2,balanced,1.0479551696777345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,32,2,balanced,0.03659647941589356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,32,2,balanced,0.04413568019866944
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,32,2,balanced,0.06117631912231445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,32,2,balanced,0.09232128143310547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,32,2,balanced,0.14001791954040527
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,32,2,balanced,1.4970445251464843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,32,2,balanced,0.13332863807678225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,32,2,balanced,0.13477888107299804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,32,2,balanced,0.13623423576354982
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,32,2,balanced,0.13725312232971193
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,32,2,balanced,0.13954175949096678
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,32,2,balanced,0.16349439620971679
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,32,2,balanced,1.951429138183594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,32,2,balanced,0.1659494400024414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,32,2,balanced,0.1694528007507324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,32,2,balanced,0.17267200469970703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,32,2,balanced,0.18084224700927734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,32,2,balanced,0.19355775833129882
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,32,2,balanced,0.2094041633605957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,32,2,balanced,0.24014335632324219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,32,2,balanced,0.2684505653381348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,32,2,balanced,0.32834686279296876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,32,2,balanced,0.38653568267822264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,32,2,balanced,0.17010944366455077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,32,2,balanced,0.6015180969238282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,32,2,balanced,0.7165977478027343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,32,2,balanced,1.0487769317626952
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,32,2,balanced,1.3862399291992187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,16,4,power_law_1.2,0.4886796951293945
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,16,4,power_law_1.2,0.6592896270751953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,16,4,power_law_1.2,0.9900377655029298
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,16,4,power_law_1.2,1.1361203002929687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,32,2,balanced,0.045432319641113286
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,32,2,balanced,0.10052864074707031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,32,2,balanced,0.044465918540954594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,32,2,balanced,0.10071040153503419
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,32,2,balanced,0.10166015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,32,2,balanced,0.10306559562683107
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,32,2,balanced,0.048936958312988284
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,32,2,balanced,0.056235518455505375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,32,2,balanced,0.10352383613586427
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,32,2,balanced,0.10452735900878907
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,32,2,balanced,0.10580863952636718
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,32,2,balanced,0.1074713611602783
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,32,2,balanced,0.10964351654052736
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,32,2,balanced,0.12027263641357422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,32,2,balanced,0.13057920455932617
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,32,2,balanced,0.14031744003295898
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,32,2,balanced,0.16089984893798828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,32,2,balanced,0.1827097511291504
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,32,2,balanced,0.07201024055480956
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,32,2,balanced,0.09933823585510254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,32,2,power_law_1.01,0.0313369607925415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,32,2,power_law_1.01,0.035167999267578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,32,2,power_law_1.01,0.05798399925231934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,32,2,power_law_1.01,0.0666764783859253
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,32,2,power_law_1.01,0.08167679786682129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,32,2,power_law_1.01,0.08905471801757812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,32,2,power_law_1.01,0.0997657585144043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,32,2,power_law_1.01,0.10264448165893554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,32,2,power_law_1.01,0.10350975990295411
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,32,2,power_law_1.01,0.10479104042053222
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,32,2,power_law_1.01,0.1091648006439209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,32,2,power_law_1.01,0.1941721534729004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,32,2,power_law_1.01,0.19625856399536132
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,32,2,power_law_1.01,0.20377983093261717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,32,2,power_law_1.01,0.20692991256713866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,32,2,power_law_1.01,0.21440895080566405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,32,2,power_law_1.01,0.22767103195190427
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,32,2,power_law_1.01,0.25321088790893553
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,32,2,power_law_1.01,0.2800896072387695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,32,2,power_law_1.01,0.3334016036987305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,32,2,balanced,0.10036352157592773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,32,2,power_law_1.01,0.3859571075439453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,32,2,balanced,0.1000870418548584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,32,2,power_law_1.01,0.4917094421386719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,32,2,power_law_1.01,0.5994137573242188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,32,2,power_law_1.01,0.8373516845703126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,32,2,power_law_1.01,1.1013043212890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,32,2,power_law_1.01,0.03538687944412232
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,32,2,power_law_1.01,0.041125121116638186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,32,2,power_law_1.01,1.5489984130859376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,32,2,power_law_1.01,0.049862399101257324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,32,2,power_law_1.01,0.06684288024902343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,32,2,power_law_1.01,0.09377663612365723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,32,2,power_law_1.01,0.10107775688171387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,32,2,power_law_1.01,0.11405823707580567
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,32,2,power_law_1.01,2.0293017578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,32,2,power_law_1.01,0.11761792182922363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,32,2,power_law_1.01,0.11898624420166015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,32,2,power_law_1.01,0.12372991561889649
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,32,2,power_law_1.01,0.127128324508667
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,32,2,power_law_1.01,0.14510208129882812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,32,2,power_law_1.01,0.14790783882141115
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,32,2,power_law_1.01,0.15748607635498046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,32,2,power_law_1.01,0.1568883228302002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,32,2,power_law_1.01,0.165482234954834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,32,2,power_law_1.01,0.17130752563476564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,32,2,power_law_1.01,0.18863231658935548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,32,2,power_law_1.01,0.2137843132019043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,32,2,power_law_1.01,0.24977407455444336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,32,2,power_law_1.01,0.3088473510742188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,32,2,power_law_1.01,0.38911231994628903
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,32,2,power_law_1.01,0.4686604690551758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,32,2,power_law_1.01,0.6408844757080078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,32,2,power_law_1.01,0.8008716583251954
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,32,2,power_law_1.01,0.046268157958984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,32,2,power_law_1.01,1.126065902709961
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,32,2,power_law_1.01,0.049350399971008305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,32,2,balanced,0.2825817680358887
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,32,2,balanced,0.32702590942382814
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,32,2,power_law_1.01,0.05498496055603027
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,32,2,balanced,0.4664947128295898
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,32,2,power_law_1.01,0.06951935768127442
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,32,2,power_law_1.01,1.4482240295410156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,32,2,power_law_1.01,0.07554815769195557
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,32,2,power_law_1.01,0.08704000473022461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,32,2,power_law_1.01,0.0906060791015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,32,2,power_law_1.01,0.08845696449279786
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,32,2,power_law_1.01,0.09269760131835938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,32,2,power_law_1.01,0.09588607788085937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,32,2,power_law_1.01,0.09867136001586914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,32,2,power_law_1.01,0.10493824005126953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,32,2,power_law_1.01,0.10927103996276855
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,32,2,power_law_1.01,0.11425151824951171
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,32,2,power_law_1.01,0.12312447547912597
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,32,2,power_law_1.01,0.1537600040435791
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,32,2,power_law_1.01,0.17877248764038084
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,32,2,power_law_1.01,0.09041664123535156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,32,2,power_law_1.01,0.0608460807800293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,32,2,balanced,0.6035289764404297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,32,2,power_law_1.01,0.21388671874999998
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,32,2,power_law_1.01,0.2622694396972656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,32,2,power_law_1.01,0.45495040893554684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,32,2,power_law_1.2,0.034149119853973384
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,32,2,power_law_1.2,0.03484544038772583
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,32,2,power_law_1.01,0.3575398254394531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,32,2,power_law_1.2,0.05771903991699219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,32,2,power_law_1.2,0.0648691177368164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,32,2,power_law_1.2,0.07758975982666015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,32,2,power_law_1.2,0.09196800231933594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,32,2,power_law_1.2,0.10017919540405273
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,32,2,power_law_1.01,0.5953664016723633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,32,2,power_law_1.2,0.10116352081298827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,32,2,power_law_1.2,0.10353023529052734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,32,2,power_law_1.2,0.10736000061035156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,32,2,power_law_1.2,0.1125107192993164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,32,2,power_law_1.2,0.19255552291870118
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,32,2,power_law_1.2,0.19557504653930663
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,32,2,power_law_1.2,0.20515455245971678
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,32,2,power_law_1.2,0.2073651123046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,32,2,power_law_1.2,0.21484031677246093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,32,2,power_law_1.2,0.22861696243286134
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,32,2,power_law_1.2,0.25572864532470707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,32,2,power_law_1.2,0.28136831283569336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,32,2,power_law_1.2,0.33486591339111327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,32,2,power_law_1.2,0.3901721572875977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,32,2,power_law_1.2,0.49550720214843746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,32,2,power_law_1.2,0.6022745513916016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,32,2,power_law_1.2,0.868373794555664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,32,2,power_law_1.2,0.03897471904754639
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,32,2,power_law_1.2,0.041140480041503905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,32,2,power_law_1.2,1.1087462615966797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,32,2,power_law_1.2,0.04975615978240967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,32,2,power_law_1.2,1.575389404296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,32,2,power_law_1.2,0.06699135780334473
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,32,2,power_law_1.2,0.09004287719726563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,32,2,power_law_1.2,0.11411199569702149
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,32,2,power_law_1.2,0.11730943679809572
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,32,2,power_law_1.2,0.11985280036926269
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,32,2,power_law_1.2,2.0657139587402344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,32,2,power_law_1.2,0.1212275218963623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,32,2,power_law_1.2,0.1273305606842041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,32,2,power_law_1.2,0.1460300827026367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,32,2,power_law_1.2,0.09880831718444824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,32,2,power_law_1.2,0.1502400016784668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,32,2,power_law_1.2,0.1588710403442383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,32,2,power_law_1.2,0.16000383377075195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,32,2,power_law_1.2,0.16607872009277344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,32,2,power_law_1.2,0.17618047714233398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,32,2,power_law_1.2,0.19541759490966798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,32,2,power_law_1.2,0.22158336639404297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,32,2,power_law_1.2,0.2801203155517578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,32,2,power_law_1.2,0.3140979194641113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,32,2,power_law_1.2,0.3988735961914062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,32,2,power_law_1.2,0.47843070983886715
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,32,2,power_law_1.01,1.2457958221435548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,32,2,power_law_1.2,0.6491980743408203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,32,2,power_law_1.2,0.8111052703857421
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,32,2,power_law_1.2,1.1310054779052734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,32,2,power_law_1.2,0.04640128135681153
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,32,2,power_law_1.2,0.05375616073608398
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,32,2,power_law_1.2,0.059850239753723146
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,32,2,power_law_1.2,0.0647872018814087
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,32,2,power_law_1.2,0.07269248008728027
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,32,2,power_law_1.2,0.0885542392730713
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,32,2,power_law_1.2,0.08851200103759765
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,32,2,power_law_1.2,0.08760064125061036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,32,2,power_law_1.2,1.468783416748047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,32,2,power_law_1.2,0.08985600471496583
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,32,2,power_law_1.2,0.09460479736328124
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,32,2,power_law_1.2,0.09775360107421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,32,2,power_law_1.2,0.10223872184753417
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,32,2,power_law_1.2,0.048682241439819335
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,32,2,power_law_1.2,0.1153536033630371
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,32,2,power_law_1.2,0.12279680252075195
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,32,2,power_law_1.2,0.13999999999999999
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,32,2,power_law_1.2,0.20786304473876954
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,32,2,power_law_1.2,0.26273279190063475
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,32,2,power_law_1.2,0.33267585754394535
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,32,2,power_law_1.2,0.10855039596557617
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,32,2,power_law_1.01,1.7024140930175782
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,32,2,power_law_1.2,0.17576704025268555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,32,2,power_law_1.2,0.49053054809570307
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,64,balanced,0.019875839948654175
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,64,balanced,0.019560960531234742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,64,balanced,0.019880959987640383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,64,balanced,0.019865599870681764
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,64,balanced,0.019870719909667968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,64,balanced,0.03214591979980469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,64,balanced,0.03298559904098511
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,32,2,power_law_1.2,0.5840140914916991
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,64,balanced,0.03337728023529053
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,64,balanced,0.03455872058868408
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,64,balanced,0.034197759628295896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,64,balanced,0.03581183910369873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,64,balanced,0.03541759967803955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,64,balanced,0.03703808069229126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,64,balanced,0.037793281078338625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,64,balanced,0.03683712005615235
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,32,2,power_law_1.2,0.9618137359619141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,64,balanced,0.037482240200042725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,64,balanced,0.04134143829345703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,64,balanced,0.04074111938476562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,64,balanced,0.044176640510559084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,64,balanced,0.048474879264831544
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,64,balanced,0.05384960174560547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,64,balanced,0.06101376056671143
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,64,balanced,0.07776639938354493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,64,balanced,0.09758336067199706
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,32,2,power_law_1.2,1.271102752685547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,64,balanced,0.13181695938110352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,64,balanced,0.16591487884521486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,64,balanced,0.024332799911499024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,64,balanced,0.023901441097259522
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,64,balanced,0.02393984079360962
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,64,balanced,0.024364800453186036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,64,balanced,0.02398591995239258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,64,balanced,0.02930304050445557
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,64,balanced,0.029699840545654298
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,64,balanced,0.029632000923156737
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,64,balanced,0.03053440093994141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,64,balanced,0.030215680599212646
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,64,balanced,0.03070208072662354
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,64,balanced,0.03147264003753662
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,64,balanced,0.03133440017700195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,64,balanced,0.03307264089584351
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,64,balanced,0.03158655881881714
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,64,balanced,0.03255167961120605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,64,balanced,0.033381121158599855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,64,balanced,0.036290559768676754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,64,balanced,0.03255552053451538
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,64,balanced,0.03455487966537475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,64,balanced,0.037465600967407225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,64,balanced,0.04227071762084961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,64,balanced,0.04479487895965576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,64,balanced,0.05329535961151123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,64,balanced,0.0653004789352417
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,64,balanced,0.08710016250610352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,64,balanced,0.0460748815536499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,64,balanced,0.1070963191986084
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,1,64,balanced,0.014882559776306153
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,1,64,balanced,0.01449728012084961
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,1,64,balanced,0.014862079620361329
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,1,64,balanced,0.014958080053329468
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,1,64,balanced,0.01495296001434326
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,1,64,balanced,0.014885120391845702
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,1,64,balanced,0.014970879554748535
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,1,64,balanced,0.014951679706573486
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,1,64,balanced,0.015357439517974853
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,1,64,balanced,0.01486847996711731
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,1,64,balanced,0.014949120283126833
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,1,64,balanced,0.014539519548416138
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,1,64,balanced,0.015260159969329834
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,1,64,balanced,0.015368959903717042
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,1,64,balanced,0.014973440170288085
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,1,64,balanced,0.014935040473937988
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,1,64,balanced,0.014608639478683471
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,1,64,balanced,0.015025919675827027
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,1,64,balanced,0.015386879444122314
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,1,64,balanced,0.015363839864730836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,32,2,power_law_1.2,2.8882266235351564
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,1,64,balanced,0.016369919776916503
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,1,64,balanced,0.01821951985359192
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,1,64,balanced,0.01657088041305542
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,64,power_law_1.01,0.02022144079208374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,64,power_law_1.01,0.020276479721069336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,64,power_law_1.01,0.019775999784469606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,64,power_law_1.01,0.03302783966064453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,64,power_law_1.01,0.03374079942703247
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,1,64,balanced,0.013771519660949708
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,64,power_law_1.01,0.03410559892654419
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,64,power_law_1.01,0.03462912082672119
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,64,power_law_1.01,0.03622656106948853
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,64,power_law_1.01,0.03666687965393066
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,1,64,balanced,0.017431039810180664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,64,power_law_1.01,0.04010752201080322
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,64,power_law_1.01,0.0410752010345459
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,1,64,balanced,0.02147327899932861
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,64,power_law_1.01,0.04360191822052002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,64,power_law_1.01,0.03999104022979737
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,64,power_law_1.01,0.04397696018218994
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,64,power_law_1.01,0.045669121742248533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,64,power_law_1.01,0.04565887928009033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,64,power_law_1.01,0.0522214412689209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,64,power_law_1.01,0.06208255767822266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,64,power_law_1.01,0.08440959930419922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,64,power_law_1.01,0.105665283203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,64,power_law_1.01,0.14180095672607423
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,64,power_law_1.01,0.21121408462524416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,64,power_law_1.01,0.30424448013305666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,64,power_law_1.01,0.4174937438964844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,64,power_law_1.01,0.5216191864013672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,64,power_law_1.01,0.024778239727020264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,64,power_law_1.01,0.024435200691223145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,64,power_law_1.01,0.024362239837646484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,64,power_law_1.01,0.0298304009437561
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,64,power_law_1.01,0.029684479236602786
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,64,power_law_1.01,0.030453760623931885
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,64,power_law_1.01,0.030641920566558838
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,64,power_law_1.01,0.031485440731048586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,64,power_law_1.01,0.03214207887649536
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,64,power_law_1.01,0.03145472049713135
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,64,power_law_1.01,0.03348992109298706
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,64,power_law_1.01,0.03643136024475098
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,64,power_law_1.01,0.7603033447265626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,64,power_law_1.01,0.03469311952590942
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,64,power_law_1.01,0.036634879112243654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,64,power_law_1.01,0.03623424053192138
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,64,power_law_1.01,0.04073984146118164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,64,power_law_1.01,0.04762112140655518
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,64,power_law_1.01,0.05672575950622558
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,64,power_law_1.01,0.06812032222747802
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,64,power_law_1.01,0.09788800239562988
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,64,power_law_1.01,0.11738368034362794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,64,power_law_1.01,0.17474687576293946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,64,power_law_1.01,0.2260915184020996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,64,power_law_1.01,0.30562944412231446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,64,power_law_1.01,0.03253504037857056
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,1,64,power_law_1.01,0.014536960124969484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,64,power_law_1.01,0.035109119415283205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,64,power_law_1.01,0.4009907150268555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,1,64,power_law_1.01,0.014895360469818115
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,1,64,power_law_1.01,0.015024640560150146
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,1,64,power_law_1.01,0.014940160512924194
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,1,64,power_law_1.01,0.014931199550628663
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,1,64,power_law_1.01,0.015336960554122925
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,1,64,power_law_1.01,0.01536128044128418
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,1,64,power_law_1.01,0.015023360252380371
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,1,64,power_law_1.01,0.015363839864730836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,1,64,power_law_1.01,0.015034879446029664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,1,64,power_law_1.01,0.014853119850158691
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,1,64,power_law_1.01,0.01533951997756958
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,1,64,power_law_1.01,0.01528447985649109
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,1,64,power_law_1.01,0.015329279899597167
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,1,64,power_law_1.01,0.014956799745559694
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,1,64,balanced,0.024447999000549316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,64,power_law_1.01,0.034544639587402344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,1,64,power_law_1.01,0.020295679569244385
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,1,64,power_law_1.01,0.014951679706573486
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,1,64,power_law_1.01,0.015453439950942994
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,1,64,power_law_1.01,0.01709439992904663
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,1,64,power_law_1.01,0.016600320339202883
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,1,64,power_law_1.01,0.0178547203540802
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,64,power_law_1.2,0.020273919105529788
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,64,power_law_1.2,0.020262401103973388
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,64,power_law_1.2,0.020697600841522217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,64,power_law_1.2,0.03255552053451538
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,64,power_law_1.2,0.033327360153198246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,64,power_law_1.2,0.03377919912338257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,64,power_law_1.2,0.035281920433044435
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,64,power_law_1.2,0.036652801036834715
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,1,64,power_law_1.01,0.025183999538421632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,64,power_law_1.2,0.036646399497985836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,64,power_law_1.2,0.03831935882568359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,64,power_law_1.2,0.04113279819488526
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,64,power_law_1.2,0.04110464096069336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,64,power_law_1.2,0.042769918441772456
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,64,power_law_1.2,0.04359551906585694
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,64,power_law_1.2,0.046856961250305175
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,64,power_law_1.2,0.046900482177734376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,64,power_law_1.2,0.05214335918426514
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,64,power_law_1.2,0.06187007904052735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,64,power_law_1.2,0.08109439849853516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,64,power_law_1.2,0.11018879890441895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,64,power_law_1.2,0.15153023719787598
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,64,power_law_1.2,0.1976038360595703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,64,power_law_1.2,0.28738304138183596
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,64,power_law_1.2,0.4283468627929688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,1,64,power_law_1.01,0.02887808084487915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,64,power_law_1.2,0.8021683502197267
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,64,power_law_1.2,0.024435200691223145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,64,power_law_1.2,0.023960320949554442
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,64,power_law_1.2,0.029290239810943603
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,64,power_law_1.2,0.02962048053741455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,64,power_law_1.2,0.030088961124420166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,64,power_law_1.2,0.03096575975418091
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,64,power_law_1.2,0.03173120021820068
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,1,64,power_law_1.01,0.045772800445556636
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,64,power_law_1.2,0.031737599372863766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,64,power_law_1.2,0.032112638950347897
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,64,power_law_1.2,0.032496640682220454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,64,power_law_1.2,0.03339135885238648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,64,power_law_1.2,0.03504384040832519
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,64,power_law_1.2,0.03455744028091431
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,64,power_law_1.2,0.034544639587402344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,1,64,power_law_1.01,0.060648961067199705
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,1,64,power_law_1.01,0.07475711822509765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,64,power_law_1.2,0.035361280441284185
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,64,power_law_1.2,0.03804415941238403
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,64,power_law_1.2,0.04285439968109131
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,64,power_law_1.2,0.053140478134155275
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,64,power_law_1.2,0.07067520141601562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,64,power_law_1.2,0.0837388801574707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,64,power_law_1.2,0.13370752334594727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,64,power_law_1.2,0.03668864011764526
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,64,power_law_1.2,0.1764646339416504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,64,power_law_1.2,0.5912511825561524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,64,power_law_1.2,0.3401023864746094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,1,64,power_law_1.2,0.0149017596244812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,1,64,power_law_1.2,0.014885120391845702
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,1,64,power_law_1.2,0.014977279901504516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,1,64,power_law_1.2,0.014942719936370849
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,1,64,power_law_1.2,0.01496832013130188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,1,64,power_law_1.2,0.014877439737319947
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,1,64,power_law_1.2,0.015352319478988647
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,1,64,power_law_1.2,0.015283199548721312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,1,64,power_law_1.2,0.014928640127182008
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,1,64,power_law_1.2,0.014890240430831907
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,1,64,power_law_1.2,0.015340800285339356
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,1,64,power_law_1.01,0.033696000576019285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,1,64,power_law_1.2,0.015301120281219483
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,1,64,power_law_1.2,0.014965759515762327
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,1,64,power_law_1.2,0.015447039604187012
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,64,power_law_1.2,1.1914252471923827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,64,power_law_1.2,0.28481536865234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,64,power_law_1.2,0.5327807998657226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,64,power_law_1.2,0.7168434906005859
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,1,64,power_law_1.2,0.014657280445098878
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,1,64,power_law_1.2,0.015709439516067507
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,1,64,power_law_1.2,0.016119040250778198
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,1,64,power_law_1.2,0.016910079717636108
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,1,64,power_law_1.2,0.01788032054901123
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,1,64,power_law_1.2,0.020236799716949465
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,1,64,power_law_1.2,0.023976960182189942
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,32,balanced,0.01823232054710388
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,32,balanced,0.017824000120162962
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,1,64,power_law_1.2,0.08782719612121583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,32,balanced,0.018847999572753907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,32,balanced,0.019427839517593384
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,32,balanced,0.03010175943374634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,32,balanced,0.03133568048477173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,32,balanced,0.03174911975860596
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,32,balanced,0.03219712018966675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,32,balanced,0.03337088108062744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,32,balanced,0.03337856054306031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,32,balanced,0.034977281093597413
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,32,balanced,0.03542272090911865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,32,balanced,0.0374835205078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,32,balanced,0.03789952039718628
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,32,balanced,0.039580159187316895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,32,balanced,0.03783551931381225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,32,balanced,0.04116608142852783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,32,balanced,0.04314496040344239
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,32,balanced,0.049420800209045414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,32,balanced,0.053454079627990723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,32,balanced,0.016631040573120117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,32,balanced,0.06867584228515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,32,balanced,0.1152460765838623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,32,balanced,0.08241151809692383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,32,balanced,0.14104960441589356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,32,balanced,0.19054079055786133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,32,balanced,0.2484467124938965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,32,balanced,0.023170559406280516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,32,balanced,0.02254591941833496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,32,balanced,0.02429824113845825
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,1,64,power_law_1.2,0.02961024045944214
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,32,balanced,0.023557119369506836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,32,balanced,0.028492801189422608
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,32,balanced,0.02968064069747925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,32,balanced,0.02930560111999512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,32,balanced,0.030559999942779543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,32,balanced,0.030580480098724366
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,32,balanced,0.030521600246429442
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,32,balanced,0.031299839019775386
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,32,balanced,0.03255808115005493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,32,balanced,0.03362175941467285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,32,balanced,0.03503871917724609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,32,balanced,0.03089024066925049
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,32,balanced,0.03255808115005493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,32,balanced,0.02309760093688965
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,1,64,power_law_1.2,0.037311999797821044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,32,balanced,0.03505151987075806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,32,balanced,0.037080318927764894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,32,balanced,0.040285439491271974
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,32,balanced,0.05562880039215088
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,32,balanced,0.049053440093994136
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,32,balanced,0.07654784202575685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,32,balanced,0.0953651237487793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,32,balanced,0.1329689598083496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,32,balanced,0.17030527114868163
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,1,64,power_law_1.2,0.0607257604598999
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,1,64,power_law_1.2,0.049072642326354984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,32,balanced,0.029835519790649412
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,1,64,power_law_1.2,0.11052543640136718
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,2,32,balanced,0.014942719936370849
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,2,32,balanced,0.015383039712905883
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,2,32,balanced,0.014896639585494994
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,2,32,balanced,0.01537279963493347
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,2,32,balanced,0.015308799743652343
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,2,32,balanced,0.015333119630813598
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,2,32,balanced,0.01576575994491577
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,2,32,balanced,0.015776000022888183
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,2,32,balanced,0.015555839538574218
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,2,32,balanced,0.015416320562362671
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,2,32,balanced,0.015718400478363037
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,2,32,balanced,0.015802880525588987
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,2,32,balanced,0.015802880525588987
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,2,32,balanced,0.016129280328750613
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,2,32,balanced,0.016235519647598267
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,2,32,balanced,0.014469120502471924
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,2,32,balanced,0.0157695996761322
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,2,32,balanced,0.01528447985649109
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,2,32,balanced,0.015357439517974853
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,2,32,balanced,0.015357439517974853
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,32,power_law_1.01,0.01864192008972168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,32,power_law_1.01,0.018568960428237916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,32,power_law_1.01,0.018622720241546632
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,2,32,balanced,0.015269119739532471
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,32,power_law_1.01,0.02949631929397583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,32,power_law_1.01,0.031249918937683106
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,32,power_law_1.01,0.031650559902191164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,32,power_law_1.01,0.033807361125946046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,32,power_law_1.01,0.03443711996078491
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,32,power_law_1.01,0.03493887901306152
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,32,power_law_1.01,0.03626751899719238
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,32,power_law_1.01,0.03880959987640381
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,32,power_law_1.01,0.039444479942321774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,32,power_law_1.01,0.04274303913116455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,32,power_law_1.01,0.03990655899047851
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,32,power_law_1.01,0.04124159812927246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,32,power_law_1.01,0.043217921257019044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,32,power_law_1.01,0.04667776107788086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,32,power_law_1.01,0.05936895847320557
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,32,power_law_1.01,0.06867584228515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,32,power_law_1.01,0.08373632431030273
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,2,32,balanced,0.020614399909973144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,32,power_law_1.01,0.09527680397033692
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,32,power_law_1.01,0.13615360260009765
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,2,32,balanced,0.017774080038070676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,32,power_law_1.01,0.2675801658630371
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,2,32,balanced,0.017757439613342287
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,32,power_law_1.01,0.3746803283691406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,32,power_law_1.01,0.4360575866699219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,32,power_law_1.01,0.02439039945602417
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,32,power_law_1.01,0.023633921146392824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,32,power_law_1.01,0.023562240600585937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,2,32,balanced,0.03106688022613525
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,32,power_law_1.01,0.028476159572601318
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,32,power_law_1.01,0.717286376953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,32,power_law_1.01,0.028857600688934327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,32,power_law_1.01,0.029785599708557133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,32,power_law_1.01,0.028844799995422367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,32,power_law_1.01,0.03132927894592285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,32,power_law_1.01,0.030611200332641603
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,32,power_law_1.01,0.030926079750061036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,32,power_law_1.01,0.032194559574127195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,32,power_law_1.01,0.033320960998535154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,32,power_law_1.01,0.03542144060134887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,32,power_law_1.01,0.03418623924255371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,32,power_law_1.01,0.03328128099441528
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,32,power_law_1.01,0.034919679164886475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,32,power_law_1.01,0.03667072057723999
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,32,power_law_1.01,0.041948161125183105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,32,power_law_1.01,0.1746291160583496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,32,power_law_1.01,0.04981247901916504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,32,power_law_1.01,0.057239041328430176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,32,power_law_1.01,0.062196478843688965
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,2,32,balanced,0.021896960735321044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,32,power_law_1.01,0.09544575691223145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,32,power_law_1.01,0.11760255813598633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,32,power_law_1.01,0.1473459243774414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,32,power_law_1.01,0.21033088684082032
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,2,32,power_law_1.01,0.015344640016555786
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,32,power_law_1.01,0.2815014457702637
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,2,32,power_law_1.01,0.015311360359191895
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,2,32,power_law_1.01,0.015307519435882568
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,2,32,power_law_1.01,0.015276800394058227
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,2,32,power_law_1.01,0.01533951997756958
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,2,32,power_law_1.01,0.015358719825744629
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,2,32,power_law_1.01,0.015352319478988647
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,2,32,power_law_1.01,0.01578112006187439
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,2,32,power_law_1.01,0.015709439516067507
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,2,32,power_law_1.01,0.015522559881210327
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,2,32,power_law_1.01,0.0157043194770813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,2,32,power_law_1.01,0.015751680135726927
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,2,32,power_law_1.01,0.015763200521469116
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,2,32,power_law_1.01,0.01580415964126587
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,2,32,power_law_1.01,0.01701120018959045
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,2,32,power_law_1.01,0.0156876802444458
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,32,power_law_1.01,0.3802316665649414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,2,32,balanced,0.02627840042114258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,2,32,power_law_1.01,0.01775871992111206
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,2,32,power_law_1.01,0.016622079610824587
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,2,32,power_law_1.01,0.01888383984565735
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,2,32,power_law_1.01,0.02553855895996094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,2,32,power_law_1.01,0.021457920074462893
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,32,power_law_1.2,0.019389439821243287
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,32,power_law_1.2,0.019097599983215332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,2,32,power_law_1.01,0.021089279651641847
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,32,power_law_1.2,0.019066879749298098
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,32,power_law_1.2,0.02956928014755249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,32,power_law_1.2,0.031784958839416504
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,2,32,power_law_1.01,0.029021439552307127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,32,power_law_1.2,0.03337215900421143
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,32,power_law_1.2,0.034146559238433835
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,2,32,power_law_1.01,0.04370560169219971
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,32,power_law_1.2,0.03558271884918213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,32,power_law_1.2,0.03665920019149781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,32,power_law_1.2,0.038344960212707516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,32,power_law_1.2,0.04102528095245361
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,32,power_law_1.2,0.040767998695373536
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,32,power_law_1.2,0.04146560192108154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,32,power_law_1.2,0.042731518745422366
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,32,power_law_1.2,0.04314239978790283
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,32,power_law_1.2,0.04769023895263672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,32,power_law_1.2,0.06491392135620118
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,32,power_law_1.2,0.07673984050750733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,32,power_law_1.2,0.09465855598449707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,32,power_law_1.2,0.126878719329834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,32,power_law_1.2,0.19137279510498045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,32,power_law_1.2,0.03173887968063355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,32,power_law_1.2,0.21396863937377927
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,32,power_law_1.2,0.023978240489959717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,32,power_law_1.2,0.023110399246215822
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,32,power_law_1.2,0.02308095932006836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,32,power_law_1.2,0.028017919063568115
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,32,power_law_1.2,0.02855168104171753
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,32,power_law_1.2,0.02882431983947754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,32,power_law_1.2,0.030059518814086916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,32,power_law_1.2,0.03006335973739624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,32,power_law_1.2,0.03087104082107544
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,32,power_law_1.2,0.031248641014099122
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,32,power_law_1.2,0.033077759742736815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,32,power_law_1.2,0.03512448072433472
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,32,power_law_1.2,0.4839641571044922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,32,power_law_1.2,0.034149119853973384
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,32,power_law_1.2,0.03378943920135498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,32,power_law_1.2,0.034213120937347415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,32,power_law_1.2,0.03591552019119263
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,32,power_law_1.2,0.038293759822845455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,32,power_law_1.2,0.3508095932006836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,32,power_law_1.2,0.044743680953979494
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,32,power_law_1.2,0.05354879856109619
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,32,power_law_1.2,0.06404096126556397
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,32,power_law_1.2,0.08092415809631348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,32,power_law_1.2,0.10998016357421876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,32,power_law_1.2,0.747778549194336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,32,power_law_1.2,0.1517695999145508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,32,power_law_1.2,0.21374719619750976
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,2,32,power_law_1.2,0.015383039712905883
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,2,32,power_law_1.2,0.015296000242233276
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,32,power_law_1.2,0.988279037475586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,2,32,power_law_1.2,0.015278079509735107
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,2,32,power_law_1.2,0.015331840515136719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,2,32,power_law_1.2,0.01540992021560669
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,32,power_law_1.2,0.32238975524902347
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,2,32,power_law_1.2,0.015406080484390258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,2,32,power_law_1.2,0.01529728055000305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,2,32,power_law_1.2,0.01542655944824219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,2,32,power_law_1.2,0.015349760055541992
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,2,32,power_law_1.2,0.015411200523376464
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,2,32,power_law_1.2,0.01575935959815979
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,2,32,power_law_1.2,0.015751680135726927
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,2,32,power_law_1.2,0.015749119520187378
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,2,32,power_law_1.2,0.016632319688796998
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,2,32,power_law_1.2,0.016984319686889647
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,2,32,power_law_1.2,0.01570688009262085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,32,power_law_1.2,0.46850944519042964
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,2,32,power_law_1.01,0.05292672157287598
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,2,32,power_law_1.01,0.06108416080474853
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,2,32,power_law_1.01,0.07913343906402588
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,2,32,power_law_1.2,0.019801599979400637
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,2,32,power_law_1.2,0.018278399705886843
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,2,32,power_law_1.2,0.017470719814300536
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,2,32,power_law_1.2,0.021524479389190672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,2,32,power_law_1.2,0.024039680957794192
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,4,16,balanced,0.017506560087203978
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,4,16,balanced,0.017849600315093993
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,16,balanced,0.018211840391159057
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,16,balanced,0.018191360235214234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,16,balanced,0.020582399368286132
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,2,32,power_law_1.2,0.03583872079849243
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,2,32,power_law_1.2,0.04205567836761474
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,16,balanced,0.03131903886795044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,16,balanced,0.034206719398498536
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,16,balanced,0.03418368101119995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,16,balanced,0.03460736036300659
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,16,balanced,0.033724160194396974
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,16,balanced,0.03618304014205932
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,32,power_law_1.2,0.6071884918212891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,16,balanced,0.03805056095123291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,16,balanced,0.0395417594909668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,16,balanced,0.03623552083969116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,16,balanced,0.039508481025695805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,16,balanced,0.03755520105361938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,16,balanced,0.04272895812988281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,16,balanced,0.04687744140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,16,balanced,0.05039743900299072
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,16,balanced,0.0643225622177124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,16,balanced,0.07681024074554443
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,16,balanced,0.10652031898498535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,16,balanced,0.13015040397644043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,16,balanced,0.1782476806640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,16,balanced,0.22582656860351563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,16,balanced,0.022312960624694823
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,16,balanced,0.31801216125488285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,16,balanced,0.023096320629119874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,16,balanced,0.022780160903930664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,16,balanced,0.4145843124389648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,16,balanced,0.023878400325775144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,16,balanced,0.023956480026245116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,16,balanced,0.029667840003967282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,16,balanced,0.03175551891326904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,16,balanced,0.03090431928634644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,16,balanced,0.030854399204254153
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,16,balanced,0.03163520097732544
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,16,balanced,0.032958719730377194
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,16,balanced,0.03379839897155762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,16,balanced,0.034973440170288084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,16,balanced,0.0309171199798584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,16,balanced,0.03255424022674561
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,16,balanced,0.03453439950942993
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,16,balanced,0.035526399612426755
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,16,balanced,0.03705984115600586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,16,balanced,0.04181119918823242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,16,balanced,0.04600448131561279
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,16,balanced,0.05296895980834961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,16,balanced,0.07646848201751709
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,16,balanced,0.09043328285217285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,16,balanced,0.12650112152099607
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,16,balanced,0.15944576263427734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,16,balanced,0.2287654495239258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,16,balanced,0.2976140785217285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,2,32,power_law_1.2,0.0574886417388916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,2,32,power_law_1.2,0.06028031826019287
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,4,16,balanced,0.015776000022888183
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,4,16,balanced,0.01690623998641968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,4,16,balanced,0.01660159945487976
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,4,16,balanced,0.015752960443496705
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,4,16,balanced,0.0169267201423645
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,4,16,balanced,0.01654144048690796
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,4,16,balanced,0.01659775972366333
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,4,16,balanced,0.01693184018135071
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,4,16,balanced,0.01698815941810608
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,4,16,balanced,0.016532479524612426
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,4,16,balanced,0.01660416007041931
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,4,16,balanced,0.01689087986946106
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,2,32,power_law_1.2,0.08596223831176758
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,4,16,balanced,0.016980479955673217
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,4,16,balanced,0.015776000022888183
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,4,16,balanced,0.01734784007072449
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,4,16,balanced,0.01718016028404236
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,4,16,balanced,0.017840640544891356
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,4,16,balanced,0.01777791976928711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,4,16,balanced,0.016231679916381837
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,2,32,power_law_1.2,0.11837311744689942
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,4,16,balanced,0.016550400257110597
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,4,16,power_law_1.01,0.018984960317611693
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,4,16,power_law_1.01,0.018615039587020873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,16,power_law_1.01,0.019843839406967163
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,16,power_law_1.01,0.02029695987701416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,16,power_law_1.01,0.030908160209655762
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,4,16,balanced,0.02511359930038452
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,16,power_law_1.01,0.03243776082992554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,16,power_law_1.01,0.03350271940231323
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,4,16,balanced,0.021108479499816896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,16,power_law_1.01,0.0347379207611084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,16,power_law_1.01,0.03581824064254761
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,4,16,balanced,0.01707263946533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,16,power_law_1.01,0.03610624074935913
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,4,16,balanced,0.021527040004730224
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,16,power_law_1.01,0.03990272045135498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,16,power_law_1.01,0.041649918556213375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,16,power_law_1.01,0.039984641075134275
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,16,power_law_1.01,0.03888384103775024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,16,power_law_1.01,0.04125696182250976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,16,power_law_1.01,0.043162879943847654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,16,power_law_1.01,0.046488318443298336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,16,power_law_1.01,0.06612095832824708
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,16,power_law_1.01,0.07450240135192872
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,16,power_law_1.01,0.08815232276916504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,16,power_law_1.01,0.10598272323608397
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,16,power_law_1.01,0.1386406421661377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,16,power_law_1.01,0.17814144134521484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,16,power_law_1.01,0.26398591995239257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,16,power_law_1.01,0.3076902389526367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,16,power_law_1.01,0.024789760112762453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,16,power_law_1.01,0.02349695920944214
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,16,power_law_1.01,0.4977945709228516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,16,power_law_1.01,0.023613440990447997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,16,power_law_1.01,0.02434943914413452
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,16,power_law_1.01,0.6726822662353515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,16,power_law_1.01,0.030092799663543703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,16,power_law_1.01,0.02973952054977417
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,16,power_law_1.01,0.03130111932754517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,16,power_law_1.01,0.031320319175720215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,16,power_law_1.01,0.03246592044830322
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,16,power_law_1.01,0.03215359926223755
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,16,power_law_1.01,0.03549567937850952
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,16,power_law_1.01,0.03464576005935669
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,16,power_law_1.01,0.03316096067428589
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,16,power_law_1.01,0.03507839918136597
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,16,power_law_1.01,0.036038401126861574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,16,power_law_1.01,0.038113279342651366
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,16,power_law_1.01,0.04235648155212403
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,16,power_law_1.01,0.05155327796936036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,16,power_law_1.01,0.06062592029571533
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,4,16,balanced,0.028419840335845947
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,16,power_law_1.01,0.09790335655212402
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,16,power_law_1.01,0.1242854404449463
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,16,power_law_1.01,0.18795904159545898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,16,power_law_1.01,0.27741695404052735
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,4,16,power_law_1.01,0.016596479415893556
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,16,power_law_1.01,0.38794750213623047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,16,power_law_1.01,0.03377408027648926
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,4,16,power_law_1.01,0.01656191945075989
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,4,16,power_law_1.01,0.016559360027313234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,4,16,power_law_1.01,0.016522239446640014
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,4,16,power_law_1.01,0.01662335991859436
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,4,16,power_law_1.01,0.016588799953460694
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,4,16,power_law_1.01,0.01654144048690796
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,4,16,power_law_1.01,0.016610560417175294
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,4,16,power_law_1.01,0.016546560525894167
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,4,16,power_law_1.01,0.01681663990020752
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,4,16,power_law_1.01,0.016999679803848266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,16,power_law_1.01,0.08004351615905761
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,4,16,power_law_1.01,0.016586240530014038
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,4,16,power_law_1.01,0.016948479413986205
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,4,16,power_law_1.01,0.017157119512557984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,4,16,power_law_1.01,0.017350399494171144
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,4,16,power_law_1.01,0.01895424008369446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,16,power_law_1.01,0.46204799652099615
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,4,16,balanced,0.03645951986312866
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,4,16,balanced,0.044072961807250975
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,4,16,power_law_1.01,0.024446721076965335
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,4,16,power_law_1.01,0.026667520999908444
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,4,16,power_law_1.01,0.03328511953353882
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,4,16,power_law_1.01,0.021080319881439206
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,4,16,power_law_1.01,0.01822975993156433
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,4,16,power_law_1.2,0.01863039970397949
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,4,16,power_law_1.2,0.01947391986846924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,16,power_law_1.2,0.019879679679870605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,16,power_law_1.2,0.021077759265899658
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,16,power_law_1.2,0.030499839782714845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,16,power_law_1.2,0.032529919147491454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,16,power_law_1.2,0.03380095958709717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,16,power_law_1.2,0.03461888074874878
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,4,16,power_law_1.01,0.019856640100479127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,16,power_law_1.2,0.03543040037155151
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,16,power_law_1.2,0.03664511919021606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,16,power_law_1.2,0.03949440002441406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,16,power_law_1.2,0.041463041305541994
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,16,power_law_1.2,0.040378880500793454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,16,power_law_1.2,0.04032256126403809
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,16,power_law_1.2,0.04236671924591064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,16,power_law_1.2,0.044972801208496095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,16,power_law_1.2,0.05171455860137939
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,16,power_law_1.2,0.07802879810333252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,16,power_law_1.2,0.09351679801940918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,16,power_law_1.2,0.11858943939208984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,16,power_law_1.2,0.1654924774169922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,16,power_law_1.2,0.06808320045471192
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,16,power_law_1.2,0.2118707275390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,4,16,power_law_1.01,0.040919041633605956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,16,power_law_1.2,0.30375295639038086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,16,power_law_1.2,0.02351871967315674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,16,power_law_1.2,0.023658239841461183
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,16,power_law_1.2,0.02395008087158203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,16,power_law_1.2,0.023970561027526854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,16,power_law_1.2,0.029706239700317383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,16,power_law_1.2,0.03057152032852173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,16,power_law_1.2,0.030959360599517823
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,16,power_law_1.2,0.03166079998016357
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,16,power_law_1.2,0.031760640144348144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,16,power_law_1.2,0.42499328613281245
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,16,power_law_1.2,0.03296639919281006
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,4,16,power_law_1.01,0.04852352142333984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,16,power_law_1.2,0.034579200744628905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,16,power_law_1.2,0.03502079963684082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,16,power_law_1.2,0.034593279361724856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,16,power_law_1.2,0.03341567993164062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,16,power_law_1.2,0.03631999969482422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,16,power_law_1.2,0.0374399995803833
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,16,power_law_1.2,0.03895040035247803
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,16,power_law_1.2,0.04804351806640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,16,power_law_1.2,0.055239682197570805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,16,power_law_1.2,0.06501120090484619
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,16,power_law_1.2,0.08110079765319825
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,16,power_law_1.2,0.10690815925598145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,16,power_law_1.2,0.1397747230529785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,16,power_law_1.2,0.2370560073852539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,16,power_law_1.2,0.2967935943603516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,16,power_law_1.2,0.6212876892089844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,4,16,power_law_1.01,0.056163840293884285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,4,16,power_law_1.2,0.016569600105285645
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,4,16,power_law_1.2,0.016641279458999632
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,4,16,power_law_1.2,0.01649791955947876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,4,16,power_law_1.2,0.016537599563598633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,4,16,power_law_1.2,0.0163481605052948
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,4,16,power_law_1.01,0.07829376220703124
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,4,16,power_law_1.2,0.016741119623184204
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,4,16,power_law_1.01,0.10159999847412109
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,4,16,power_law_1.2,0.016640000343322754
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,4,16,power_law_1.2,0.016573439836502075
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,4,16,power_law_1.2,0.01694208025932312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,4,16,power_law_1.2,0.01696895956993103
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,4,16,power_law_1.2,0.016976640224456788
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,4,16,power_law_1.2,0.017006080150604248
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,4,16,power_law_1.2,0.018617600202560425
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,4,16,power_law_1.2,0.0165721595287323
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,16,power_law_1.2,0.4575462341308594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,4,16,power_law_1.2,0.024732160568237304
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,4,16,power_law_1.2,0.018256640434265135
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,4,16,power_law_1.2,0.018237440586090087
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,4,16,power_law_1.2,0.019429119825363158
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,4,16,power_law_1.2,0.021502718925476075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,16,power_law_1.2,0.8485657501220704
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,4,16,power_law_1.2,0.02228096008300781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,16,power_law_1.2,0.5975206375122071
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,64,balanced,0.02764287948608398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,64,balanced,0.027639040946960448
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,64,balanced,0.028064000606536865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,64,balanced,0.027628800868988036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,64,balanced,0.030123519897460937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,64,balanced,0.06541183948516846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,64,balanced,0.06646399974822999
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,64,balanced,0.06629119873046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,64,balanced,0.06185728073120117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,64,balanced,0.0609600019454956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,64,balanced,0.06227200031280518
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,64,balanced,0.08022527694702149
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,64,balanced,0.08265088081359864
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,64,balanced,0.08534784317016601
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,64,balanced,0.08233856201171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,64,balanced,0.08253055572509765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,64,balanced,0.08759167671203613
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,64,balanced,0.10132224082946777
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,64,balanced,0.09684991836547852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,64,balanced,0.10657024383544922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,64,balanced,0.11475456237792969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,64,balanced,0.12575103759765624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,64,balanced,0.1391577625274658
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,64,balanced,0.17237119674682616
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,64,balanced,0.20636032104492186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,64,balanced,0.29705215454101563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,64,balanced,0.37266433715820313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,64,balanced,0.03051136016845703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,4,16,power_law_1.2,0.03276031970977783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,64,balanced,0.030897920131683347
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,64,balanced,0.029850881099700927
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,64,balanced,0.030497279167175297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,64,balanced,0.03068543910980225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,64,balanced,0.03789184093475342
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,64,balanced,0.03823359966278076
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,64,balanced,0.038353919982910156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,64,balanced,0.03936768054962158
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,64,balanced,0.03953151941299439
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,64,balanced,0.0399232006072998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,64,balanced,0.04081791877746582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,64,balanced,0.04190847873687744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,64,balanced,0.04319231986999512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,64,balanced,0.04018815994262696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,64,balanced,0.04108287811279297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,64,balanced,0.04238080024719239
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,64,balanced,0.04526847839355469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,64,balanced,0.04338687896728516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,64,balanced,0.049354238510131834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,64,balanced,0.05576320171356201
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,64,balanced,0.06228223800659179
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,64,balanced,0.07197824001312256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,64,balanced,0.08912896156311036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,64,balanced,0.10996992111206054
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,64,balanced,0.1540774440765381
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,64,balanced,0.19659391403198243
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,4,16,power_law_1.2,0.036887040138244634
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,4,16,power_law_1.2,0.048078079223632816
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,4,16,power_law_1.2,0.07825664043426514
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,4,16,power_law_1.2,0.06729472160339356
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,4,16,power_law_1.2,0.11598848342895507
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,1,64,balanced,0.017008639574050903
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,1,64,balanced,0.016963839530944824
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,1,64,balanced,0.017415679693222046
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,1,64,balanced,0.017427200078964235
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,1,64,balanced,0.017401599884033205
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,1,64,balanced,0.01741055965423584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,1,64,balanced,0.017345279455184937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,1,64,balanced,0.017408000230789183
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,1,64,balanced,0.016953599452972413
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,1,64,balanced,0.017344000339508055
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,1,64,balanced,0.017619199752807617
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,1,64,balanced,0.017392640113830564
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,1,64,balanced,0.01775871992111206
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,1,64,balanced,0.017826559543609618
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,1,64,balanced,0.01820799946784973
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,1,64,balanced,0.01820799946784973
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,1,64,balanced,0.01824128031730652
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,1,64,balanced,0.018191360235214234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,1,64,balanced,0.017404160499572753
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,1,64,balanced,0.01739135980606079
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,1,64,balanced,0.017433600425720213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,64,power_law_1.01,0.029619200229644777
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,64,power_law_1.01,0.03486848115921021
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,64,power_law_1.01,0.03532543897628784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,64,power_law_1.01,0.06744063854217529
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,64,power_law_1.01,0.06957952022552491
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,64,power_law_1.01,0.07234560012817383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,64,power_law_1.01,0.08317055702209472
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,64,power_law_1.01,0.08648832321166992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,64,power_law_1.01,0.08589568138122558
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,64,power_law_1.01,0.08911231994628907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,64,power_law_1.01,0.08632191658020019
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,64,power_law_1.01,0.09274239540100097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,64,power_law_1.01,0.09604736328124999
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,64,power_law_1.01,0.10129152297973634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,64,power_law_1.01,0.10123007774353028
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,64,power_law_1.01,0.10451583862304688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,64,power_law_1.01,0.11105024337768554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,64,power_law_1.01,0.1327564811706543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,64,power_law_1.01,0.15178239822387696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,64,power_law_1.01,0.1781068801879883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,64,power_law_1.01,0.2450547218322754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,64,power_law_1.01,0.3763558578491211
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,1,64,balanced,0.020211200714111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,1,64,balanced,0.023850240707397462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,64,power_law_1.01,0.43228801727294924
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,1,64,balanced,0.020913920402526855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,64,power_law_1.01,0.6416191864013672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,1,64,balanced,0.028494079113006592
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,64,power_law_1.01,0.03086335897445679
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,64,power_law_1.01,0.8980300903320313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,64,power_law_1.01,0.030855679512023927
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,64,power_law_1.01,0.0389081597328186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,64,power_law_1.01,1.2987225341796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,64,power_law_1.01,0.03912447929382325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,64,power_law_1.01,0.030696959495544435
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,64,power_law_1.01,0.03914367914199829
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,64,power_law_1.01,0.04157311916351318
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,64,power_law_1.01,0.04080639839172363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,64,power_law_1.01,0.0415334415435791
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,64,power_law_1.01,0.04293759822845459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,64,power_law_1.01,1.859453430175781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,64,power_law_1.01,0.04525440216064453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,64,power_law_1.01,0.04860799789428711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,64,power_law_1.01,0.046975998878479
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,64,power_law_1.01,0.037564160823822024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,64,power_law_1.01,0.05785984039306641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,64,power_law_1.01,0.06704127788543701
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,64,power_law_1.01,0.07689343929290772
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,64,power_law_1.01,0.0403059196472168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,64,power_law_1.01,0.09325823783874512
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,1,64,balanced,0.035979518890380854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,64,power_law_1.01,0.12625151634216308
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,64,power_law_1.01,0.17152896881103516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,1,64,balanced,0.04273536205291748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,64,power_law_1.01,0.2217203140258789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,64,power_law_1.01,0.3467071914672851
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,1,64,power_law_1.01,0.01740031957626343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,64,power_law_1.01,0.45680767059326166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,64,power_law_1.01,0.053917441368103024
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,1,64,power_law_1.01,0.017544959783554075
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,1,64,power_law_1.01,0.017341439723968503
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,1,64,power_law_1.01,0.017331199645996092
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,1,64,power_law_1.01,0.017544959783554075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,64,power_law_1.01,0.6791244506835937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,1,64,power_law_1.01,0.017640960216522217
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,1,64,power_law_1.01,0.01776255965232849
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,64,power_law_1.01,0.7565196990966797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,1,64,power_law_1.01,0.017753599882125853
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,1,64,power_law_1.01,0.018137600421905518
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,1,64,power_law_1.01,0.017827839851379396
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,1,64,power_law_1.01,0.01783679962158203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,1,64,power_law_1.01,0.018141440153121947
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,1,64,power_law_1.01,0.018191360235214234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,1,64,power_law_1.01,0.01935487985610962
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,1,64,power_law_1.01,0.01740031957626343
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,1,64,power_law_1.01,0.017826559543609618
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,4,16,power_law_1.2,0.14325504302978515
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,1,64,power_law_1.01,0.019722239971160887
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,1,64,power_law_1.01,0.02393343925476074
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,1,64,power_law_1.01,0.021451520919799804
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,1,64,power_law_1.01,0.02886528015136719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,64,power_law_1.2,0.030369279384613035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,64,power_law_1.2,0.029726719856262206
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,1,64,power_law_1.01,0.033402879238128666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,64,power_law_1.2,0.029995520114898682
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,64,power_law_1.2,0.05948927879333497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,64,power_law_1.2,0.06900479793548583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,64,power_law_1.2,0.07344128131866455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,64,power_law_1.2,0.06680448055267334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,64,power_law_1.2,0.0891596794128418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,64,power_law_1.2,0.0845299243927002
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,1,64,power_law_1.01,0.058442239761352544
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,64,power_law_1.2,0.08883711814880371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,64,power_law_1.2,0.07999104022979736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,64,power_law_1.2,0.09597311973571777
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,64,power_law_1.2,0.10231040000915528
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,64,power_law_1.2,0.10492032051086427
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,64,power_law_1.2,0.10804736137390136
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,64,power_law_1.2,0.11183232307434081
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,64,power_law_1.2,0.12092927932739257
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,1,64,power_law_1.01,0.04686463832855224
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,64,power_law_1.2,0.1538585567474365
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,64,power_law_1.2,0.17708543777465818
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,64,power_law_1.2,0.23595903396606444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,64,power_law_1.2,0.3281740951538086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,1,64,power_law_1.01,0.08263168334960938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,64,power_law_1.2,0.48829185485839843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,64,power_law_1.2,0.6442790222167969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,64,power_law_1.2,0.030519039630889894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,64,power_law_1.2,0.030950400829315182
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,64,power_law_1.2,0.03783040046691895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,64,power_law_1.2,0.03794559955596924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,64,power_law_1.2,0.039061760902404784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,64,power_law_1.2,0.03956223964691162
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,64,power_law_1.2,0.040761599540710444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,64,power_law_1.2,0.041212158203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,64,power_law_1.2,0.04210303783416748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,64,power_law_1.2,0.041567997932434084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,64,power_law_1.2,0.04340608119964599
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,64,power_law_1.2,0.044650239944458006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,64,power_law_1.2,0.04716032028198242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,64,power_law_1.2,1.0936409759521486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,64,power_law_1.2,0.05207551956176758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,64,power_law_1.2,0.050647039413452145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,64,power_law_1.2,0.055339522361755364
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,64,power_law_1.2,0.05958655834197998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,64,power_law_1.2,1.9908006286621096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,64,power_law_1.2,0.07446656227111817
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,64,power_law_1.2,0.08821248054504395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,64,power_law_1.2,0.11732735633850097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,64,power_law_1.2,0.16780799865722656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,64,power_law_1.2,0.35001216888427733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,64,power_law_1.2,0.2692620849609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,64,power_law_1.2,0.542553596496582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,64,power_law_1.2,1.4067929077148438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,1,64,power_law_1.2,0.017446399927139283
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,1,64,power_law_1.2,0.017465599775314332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,1,64,power_law_1.2,0.017390079498291015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,1,64,power_law_1.2,0.017548799514770508
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,1,64,power_law_1.2,0.01738495945930481
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,1,64,power_law_1.2,0.017624319791793824
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,1,64,power_law_1.2,0.017370879650115967
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,1,64,power_law_1.2,0.017817599773406984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,1,64,power_law_1.2,0.017794560194015503
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,1,64,power_law_1.2,0.018237440586090087
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,1,64,power_law_1.2,0.017893120050430297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,64,power_law_1.2,0.6894771575927734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,1,64,power_law_1.2,0.018254079818725587
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,1,64,power_law_1.01,0.10348799705505371
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,1,64,power_law_1.01,0.149749755859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,64,power_law_1.2,2.830824890136719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,1,64,power_law_1.2,0.019875839948654175
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,1,64,power_law_1.2,0.02288383960723877
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,1,64,power_law_1.01,0.18711040496826173
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,1,64,power_law_1.2,0.019063040018081664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,64,power_law_1.2,1.0573734283447265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,64,power_law_1.2,1.3395762634277344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,1,64,power_law_1.2,0.018191360235214234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,1,64,power_law_1.2,0.019028480052948
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,1,64,power_law_1.2,0.020677120685577394
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,1,64,power_law_1.2,0.028450560569763184
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,1,64,power_law_1.2,0.039521279335021975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,32,balanced,0.026036479473114015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,32,balanced,0.02631808042526245
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,32,balanced,0.02682624101638794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,32,balanced,0.026001920700073244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,32,balanced,0.03533184051513672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,1,64,power_law_1.2,0.04528255939483643
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,32,balanced,0.05612671852111817
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,32,balanced,0.05922688007354736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,32,balanced,0.06408576011657716
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,32,balanced,0.06092415809631348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,32,balanced,0.06702591896057128
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,32,balanced,0.06590208053588867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,32,balanced,0.07736063957214355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,32,balanced,0.06975615978240966
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,32,balanced,0.08458239555358886
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,32,balanced,0.07525248050689697
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,32,balanced,0.09143168449401855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,32,balanced,0.08531583786010742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,32,balanced,0.09470975875854493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,32,balanced,0.09882623672485351
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,32,balanced,0.1120729637145996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,32,balanced,0.12178303718566894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,32,balanced,0.14560256004333497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,32,balanced,0.1707865524291992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,32,balanced,0.22199424743652343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,32,balanced,0.27215871810913084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,32,balanced,0.3998463821411133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,32,balanced,0.027834880352020263
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,32,balanced,0.5117452621459961
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,1,64,power_law_1.2,0.07212031841278076
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,32,balanced,0.028003840446472167
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,32,balanced,0.030467839241027833
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,32,balanced,0.02854144096374512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,32,balanced,0.037080318927764894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,32,balanced,0.03722752094268798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,32,balanced,0.039137279987335204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,32,balanced,0.038748159408569335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,32,balanced,0.040769281387329104
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,32,balanced,0.04122879981994629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,32,balanced,0.03829247951507568
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,32,balanced,0.038696959018707275
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,32,balanced,0.02758143901824951
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,32,balanced,0.04053120136260986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,32,balanced,0.04240255832672119
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,32,balanced,0.04517632007598877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,32,balanced,0.04157311916351318
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,32,balanced,0.0448422384262085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,32,balanced,0.04886528015136719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,32,balanced,0.05525760173797607
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,32,balanced,0.06339456081390381
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,32,balanced,0.08120832443237305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,32,balanced,0.09446528434753418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,32,balanced,0.12552063941955566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,32,balanced,0.15796095848083497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,32,balanced,0.22006912231445314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,32,balanced,0.29523071289062497
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,1,64,power_law_1.2,0.08649344444274902
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,1,64,power_law_1.2,0.12531200408935547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,2,32,balanced,0.017422080039978027
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,1,64,power_law_1.2,0.15196672439575196
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,2,32,balanced,0.017827839851379396
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,2,32,balanced,0.01814911961555481
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,2,32,balanced,0.017401599884033205
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,2,32,balanced,0.01790336012840271
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,2,32,balanced,0.017825280427932737
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,2,32,balanced,0.018216960430145264
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,2,32,balanced,0.018210560083389282
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,2,32,balanced,0.018137600421905518
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,2,32,balanced,0.017472000122070314
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,2,32,balanced,0.01813375949859619
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,2,32,balanced,0.018254079818725587
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,2,32,balanced,0.018406399488449098
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,2,32,balanced,0.018638080358505248
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,2,32,balanced,0.01897984027862549
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,2,32,balanced,0.019048960208892823
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,2,32,balanced,0.01938048005104065
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,2,32,balanced,0.019417599439620972
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,1,64,power_law_1.2,0.20426111221313475
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,2,32,balanced,0.017415679693222046
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,2,32,balanced,0.01824895977973938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,1,64,power_law_1.2,0.2729408073425293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,32,power_law_1.01,0.03331968069076538
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,32,power_law_1.01,0.033075199127197266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,32,power_law_1.01,0.0354150390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,2,32,balanced,0.017820160388946533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,32,power_law_1.01,0.047400960922241206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,32,power_law_1.01,0.05302400112152099
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,2,32,balanced,0.028167679309844974
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,32,power_law_1.01,0.05310848236083985
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,2,32,balanced,0.023128321170806883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,32,power_law_1.01,0.05522431850433349
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,32,power_law_1.01,0.06955391883850097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,32,power_law_1.01,0.07501952171325683
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,2,32,balanced,0.02271104097366333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,32,power_law_1.01,0.07538176059722901
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,32,power_law_1.01,0.08551808357238769
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,32,power_law_1.01,0.0925990390777588
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,32,power_law_1.01,0.09431039810180664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,32,power_law_1.01,0.09362048149108887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,32,power_law_1.01,0.09620479583740235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,32,power_law_1.01,0.10645248413085937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,32,power_law_1.01,0.11739007949829101
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,32,power_law_1.01,0.14628479957580567
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,32,power_law_1.01,0.16945407867431642
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,32,power_law_1.01,0.20993791580200197
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,32,power_law_1.01,0.29877376556396484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,32,power_law_1.01,0.3994905471801758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,32,power_law_1.01,0.07898752212524414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,32,power_law_1.01,0.5436172866821289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,32,power_law_1.01,0.02972800016403198
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,32,power_law_1.01,0.030507519245147708
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,32,power_law_1.01,0.7760281372070312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,32,power_law_1.01,0.03048703908920288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,32,power_law_1.01,0.03513216018676758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,32,power_law_1.01,0.03749504089355469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,32,power_law_1.01,0.03662336111068726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,32,power_law_1.01,1.1121497344970703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,32,power_law_1.01,0.03749504089355469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,32,power_law_1.01,0.04204671859741211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,32,power_law_1.01,0.038388481140136724
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,32,power_law_1.01,0.04036863803863525
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,32,power_law_1.01,0.03951872110366821
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,32,power_law_1.01,0.042780160903930664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,32,power_law_1.01,0.043613438606262205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,32,power_law_1.01,1.4015257263183594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,32,power_law_1.01,0.044803838729858395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,32,power_law_1.01,0.046039037704467774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,32,power_law_1.01,0.04732160091400146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,32,power_law_1.01,0.0546342420578003
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,32,power_law_1.01,0.06196991920471191
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,32,power_law_1.01,0.0719974422454834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,32,power_law_1.01,0.0936575984954834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,32,power_law_1.01,0.11787775993347167
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,32,power_law_1.01,0.15217663764953612
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,32,power_law_1.01,0.21463039398193356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,32,power_law_1.01,0.3328691101074219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,2,32,power_law_1.01,0.017816319465637206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,32,power_law_1.01,0.46719615936279296
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,2,32,power_law_1.01,0.01778048038482666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,2,32,power_law_1.01,0.018202879428863526
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,32,power_law_1.01,0.6253721618652344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,2,32,power_law_1.01,0.017868800163269045
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,2,32,power_law_1.01,0.018145279884338377
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,2,32,power_law_1.01,0.017853440046310426
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,2,32,power_law_1.01,0.01823359966278076
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,2,32,power_law_1.01,0.018079999685287475
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,2,32,power_law_1.01,0.018143999576568603
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,32,power_law_1.01,0.8911949157714844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,2,32,power_law_1.01,0.01836159944534302
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,2,32,power_law_1.01,0.01835647940635681
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,2,32,power_law_1.01,0.018621439933776858
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,2,32,power_law_1.01,0.018650879859924314
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,2,32,power_law_1.01,0.01905151963233948
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,2,32,power_law_1.01,0.01980031967163086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,2,32,power_law_1.01,0.02066943883895874
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,2,32,power_law_1.01,0.02233599901199341
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,2,32,balanced,0.03502975940704346
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,2,32,balanced,0.04504064083099365
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,2,32,balanced,0.05537280082702637
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,2,32,power_law_1.01,0.019055360555648805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,2,32,power_law_1.01,0.04420608043670654
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,2,32,power_law_1.01,0.024416000843048097
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,2,32,power_law_1.01,0.03705343961715698
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,32,power_law_1.2,0.0336243200302124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,32,power_law_1.2,0.034977281093597413
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,32,power_law_1.2,0.04002560138702392
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,32,power_law_1.2,0.05635968208312988
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,32,power_law_1.2,0.06326911926269532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,32,power_law_1.2,0.06985599994659424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,32,power_law_1.2,0.06221439838409424
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,2,32,power_law_1.01,0.03011967897415161
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,32,power_law_1.2,0.07138432025909423
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,32,power_law_1.2,0.07795455932617187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,32,power_law_1.2,0.0847321605682373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,32,power_law_1.2,0.0846003246307373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,32,power_law_1.2,0.08959487915039063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,32,power_law_1.2,0.09464703559875488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,32,power_law_1.2,0.09679360389709472
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,32,power_law_1.2,0.098538236618042
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,32,power_law_1.2,0.102359037399292
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,32,power_law_1.2,0.11049471855163576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,32,power_law_1.2,0.1329305553436279
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,32,power_law_1.2,0.16254720687866211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,32,power_law_1.2,0.2123276710510254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,32,power_law_1.2,0.2537779235839844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,32,power_law_1.2,0.40356094360351563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,32,power_law_1.2,0.5001996612548828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,2,32,power_law_1.01,0.05788032054901123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,32,power_law_1.2,0.7819647979736328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,32,power_law_1.2,0.030261759757995603
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,32,power_law_1.2,0.030522880554199217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,32,power_law_1.2,0.030265600681304933
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,32,power_law_1.2,0.03665535926818848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,32,power_law_1.2,1.4585101318359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,32,power_law_1.2,0.03782399892807007
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,2,32,power_law_1.01,0.07621759891510009
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,32,power_law_1.2,0.04124032020568848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,32,power_law_1.2,0.03911679983139038
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,32,power_law_1.2,0.04024064064025879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,32,power_law_1.2,0.03987328052520752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,32,power_law_1.2,0.04280447959899902
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,32,power_law_1.2,0.044326400756835936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,32,power_law_1.2,0.04370944023132324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,32,power_law_1.2,0.046835198402404785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,32,power_law_1.2,0.04705535888671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,32,power_law_1.2,0.055010561943054204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,32,power_law_1.2,0.03666431903839111
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,32,power_law_1.2,0.0556544017791748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,32,power_law_1.2,0.07146751880645752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,32,power_law_1.2,0.08666239738464356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,32,power_law_1.2,0.11283712387084961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,32,power_law_1.2,0.14175231933593752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,32,power_law_1.2,0.2837401580810547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,32,power_law_1.2,0.22873727798461915
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,2,32,power_law_1.01,0.09401599884033203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,32,power_law_1.2,0.5672550582885743
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,2,32,power_law_1.2,0.01815935969352722
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,2,32,power_law_1.2,0.017812479734420777
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,32,power_law_1.2,0.4351948928833008
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,2,32,power_law_1.2,0.018140159845352173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,32,power_law_1.2,1.0523955535888672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,2,32,power_law_1.2,0.017765120267868043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,2,32,power_law_1.2,0.017811199426651
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,2,32,power_law_1.2,0.01821568012237549
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,2,32,power_law_1.2,0.018165760040283203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,2,32,power_law_1.2,0.018265600204467776
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,2,32,power_law_1.2,0.018211840391159057
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,2,32,power_law_1.2,0.01864192008972168
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,2,32,power_law_1.2,0.01859071969985962
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,2,32,power_law_1.2,0.018644479513168336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,2,32,power_law_1.2,0.018814719915390014
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,2,32,power_law_1.2,0.020666880607604982
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,2,32,power_law_1.2,0.017820160388946533
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,2,32,power_law_1.01,0.1398655986785889
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,2,32,power_law_1.01,0.16626432418823242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,32,power_law_1.2,0.845486068725586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,2,32,power_law_1.2,0.01988736033439636
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,2,32,power_law_1.2,0.021502718925476075
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,2,32,power_law_1.2,0.02424959897994995
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,2,32,power_law_1.2,0.025523200035095218
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,2,32,power_law_1.2,0.032971520423889164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,32,power_law_1.2,1.8986560058593749
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,2,32,power_law_1.2,0.04451839923858643
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,16,balanced,0.025615360736846927
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,16,balanced,0.025589759349823
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,16,balanced,0.024360959529876706
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,16,balanced,0.026334719657897947
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,16,balanced,0.03132672071456909
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,16,balanced,0.056655359268188474
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,16,balanced,0.057448959350585936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,16,balanced,0.0621017599105835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,16,balanced,0.06260863780975343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,16,balanced,0.06278272151947022
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,16,balanced,0.07441152095794677
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,16,balanced,0.08267264366149903
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,16,balanced,0.08678784370422363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,16,balanced,0.08422911643981934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,16,balanced,0.08616448402404785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,16,balanced,0.08814335823059082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,16,balanced,0.09505023956298828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,16,balanced,0.1053273582458496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,16,balanced,0.11254528045654297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,16,balanced,0.1362598419189453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,16,balanced,0.15424768447875975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,16,balanced,0.19691648483276367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,16,balanced,0.23922943115234374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,16,balanced,0.32191871643066405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,16,balanced,0.40456192016601566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,16,balanced,0.5975667190551758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,16,balanced,0.7781094360351563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,16,balanced,0.02930304050445557
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,16,balanced,0.029034240245819094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,16,balanced,0.028115200996398925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,16,balanced,0.02934528112411499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,16,balanced,0.030531840324401854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,16,balanced,0.03704319953918457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,16,balanced,0.03870208024978637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,16,balanced,0.040252161026000974
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,16,balanced,0.039508481025695805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,16,balanced,0.040316162109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,16,balanced,0.041565442085266115
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,16,balanced,0.044418559074401856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,16,balanced,0.0464844799041748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,16,balanced,0.0428659200668335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,16,balanced,0.0445798397064209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,16,balanced,0.04551551818847656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,16,balanced,0.04892799854278564
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,2,32,power_law_1.2,0.08260479927062989
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,16,balanced,0.051430401802062986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,16,balanced,0.05963391780853271
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,16,balanced,0.07639039993286133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,32,power_law_1.2,1.0800563049316405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,16,balanced,0.08764543533325195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,16,balanced,0.11894016265869141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,16,balanced,0.14285568237304686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,16,balanced,0.19838336944580076
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,16,balanced,0.2551027107238769
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,16,balanced,0.36243457794189454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,16,balanced,0.48138496398925784
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,2,32,power_law_1.2,0.09419903755187989
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,2,32,power_law_1.2,0.06386688232421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,2,32,power_law_1.2,0.12931967735290528
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,2,32,power_law_1.2,0.20545280456542966
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,4,16,balanced,0.018222080469131468
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,4,16,balanced,0.018222080469131468
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,4,16,balanced,0.01780992031097412
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,4,16,balanced,0.02035840034484863
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,4,16,balanced,0.0202675199508667
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,4,16,balanced,0.019379199743270875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,4,16,balanced,0.02064768075942993
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,4,16,balanced,0.020784640312194826
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,4,16,balanced,0.021105918884277344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,4,16,balanced,0.020673279762268064
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,4,16,balanced,0.021071360111236573
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,4,16,balanced,0.02047231912612915
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,4,16,balanced,0.021057279109954835
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,4,16,balanced,0.020847361087799072
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,4,16,balanced,0.02110975980758667
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,4,16,balanced,0.022726399898529055
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,4,16,balanced,0.022731521129608155
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,4,16,balanced,0.020285439491271973
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,4,16,balanced,0.020682239532470705
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,4,16,balanced,0.020670719146728515
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,2,32,power_law_1.2,0.2462067222595215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,16,power_law_1.01,0.028456959724426266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,16,power_law_1.01,0.029256958961486817
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,16,power_law_1.01,0.035783679485321046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,16,power_law_1.01,0.03460736036300659
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,16,power_law_1.01,0.05934591770172119
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,16,power_law_1.01,0.06490623950958252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,16,power_law_1.01,0.06754432201385499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,16,power_law_1.01,0.06641407966613769
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,16,power_law_1.01,0.07497087955474854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,16,power_law_1.01,0.08034815788269042
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,16,power_law_1.01,0.08194944381713867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,16,power_law_1.01,0.08895232200622558
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,16,power_law_1.01,0.08894847869873047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,16,power_law_1.01,0.09094400405883789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,16,power_law_1.01,0.0944320011138916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,16,power_law_1.01,0.09837568283081055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,16,power_law_1.01,0.10441984176635741
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,16,power_law_1.01,0.12062463760375977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,16,power_law_1.01,0.13587072372436523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,16,power_law_1.01,0.1681715202331543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,16,power_law_1.01,0.2205708885192871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,16,power_law_1.01,0.26844287872314454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,16,power_law_1.01,0.3229401779174805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,16,power_law_1.01,0.5488332748413086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,16,power_law_1.01,0.666270751953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,4,16,balanced,0.020568320751190187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,16,power_law_1.01,1.0296959686279297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,16,power_law_1.01,0.02886784076690674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,16,power_law_1.01,0.029268479347229003
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,16,power_law_1.01,0.02968064069747925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,16,power_law_1.01,0.030254080295562747
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,16,power_law_1.01,1.4022221374511719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,16,power_law_1.01,0.03581824064254761
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,16,power_law_1.01,0.037487359046936036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,16,power_law_1.01,0.038691840171813964
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,16,power_law_1.01,0.03948928117752075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,16,power_law_1.01,0.0403391981124878
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,16,power_law_1.01,0.042328319549560546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,16,power_law_1.01,0.04445184230804443
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,4,16,balanced,0.048887038230896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,16,power_law_1.01,0.04559103965759277
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,16,power_law_1.01,0.04501120090484619
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,4,16,balanced,0.03909631967544556
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,16,power_law_1.01,0.04564735889434814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,16,power_law_1.01,0.04811647891998291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,16,power_law_1.01,0.04989312171936035
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,4,16,balanced,0.02829952001571655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,16,power_law_1.01,0.05452032089233398
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,4,16,balanced,0.030557439327239987
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,16,power_law_1.01,0.06693376064300537
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,16,power_law_1.01,0.07542143821716309
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,16,power_law_1.01,0.08811903953552246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,16,power_law_1.01,0.11871871948242188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,16,power_law_1.01,0.15261311531066896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,16,power_law_1.01,0.21951616287231443
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,16,power_law_1.01,0.28322687149047854
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,4,16,power_law_1.01,0.0202726411819458
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,16,power_law_1.01,0.3908492660522461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,4,16,balanced,0.06474239826202392
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,4,16,power_law_1.01,0.020302081108093263
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,4,16,power_law_1.01,0.020288000106811525
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,4,16,balanced,0.07882239818572997
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,4,16,power_law_1.01,0.019833600521087645
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,4,16,power_law_1.01,0.020316159725189208
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,4,16,power_law_1.01,0.020599040985107422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,4,16,power_law_1.01,0.020673279762268064
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,4,16,power_law_1.01,0.02054527997970581
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,4,16,power_law_1.01,0.020657920837402345
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,4,16,power_law_1.01,0.02061311960220337
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,16,power_law_1.01,0.8342668914794922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,4,16,power_law_1.01,0.020709118843078612
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,4,16,power_law_1.01,0.021107199192047118
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,4,16,power_law_1.01,0.02202239990234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,4,16,power_law_1.01,0.020198400020599368
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,16,power_law_1.01,0.5950771331787109
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,4,16,power_law_1.01,0.020646400451660156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,4,16,power_law_1.01,0.021103360652923585
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,4,16,power_law_1.01,0.022753279209136962
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,4,16,power_law_1.01,0.026152958869934084
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,4,16,power_law_1.01,0.034013440608978276
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,4,16,power_law_1.01,0.028343040943145752
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,4,16,power_law_1.01,0.04614272117614746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,16,power_law_1.2,0.032116479873657226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,16,power_law_1.2,0.03540735960006714
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,16,power_law_1.2,0.037327361106872556
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,16,power_law_1.2,0.03996416091918945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,16,power_law_1.2,0.05529727935791016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,16,power_law_1.2,0.06817920207977295
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,4,16,power_law_1.01,0.06811264038085937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,16,power_law_1.2,0.06956416130065918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,4,16,power_law_1.01,0.05172607898712158
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,16,power_law_1.2,0.07334784030914307
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,16,power_law_1.2,0.07852928161621094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,16,power_law_1.2,0.08081664085388184
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,16,power_law_1.2,0.09055359840393067
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,16,power_law_1.2,0.0895910358428955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,16,power_law_1.2,0.08989439964294434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,16,power_law_1.2,0.09112064361572265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,16,power_law_1.2,0.09560064315795899
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,16,power_law_1.2,0.09920639991760254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,16,power_law_1.2,0.1076313591003418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,16,power_law_1.2,0.12365056037902833
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,4,16,power_law_1.01,0.08299391746520997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,16,power_law_1.2,0.1463807964324951
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,16,power_law_1.2,0.1836172866821289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,16,power_law_1.2,0.22336896896362304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,16,power_law_1.2,0.3190105628967285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,16,power_law_1.2,0.3971878433227539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,16,power_law_1.2,0.029646079540252685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,16,power_law_1.2,0.029410560131073
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,16,power_law_1.2,0.03001215934753418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,16,power_law_1.2,0.030127360820770267
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,16,power_law_1.2,0.036547839641571045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,16,power_law_1.2,0.037681920528411864
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,16,power_law_1.2,0.03870336055755615
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,16,power_law_1.2,1.19797119140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,16,power_law_1.2,0.03999232053756714
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,16,power_law_1.2,0.040759038925170896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,16,power_law_1.2,0.042351360321044924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,16,power_law_1.2,0.04646783828735351
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,16,power_law_1.2,0.04712831974029541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,16,power_law_1.2,0.04405248165130615
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,16,power_law_1.2,0.6348723220825195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,16,power_law_1.2,0.04936063766479492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,16,power_law_1.2,0.04731391906738282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,16,power_law_1.2,0.05299456119537353
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,16,power_law_1.2,0.0569817590713501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,16,power_law_1.2,0.0680191993713379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,16,power_law_1.2,0.08328191757202148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,16,power_law_1.2,0.11912960052490233
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,16,power_law_1.2,0.12894975662231445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,16,power_law_1.2,0.24232063293457032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,16,power_law_1.2,0.20131200790405274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,16,power_law_1.2,0.8363302612304688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,16,power_law_1.2,0.3846464157104492
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,4,16,power_law_1.2,0.02024960041046143
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,4,16,power_law_1.2,0.020229120254516602
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,4,16,power_law_1.2,0.020414719581604006
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,4,16,power_law_1.2,0.02024319887161255
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,4,16,power_law_1.2,0.019980800151824952
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,4,16,power_law_1.2,0.02029952049255371
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,4,16,power_law_1.2,0.020302081108093263
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,4,16,power_law_1.2,0.020655360221862793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,4,16,power_law_1.2,0.020632319450378418
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,4,16,power_law_1.2,0.02063999891281128
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,4,16,power_law_1.2,0.02064768075942993
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,4,16,power_law_1.2,0.020582399368286132
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,4,16,power_law_1.2,0.020661759376525878
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,4,16,power_law_1.2,0.021079039573669432
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,4,16,power_law_1.2,0.021923840045928955
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,4,16,power_law_1.2,0.022691841125488284
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,16,power_law_1.2,0.5039577484130859
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,4,16,power_law_1.01,0.12099200248718263
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,4,16,power_law_1.01,0.21108736038208006
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,4,16,power_law_1.01,0.15679871559143066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,16,power_law_1.2,0.7483238220214844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,16,power_law_1.2,1.6235891723632812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,4,16,power_law_1.2,0.024323840141296387
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,4,16,power_law_1.2,0.0286080002784729
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,4,16,power_law_1.2,0.032884480953216555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,4,16,power_law_1.2,0.037194240093231204
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,4,16,power_law_1.2,0.05391232013702393
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,16,power_law_1.2,0.952907485961914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,8,8,balanced,0.021907200813293455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,8,8,balanced,0.02436608076095581
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,8,8,balanced,0.02556544065475464
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,8,8,balanced,0.02885504007339478
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,8,8,balanced,0.03836031913757324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,8,8,balanced,0.06448383808135986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,8,8,balanced,0.066048002243042
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,8,8,balanced,0.07191167831420898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,8,8,balanced,0.08301055908203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,8,8,balanced,0.08025600433349608
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,8,8,balanced,0.08115839958190918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,8,8,balanced,0.08658944129943848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,8,8,balanced,0.08915583610534668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,8,8,balanced,0.09442815780639649
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,8,8,balanced,0.09912320137023925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,8,8,balanced,0.10271743774414062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,4,16,power_law_1.2,0.07589503765106201
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,8,8,balanced,0.10922752380371095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,8,8,balanced,0.13028608322143556
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,8,8,balanced,0.1488102436065674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,8,8,balanced,0.18882303237915038
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,8,8,balanced,0.22546304702758788
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,8,8,balanced,0.30244863510131836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,8,8,balanced,0.37938175201416013
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,8,8,balanced,0.5289011383056641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,8,8,balanced,0.6798886108398438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,8,8,balanced,0.02723072052001953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,8,8,balanced,0.027238399982452394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,8,8,balanced,1.0167142486572265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,8,8,balanced,0.029315838813781737
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,8,8,balanced,0.031025919914245602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,8,8,balanced,0.03663872003555298
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,8,8,balanced,0.04398848056793213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,8,8,balanced,0.045480961799621585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,8,8,balanced,1.3257331848144531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,8,8,balanced,0.04733695983886719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,8,8,balanced,0.04895232200622558
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,8,8,balanced,0.0469542407989502
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,8,8,balanced,0.04975615978240967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,8,8,balanced,0.049455361366271974
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,8,8,balanced,0.050579199790954585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,8,8,balanced,0.05400576114654541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,8,8,balanced,0.054261760711669924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,8,8,balanced,0.05642623901367187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,8,8,balanced,0.06278272151947022
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,8,8,balanced,0.07663360118865967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,8,8,balanced,0.08791040420532227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,8,8,balanced,0.11530752182006836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,8,8,balanced,0.138275842666626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,8,8,balanced,0.19610111236572264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,8,8,balanced,0.2419340705871582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,8,8,balanced,0.34605567932128906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,8,8,balanced,0.44837886810302735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,8,8,balanced,0.6622054290771484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,8,8,balanced,0.8633869171142579
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,4,16,power_law_1.2,0.08001407623291015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,4,16,power_law_1.2,0.12030591964721679
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,4,16,power_law_1.2,0.15614975929260252
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,8,8,balanced,0.018492159843444825
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,8,8,balanced,0.023889920711517333
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,8,8,balanced,0.023905279636383055
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,8,8,balanced,0.019028480052948
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,8,8,balanced,0.02356352090835571
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,8,8,balanced,0.0235916805267334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,8,8,balanced,0.023897600173950196
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,8,8,balanced,0.023935999870300293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,8,8,balanced,0.023892478942871092
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,4,16,power_law_1.2,0.22732416152954102
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,8,8,balanced,0.023576319217681885
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,8,8,balanced,0.02391040086746216
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,8,8,balanced,0.024097280502319334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,8,8,balanced,0.025603840351104735
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,8,8,balanced,0.020016639232635497
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,8,8,balanced,0.025655040740966795
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,8,8,balanced,0.02800512075424194
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,8,8,balanced,0.0294323205947876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,8,8,balanced,0.02143359899520874
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,8,8,balanced,0.023942399024963378
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,4,16,power_law_1.2,0.29346815109252933
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,8,8,balanced,0.02348543882369995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,8,8,power_law_1.01,0.027601919174194335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,8,8,power_law_1.01,0.03109503984451294
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,8,8,power_law_1.01,0.032623360157012934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,8,8,power_law_1.01,0.04116352081298828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,8,8,power_law_1.01,0.043009281158447266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,8,8,power_law_1.01,0.06123519897460937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,8,8,power_law_1.01,0.06503551959991455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,8,8,power_law_1.01,0.07392640113830566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,8,8,power_law_1.01,0.07573760032653809
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,8,8,power_law_1.01,0.07964288234710694
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,8,8,power_law_1.01,0.08775296211242675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,8,8,power_law_1.01,0.08700799942016602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,8,8,power_law_1.01,0.08701567649841309
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,8,8,power_law_1.01,0.09208831787109376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,8,8,power_law_1.01,0.09496959686279297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,8,8,power_law_1.01,0.09921536445617676
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,8,8,balanced,0.02395008087158203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,8,8,power_law_1.01,0.10718591690063477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,8,8,power_law_1.01,0.1253606414794922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,8,8,balanced,0.04104320049285889
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,8,8,power_law_1.01,0.14469759941101074
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,8,8,balanced,0.04435711860656738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,8,8,power_law_1.01,0.18130815505981446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,8,8,power_law_1.01,0.2187571144104004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,8,8,power_law_1.01,0.29550079345703123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,8,8,power_law_1.01,0.37606143951416016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,8,8,balanced,0.05894144058227539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,8,8,power_law_1.01,0.5678374481201172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,8,8,power_law_1.01,0.029605119228363036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,8,8,power_law_1.01,0.7080025482177734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,8,8,power_law_1.01,0.032572159767150874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,8,8,power_law_1.01,0.030947840213775633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,8,8,power_law_1.01,1.066487045288086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,8,8,power_law_1.01,0.035057919025421144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,8,8,power_law_1.01,0.03538687944412232
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,8,8,power_law_1.01,0.04571392059326172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,8,8,power_law_1.01,0.042414078712463374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,8,8,power_law_1.01,1.3174578857421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,8,8,power_law_1.01,0.04496895790100098
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,8,8,power_law_1.01,0.04688384056091309
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,8,8,power_law_1.01,0.04811391830444336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,8,8,power_law_1.01,0.050032639503479005
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,8,8,power_law_1.01,0.04929408073425293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,8,8,power_law_1.01,0.0530291223526001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,8,8,power_law_1.01,0.05551616191864014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,8,8,power_law_1.01,0.05384575843811035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,8,8,power_law_1.01,0.06124032020568848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,8,8,power_law_1.01,0.07248000144958497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,8,8,power_law_1.01,0.08582783699035644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,8,8,power_law_1.01,0.11363967895507812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,8,8,power_law_1.01,0.13714176177978515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,8,8,power_law_1.01,0.17478784561157226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,8,8,power_law_1.01,0.23268735885620115
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,8,8,power_law_1.01,0.316495361328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,8,8,power_law_1.01,0.4434534454345703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,8,8,power_law_1.01,0.051060481071472166
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,8,8,power_law_1.01,0.020689918994903564
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,8,8,power_law_1.01,0.021921279430389403
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,8,8,power_law_1.01,0.021907200813293455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,8,8,power_law_1.01,0.6557170867919921
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,8,8,power_law_1.01,0.022657279968261716
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,8,8,power_law_1.01,0.02353024005889893
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,8,8,power_law_1.01,0.023292160034179686
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,8,8,power_law_1.01,0.02348288059234619
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,8,8,power_law_1.01,0.023470079898834227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,8,8,power_law_1.01,0.8376230621337891
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,8,8,power_law_1.01,0.023573760986328123
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,8,8,power_law_1.01,0.0235532808303833
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,8,8,power_law_1.01,0.023760640621185304
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,8,8,power_law_1.01,0.02411263942718506
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,8,8,power_law_1.01,0.023975679874420168
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,8,8,power_law_1.01,0.024780800342559813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,8,8,power_law_1.01,0.024944639205932616
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,8,8,power_law_1.01,0.026494719982147218
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,8,8,balanced,0.07189504146575929
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,8,8,balanced,0.09402239799499512
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,8,8,balanced,0.11824383735656738
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,8,8,power_law_1.01,0.03319040060043335
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,8,8,power_law_1.01,0.058767361640930174
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,8,8,power_law_1.01,0.038759679794311525
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,8,8,power_law_1.01,0.029694719314575197
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,8,8,power_law_1.01,0.050961918830871575
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,8,8,power_law_1.2,0.027729918956756593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,8,8,power_law_1.2,0.026451199054718016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,8,8,power_law_1.2,0.0328985595703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,8,8,power_law_1.2,0.03908864021301269
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,8,8,power_law_1.2,0.04550911903381348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,8,8,power_law_1.2,0.0678937578201294
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,8,8,power_law_1.01,0.07941120147705079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,8,8,power_law_1.2,0.0627788782119751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,8,8,power_law_1.2,0.07483391761779785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,8,8,power_law_1.2,0.07666944026947022
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,8,8,power_law_1.2,0.08356863975524902
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,8,8,power_law_1.2,0.09082624435424805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,8,8,power_law_1.2,0.08609536170959473
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,8,8,power_law_1.2,0.09132927894592285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,8,8,power_law_1.2,0.09256447792053223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,8,8,power_law_1.2,0.09693440437316894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,8,8,power_law_1.2,0.10297856330871583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,8,8,power_law_1.2,0.11167872428894043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,8,8,power_law_1.2,0.13206144332885744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,8,8,power_law_1.2,0.14878463745117188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,8,8,power_law_1.2,0.18705408096313475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,8,8,power_law_1.2,0.22647167205810548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,8,8,power_law_1.2,0.30565120697021486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,8,8,power_law_1.2,0.4017766571044922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,8,8,power_law_1.2,0.5479436874389648
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,8,8,power_law_1.01,0.0981606388092041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,8,8,power_law_1.2,0.02890239953994751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,8,8,power_law_1.2,0.03050112009048462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,8,8,power_law_1.2,0.033413119316101074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,8,8,power_law_1.2,0.03457407951354981
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,8,8,power_law_1.2,0.03500672101974488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,8,8,power_law_1.2,0.04318975925445557
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,8,8,power_law_1.2,0.043681278228759765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,8,8,power_law_1.2,0.045239038467407226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,8,8,power_law_1.2,0.046449918746948246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,8,8,power_law_1.2,1.1116659545898437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,8,8,power_law_1.2,0.049354238510131834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,8,8,power_law_1.2,0.05140351772308349
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,8,8,power_law_1.2,0.05020800113677979
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,8,8,power_law_1.2,0.04974847793579101
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,8,8,power_law_1.01,0.1276863956451416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,8,8,power_law_1.2,0.052634878158569334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,8,8,power_law_1.2,0.05594367980957031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,8,8,power_law_1.2,0.7179942321777344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,8,8,power_law_1.2,0.058745598793029784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,8,8,power_law_1.2,0.06369919776916504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,8,8,power_law_1.2,0.07906943798065186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,8,8,power_law_1.2,0.09020544052124023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,8,8,power_law_1.2,0.11943039894104004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,8,8,power_law_1.2,0.147390718460083
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,8,8,power_law_1.2,0.2545126342773437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,8,8,power_law_1.2,0.2016383934020996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,8,8,power_law_1.2,0.37391487121582034
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,8,8,power_law_1.01,0.19377920150756836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,8,8,power_law_1.2,0.02099967956542969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,8,8,power_law_1.2,0.02046720027923584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,8,8,power_law_1.2,0.022302720546722412
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,8,8,power_law_1.2,0.023192319869995117
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,8,8,power_law_1.2,0.023116800785064697
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,8,8,power_law_1.01,0.2644607925415039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,8,8,power_law_1.2,0.02312448024749756
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,8,8,power_law_1.01,0.324983024597168
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,8,8,power_law_1.2,0.023541760444641114
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,8,8,power_law_1.2,0.0235699200630188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,8,8,power_law_1.2,0.023539199829101562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,8,8,power_law_1.2,0.024577279090881348
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,8,8,power_law_1.2,0.023907840251922607
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,8,8,power_law_1.2,0.02433151960372925
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,8,8,power_law_1.2,0.024753921031951905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,8,8,power_law_1.2,0.4911692810058594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,8,8,power_law_1.2,0.02358016014099121
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,8,8,power_law_1.2,0.02721407890319824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,8,8,power_law_1.2,0.7254118347167969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,8,8,power_law_1.2,1.4776089477539063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,8,8,power_law_1.2,0.027229440212249757
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,8,8,power_law_1.2,0.031709439754486084
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,8,8,power_law_1.2,0.04426496028900147
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,8,8,power_law_1.2,0.04113408088684082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,8,8,power_law_1.2,0.061249279975891115
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,8,8,power_law_1.2,0.07115392208099365
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,16,4,balanced,0.02402048110961914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,16,4,balanced,0.02634880065917969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,16,4,balanced,0.027656960487365722
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,16,4,balanced,0.03133440017700195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,16,4,balanced,0.041520638465881346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,8,8,power_law_1.2,0.9676774597167969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,16,4,balanced,0.04762495994567871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,16,4,balanced,0.06605184078216553
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,16,4,balanced,0.08014080047607422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,16,4,balanced,0.08317055702209472
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,16,4,balanced,0.08701184272766113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,16,4,balanced,0.08716032028198242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,16,4,balanced,0.08823040008544922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,16,4,balanced,0.09362688064575195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,16,4,balanced,0.09706496238708497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,16,4,balanced,0.10026495933532716
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,16,4,balanced,0.10479743957519531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,16,4,balanced,0.11462016105651855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,16,4,balanced,0.1358041572570801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,16,4,balanced,0.15369215965270996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,16,4,balanced,0.19081855773925782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,16,4,balanced,0.22779264450073242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,16,4,balanced,0.30259328842163086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,16,4,balanced,0.3762777709960937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,16,4,balanced,0.5234239959716798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,16,4,balanced,0.6686246490478516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,16,4,balanced,0.988559341430664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,16,4,balanced,0.027630081176757814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,16,4,balanced,0.028857600688934327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,16,4,balanced,0.03323008060455322
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,16,4,balanced,1.3059648132324218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,16,4,balanced,0.037013759613037114
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,16,4,balanced,0.04576384067535401
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,16,4,balanced,0.047988481521606445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,16,4,balanced,0.05379583835601807
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,16,4,balanced,0.05483263969421387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,16,4,balanced,0.056733441352844236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,16,4,balanced,0.05834112167358398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,16,4,balanced,0.05384448051452637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,16,4,balanced,0.05557631969451904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,16,4,balanced,0.0570854377746582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,16,4,balanced,0.05958015918731689
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,16,4,balanced,0.05934847831726074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,16,4,balanced,0.06279168128967286
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,16,4,balanced,0.06818175792694092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,16,4,balanced,0.08331392288208009
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,16,4,balanced,0.09493632316589355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,16,4,balanced,0.11898240089416504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,16,4,balanced,0.1400089645385742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,16,4,balanced,0.20327039718627932
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,16,4,balanced,0.2452441596984863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,16,4,balanced,0.34958591461181643
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,16,4,balanced,0.453697280883789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,16,4,balanced,0.6705779266357422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,16,4,balanced,0.877413101196289
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,8,8,power_law_1.2,0.09407360076904296
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,8,8,power_law_1.2,0.136814079284668
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,8,8,power_law_1.2,0.1906585693359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,8,8,power_law_1.2,0.2420198440551758
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,8,8,power_law_1.2,0.3369318389892578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,16,4,balanced,0.022304000854492186
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,16,4,balanced,0.031783680915832516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,16,4,balanced,0.03224191904067993
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,16,4,balanced,0.03229952096939087
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,16,4,balanced,0.03344255924224854
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,16,4,balanced,0.03365247964859009
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,16,4,balanced,0.03359231948852539
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,16,4,balanced,0.033384959697723385
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,16,4,balanced,0.032958719730377194
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,16,4,balanced,0.033925120830535886
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,16,4,balanced,0.022649600505828857
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,16,4,balanced,0.035411200523376464
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,16,4,balanced,0.03918976068496704
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,16,4,balanced,0.04050687789916992
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,16,4,balanced,0.04600831985473633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,16,4,balanced,0.048911361694335936
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,16,4,balanced,0.02389888048171997
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,8,8,power_law_1.2,0.4526591873168945
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,16,4,balanced,0.025607678890228268
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,16,4,balanced,0.03214463949203491
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,16,4,balanced,0.031357440948486325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,16,4,power_law_1.01,0.028948481082916262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,16,4,power_law_1.01,0.02844288110733032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,16,4,power_law_1.01,0.03363584041595459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,16,4,power_law_1.01,0.041565442085266115
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,16,4,power_law_1.01,0.04225152015686035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,16,4,power_law_1.01,0.04624000072479248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,16,4,power_law_1.01,0.07274112224578858
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,16,4,power_law_1.01,0.0774515199661255
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,16,4,power_law_1.01,0.08215167999267578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,16,4,power_law_1.01,0.08599807739257812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,16,4,power_law_1.01,0.08893183708190919
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,16,4,power_law_1.01,0.08724608421325683
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,16,4,power_law_1.01,0.0900812816619873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,16,4,power_law_1.01,0.09665408134460449
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,16,4,balanced,0.07196800231933594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,16,4,power_law_1.01,0.10076800346374512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,16,4,power_law_1.01,0.10666496276855468
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,16,4,balanced,0.07802239894866944
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,16,4,power_law_1.01,0.13414784431457522
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,16,4,power_law_1.01,0.14973952293395995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,16,4,power_law_1.01,0.19299072265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,16,4,power_law_1.01,0.22630655288696291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,16,4,power_law_1.01,0.30009471893310546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,16,4,power_law_1.01,0.3748223876953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,16,4,power_law_1.01,0.11475968360900879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,16,4,power_law_1.01,0.5395686340332031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,16,4,balanced,0.10414463996887206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,16,4,power_law_1.01,0.02922879934310913
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,16,4,power_law_1.01,0.7130290985107421
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,16,4,power_law_1.01,0.032967679500579834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,16,4,power_law_1.01,0.03500159978866577
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,16,4,power_law_1.01,1.017364501953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,16,4,power_law_1.01,0.03845247983932495
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,16,4,power_law_1.01,0.040783357620239255
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,16,4,power_law_1.01,0.04279551982879638
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,16,4,power_law_1.01,0.053038082122802734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,16,4,power_law_1.01,0.054275841712951654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,16,4,power_law_1.01,0.055589118003845216
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,16,4,power_law_1.01,1.3071743774414064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,16,4,power_law_1.01,0.05797120094299316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,16,4,power_law_1.01,0.05918591976165771
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,16,4,power_law_1.01,0.055101442337036136
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,16,4,power_law_1.01,0.05633152008056641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,16,4,power_law_1.01,0.058813438415527344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,16,4,power_law_1.01,0.06236671924591064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,16,4,power_law_1.01,0.07013504028320312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,16,4,power_law_1.01,0.08518272399902344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,16,4,power_law_1.01,0.09605759620666504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,16,4,power_law_1.01,0.12617088317871095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,16,4,power_law_1.01,0.1534169578552246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,16,4,power_law_1.01,0.2101888084411621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,16,4,power_law_1.01,0.26545663833618166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,16,4,power_law_1.01,0.3650611114501953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,16,4,power_law_1.01,0.062452478408813475
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,16,4,balanced,0.13181183815002442
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,16,4,power_law_1.01,0.4828006362915039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,16,4,power_law_1.01,0.023948800563812257
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,16,4,power_law_1.01,0.025996799468994143
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,16,4,power_law_1.01,0.7202150726318359
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,16,4,power_law_1.01,0.027485439777374266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,16,4,power_law_1.01,0.02809727907180786
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,16,4,power_law_1.01,0.03132800102233887
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,16,4,power_law_1.01,0.031737599372863766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,16,4,balanced,0.18848384857177733
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,16,4,power_law_1.01,0.030919680595397948
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,16,4,power_law_1.01,0.032565760612487796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,16,4,power_law_1.01,0.9584754943847656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,16,4,power_law_1.01,0.032206079959869384
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,16,4,power_law_1.01,0.034161920547485354
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,16,4,power_law_1.01,0.03338624000549316
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,16,4,power_law_1.01,0.035403521060943605
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,16,4,power_law_1.01,0.034600958824157715
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,16,4,power_law_1.01,0.037518079280853274
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,16,4,balanced,0.24168832778930663
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,16,4,power_law_1.01,0.02488703966140747
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,16,4,power_law_1.01,0.04155903816223144
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,16,4,power_law_1.01,0.06655488014221192
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,16,4,power_law_1.01,0.05891456127166748
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,16,4,power_law_1.01,0.08798463821411133
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,16,4,power_law_1.01,0.09927040100097656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,16,4,power_law_1.01,0.13865471839904786
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,16,4,power_law_1.2,0.026776320934295654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,16,4,power_law_1.2,0.02758143901824951
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,16,4,power_law_1.2,0.031427841186523434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,16,4,power_law_1.2,0.04120063781738281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,16,4,power_law_1.2,0.04419072151184082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,16,4,power_law_1.01,0.17038080215454102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,16,4,power_law_1.2,0.07646207809448242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,16,4,power_law_1.2,0.08123392105102539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,16,4,power_law_1.2,0.08593024253845215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,16,4,power_law_1.2,0.08882559776306152
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,16,4,power_law_1.2,0.08940544128417968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,16,4,power_law_1.2,0.04928127765655518
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,16,4,power_law_1.2,0.08937600135803223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,16,4,power_law_1.2,0.09307007789611817
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,16,4,power_law_1.2,0.09772031784057618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,16,4,power_law_1.2,0.1010700798034668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,16,4,power_law_1.2,0.1059494400024414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,16,4,power_law_1.2,0.11568767547607421
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,16,4,power_law_1.01,0.5639769744873047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,16,4,power_law_1.2,0.13532416343688963
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,16,4,power_law_1.2,0.15427840232849122
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,16,4,power_law_1.2,0.19353599548339845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,16,4,power_law_1.2,0.30810239791870114
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,16,4,power_law_1.01,0.4506470489501953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,16,4,power_law_1.2,0.2294758415222168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,16,4,power_law_1.2,0.3804288101196289
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,16,4,power_law_1.01,0.32229248046875003
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,16,4,power_law_1.2,0.5575539016723633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,16,4,power_law_1.2,0.02891520023345947
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,16,4,power_law_1.2,0.03176320075988769
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,16,4,power_law_1.2,0.03536384105682373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,16,4,power_law_1.2,0.037085440158844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,16,4,power_law_1.2,0.04076288223266601
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,16,4,power_law_1.2,0.04362751960754395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,16,4,power_law_1.2,0.05304192066192627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,16,4,power_law_1.2,0.05463551998138427
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,16,4,power_law_1.2,0.05513472080230712
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,16,4,power_law_1.2,0.057093119621276854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,16,4,power_law_1.2,0.059448318481445314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,16,4,power_law_1.2,0.055530238151550296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,16,4,power_law_1.2,1.0368460845947265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,16,4,power_law_1.2,0.05756288051605225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,16,4,power_law_1.2,0.060208640098571777
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,16,4,power_law_1.2,0.06030720233917236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,16,4,power_law_1.2,0.06580992221832276
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,16,4,power_law_1.2,0.07216256141662598
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,16,4,power_law_1.2,0.08804224014282228
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,16,4,power_law_1.2,0.10016256332397462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,16,4,power_law_1.2,0.1310540771484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,16,4,power_law_1.2,0.15720447540283203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,16,4,power_law_1.2,0.26884223937988283
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,16,4,power_law_1.2,0.21300224304199217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,16,4,power_law_1.2,0.7164710235595704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,16,4,power_law_1.2,0.3886080169677734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,16,4,power_law_1.2,0.023941121101379394
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,16,4,power_law_1.2,0.02429824113845825
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,16,4,power_law_1.2,0.025410559177398685
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,16,4,power_law_1.2,0.02797823905944824
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,16,4,power_law_1.2,0.028610560894012448
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,16,4,power_law_1.2,0.03175168037414551
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,16,4,power_law_1.2,0.03248127937316895
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,16,4,power_law_1.2,0.03250688076019287
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,16,4,power_law_1.2,0.032624640464782716
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,16,4,power_law_1.2,0.03302783966064453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,16,4,power_law_1.2,0.03406847953796387
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,16,4,power_law_1.2,0.03525248050689697
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,16,4,power_law_1.2,0.03473407983779907
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,16,4,power_law_1.2,0.039445760250091555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,16,4,power_law_1.2,0.041788158416748045
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,16,4,power_law_1.2,0.04301439762115479
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,16,4,power_law_1.2,0.49557121276855465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,16,4,power_law_1.2,0.7410598754882812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,16,4,power_law_1.2,0.04944128036499024
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,16,4,power_law_1.2,0.06475647926330566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,16,4,power_law_1.2,1.3785574340820312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,16,4,power_law_1.2,0.9508748626708984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,16,4,power_law_1.2,0.07878784179687501
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,16,4,power_law_1.2,0.10684032440185547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,16,4,power_law_1.2,0.11626111984252929
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,16,4,power_law_1.2,0.15801983833312988
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,32,2,balanced,0.023539199829101562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,32,2,balanced,0.02560895919799805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,32,2,balanced,0.0289574408531189
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,32,2,balanced,0.03426687955856324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,32,2,balanced,0.044418559074401856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,32,2,balanced,0.051613440513610845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,32,2,balanced,0.05133567810058594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,32,2,balanced,0.05054336071014405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,32,2,balanced,0.09174783706665039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,32,2,balanced,0.09270784378051758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,32,2,balanced,0.09324928283691405
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,16,4,power_law_1.2,0.23026304244995116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,32,2,balanced,0.10338047981262206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,32,2,balanced,0.1012992000579834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,32,2,balanced,0.1080742359161377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,32,2,balanced,0.11636223793029785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,32,2,balanced,0.12499456405639649
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,32,2,balanced,0.14433152198791505
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,32,2,balanced,0.1117734432220459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,32,2,balanced,0.1625267219543457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,32,2,balanced,0.20183040618896486
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,16,4,power_law_1.2,0.32545665740966795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,32,2,balanced,0.23898496627807617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,32,2,balanced,0.31382783889770505
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,32,2,balanced,0.3894348907470703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,32,2,balanced,0.5360128021240234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,32,2,balanced,0.7147737884521485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,32,2,balanced,0.029287679195404055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,32,2,balanced,0.034666240215301514
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,32,2,balanced,0.044824318885803224
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,32,2,balanced,1.3468914794921876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,32,2,balanced,0.054448637962341305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,32,2,balanced,0.05839744091033936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,32,2,balanced,0.05566207885742187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,32,2,balanced,0.05627520084381103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,32,2,balanced,0.07085824012756348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,32,2,balanced,0.07089536190032959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,32,2,balanced,0.0715238380432129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,32,2,balanced,0.07410431861877441
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,32,2,balanced,0.07434751987457275
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,32,2,balanced,0.07720575809478761
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,32,2,balanced,0.07792255878448487
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,32,2,balanced,0.08163968086242676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,32,2,balanced,0.08615936279296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,32,2,balanced,0.09808511734008789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,32,2,balanced,0.10764287948608398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,32,2,balanced,0.030478079319000245
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,32,2,balanced,0.15492608070373534
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,32,2,balanced,0.2281817626953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,32,2,balanced,0.2716518402099609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,32,2,balanced,0.40226303100585936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,32,2,balanced,0.518021125793457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,32,2,balanced,0.13149951934814452
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,32,2,power_law_1.01,0.025207040309906004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,32,2,balanced,0.756263656616211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,32,2,power_law_1.01,0.02724479913711548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,32,2,power_law_1.01,0.0321727991104126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,32,2,power_law_1.01,0.044615678787231446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,32,2,balanced,1.0004531097412108
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,32,2,power_law_1.01,0.04217472076416016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,32,2,power_law_1.01,0.05492735862731933
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,32,2,power_law_1.01,0.05137792110443116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,32,2,power_law_1.01,0.052184319496154784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,32,2,power_law_1.01,0.09226112365722657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,32,2,power_law_1.01,0.0906828784942627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,32,2,power_law_1.01,0.09542783737182617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,32,2,power_law_1.01,0.10212863922119139
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,32,2,power_law_1.01,0.10837120056152343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,32,2,power_law_1.01,0.11301759719848632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,32,2,power_law_1.01,0.11784319877624512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,32,2,power_law_1.01,0.12760704040527343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,32,2,power_law_1.01,0.14736000061035157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,32,2,power_law_1.01,0.16630016326904298
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,32,2,power_law_1.01,0.20597120285034182
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,32,2,power_law_1.01,0.24455936431884764
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,32,2,power_law_1.01,0.32163070678710937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,32,2,power_law_1.01,0.398812141418457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,32,2,power_law_1.01,0.5707519912719727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,32,2,power_law_1.01,0.10079360008239746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,32,2,balanced,1.0331788635253907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,32,2,power_law_1.01,0.7449343872070313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,32,2,power_law_1.01,0.032131841182708745
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,32,2,power_law_1.01,0.03590656042098999
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,32,2,power_law_1.01,1.0712255859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,32,2,power_law_1.01,0.04442880153656006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,32,2,power_law_1.01,0.04732160091400146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,32,2,power_law_1.01,0.05300608158111573
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,32,2,power_law_1.01,1.4111795043945312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,32,2,power_law_1.01,0.05379327774047852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,32,2,power_law_1.01,0.05439231872558594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,32,2,power_law_1.01,0.06616191864013672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,32,2,power_law_1.01,0.06867455959320068
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,32,2,power_law_1.01,0.0706278419494629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,32,2,power_law_1.01,0.0731379222869873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,32,2,power_law_1.01,0.07396351814270019
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,32,2,power_law_1.01,0.07723775863647461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,32,2,power_law_1.01,0.07961599826812744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,32,2,power_law_1.01,0.08137599945068359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,32,2,power_law_1.01,0.08853631973266601
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,32,2,power_law_1.01,0.10473216056823731
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,32,2,power_law_1.01,0.12055551528930664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,32,2,power_law_1.01,0.15079039573669434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,32,2,power_law_1.01,0.18265600204467775
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,32,2,power_law_1.01,0.2555583953857422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,32,2,power_law_1.01,0.31700992584228516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,32,2,power_law_1.01,0.4378662490844727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,32,2,power_law_1.01,0.5652275085449219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,32,2,power_law_1.2,0.02847232103347778
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,32,2,power_law_1.01,0.832806396484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,32,2,power_law_1.2,0.026803200244903562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,32,2,power_law_1.2,0.034190080165863036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,32,2,power_law_1.01,1.0755315399169922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,32,2,power_law_1.2,0.036974079608917236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,32,2,power_law_1.2,0.046115841865539554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,32,2,power_law_1.2,0.04694784164428711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,32,2,power_law_1.2,0.051988477706909175
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,32,2,power_law_1.2,0.053350400924682614
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,32,2,power_law_1.2,0.08904319763183594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,32,2,power_law_1.2,0.09238656044006348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,32,2,power_law_1.2,0.09544320106506347
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,32,2,power_law_1.2,0.09918463706970215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,32,2,power_law_1.2,0.10439040184020995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,32,2,power_law_1.2,0.10952704429626466
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,32,2,power_law_1.2,0.11328384399414063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,32,2,power_law_1.2,0.11834239959716797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,32,2,power_law_1.2,0.12707327842712401
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,32,2,power_law_1.2,0.1477785587310791
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,32,2,power_law_1.2,0.1676288032531738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,32,2,power_law_1.2,0.20570367813110352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,32,2,power_law_1.2,0.2454310417175293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,32,2,power_law_1.2,0.32316673278808594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,32,2,power_law_1.2,0.39913726806640626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,32,2,power_law_1.2,0.5786214447021484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,32,2,power_law_1.01,0.02746495962142944
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,32,2,power_law_1.2,0.027384319305419923
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,32,2,power_law_1.2,1.0810765075683595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,32,2,power_law_1.2,0.03509632110595703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,32,2,power_law_1.2,0.04443647861480713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,32,2,power_law_1.2,0.04817791938781738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,32,2,power_law_1.2,0.056294398307800295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,32,2,power_law_1.2,0.05368447780609131
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,32,2,power_law_1.2,0.05538815975189208
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,32,2,power_law_1.2,0.0664358377456665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,32,2,power_law_1.2,0.06724095821380616
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,32,2,power_law_1.2,0.07104896068572998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,32,2,power_law_1.2,0.07313920021057128
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,32,2,power_law_1.2,0.07508992195129395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,32,2,power_law_1.2,0.07839615821838379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,32,2,power_law_1.2,0.08082559585571289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,32,2,power_law_1.2,0.08292223930358886
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,32,2,power_law_1.2,0.08974592208862305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,32,2,power_law_1.2,0.10697600364685059
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,32,2,power_law_1.2,0.12123135566711425
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,32,2,power_law_1.2,0.15312000274658205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,32,2,power_law_1.2,0.1849510383605957
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,16,4,power_law_1.2,0.43548927307128904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,32,2,power_law_1.2,0.3185087966918945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,32,2,power_law_1.2,0.2607449531555176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,32,2,power_law_1.2,0.032561919689178466
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,16,4,power_law_1.2,0.6038988876342773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,32,2,power_law_1.2,0.7449126434326172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,2,32,balanced,0.03908992052078247
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,2,32,balanced,0.038691840171813964
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,2,32,balanced,0.039184639453887934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,2,32,balanced,0.05874815940856933
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,2,32,balanced,0.0714355182647705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,2,32,balanced,0.12328703880310059
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,2,32,balanced,0.12499584197998045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,2,32,balanced,0.12476160049438476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,2,32,balanced,0.12512384414672853
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,32,2,power_law_1.2,0.4437004852294922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,2,32,balanced,0.12728063583374022
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,2,32,balanced,0.1312384033203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,2,32,balanced,0.13220352172851563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,2,32,balanced,0.13665663719177246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,2,32,balanced,0.13729151725769043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,2,32,balanced,0.14309887886047362
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,2,32,balanced,0.14402688026428223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,2,32,balanced,0.15334272384643555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,2,32,balanced,0.1617433547973633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,2,32,balanced,0.18524415969848634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,2,32,balanced,0.14581503868103027
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,2,32,balanced,0.20518016815185547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,2,32,balanced,0.24306432723999025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,2,32,balanced,0.297127685546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,32,2,power_law_1.2,0.5678553771972656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,2,32,balanced,0.40488449096679685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,2,32,balanced,0.5041856002807618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,2,32,balanced,0.03481856107711792
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,2,32,balanced,0.6990156555175782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,2,32,balanced,0.03459199905395508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,2,32,balanced,0.8973798370361328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,2,32,balanced,0.03375103950500488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,2,32,balanced,0.03493247985839844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,2,32,balanced,0.0559219217300415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,2,32,balanced,0.0559219217300415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,2,32,balanced,0.05993728160858154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,2,32,balanced,0.05742464065551758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,2,32,balanced,0.05998720169067383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,2,32,balanced,0.062117118835449216
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,2,32,balanced,0.060823040008544924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,2,32,balanced,0.06314623832702637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,2,32,balanced,0.06352000236511231
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,2,32,balanced,0.06918655872344971
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,2,32,balanced,0.07169919967651367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,2,32,balanced,0.06689023971557617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,2,32,balanced,0.07433087825775146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,2,32,balanced,0.0438156795501709
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,2,32,balanced,0.07608064174652099
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,2,32,balanced,0.08610048294067382
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,2,32,balanced,0.09660415649414063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,2,32,balanced,0.12272000312805176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,32,2,power_law_1.2,0.8493196868896484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,2,32,balanced,0.14621824264526367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,2,32,balanced,0.20477184295654297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,2,32,balanced,0.25523712158203127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,2,32,balanced,0.3406041717529297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,2,32,balanced,0.44876670837402344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,32,2,power_law_1.2,1.4173695373535158
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,16,4,power_law_1.2,0.9700800323486328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,2,32,balanced,0.01979647994041443
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,2,32,balanced,0.024386560916900633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,2,32,balanced,0.02431488037109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,2,32,balanced,0.024318718910217287
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,2,32,balanced,0.02437119960784912
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,2,32,balanced,0.024607999324798586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,2,32,balanced,0.01985152006149292
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,2,32,balanced,0.0194649600982666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,2,32,balanced,0.025034239292144777
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,2,32,balanced,0.024344320297241213
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,2,32,balanced,0.024723200798034667
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,2,32,balanced,0.024729599952697752
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,2,32,balanced,0.025096960067749023
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,2,32,balanced,0.02520064115524292
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,2,32,balanced,0.025597438812255856
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,2,32,balanced,0.026842880249023437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,2,32,balanced,0.020207359790802
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,2,32,balanced,0.025196800231933592
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,2,32,balanced,0.024093439579010008
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,2,32,balanced,0.021131520271301267
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,2,32,balanced,0.023045120239257814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,2,32,power_law_1.01,0.07371263980865479
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,2,32,power_law_1.01,0.07585023880004883
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,2,32,balanced,0.05331071853637696
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,2,32,balanced,0.06422272205352783
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,2,32,balanced,0.041912322044372556
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,2,32,power_law_1.01,0.11994879722595216
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,2,32,power_law_1.01,0.12362624168395997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,2,32,power_law_1.01,0.12445311546325684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,2,32,power_law_1.01,0.1300979232788086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,2,32,power_law_1.01,0.12875519752502443
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,2,32,power_law_1.01,0.1297651195526123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,2,32,power_law_1.01,0.13569536209106445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,2,32,power_law_1.01,0.1423859214782715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,2,32,power_law_1.01,0.07558400154113769
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,2,32,power_law_1.01,0.14572799682617188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,2,32,power_law_1.01,0.148472318649292
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,2,32,power_law_1.01,0.15672191619873047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,2,32,power_law_1.01,0.12248064041137696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,2,32,power_law_1.01,0.15042176246643066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,2,32,power_law_1.01,0.16927743911743165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,2,32,power_law_1.01,0.19032320022583007
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,2,32,power_law_1.01,0.2208038330078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,2,32,power_law_1.01,0.27283327102661137
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,2,32,balanced,0.04474495887756348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,2,32,power_law_1.01,0.35849857330322266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,2,32,power_law_1.01,0.5130483245849609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,2,32,power_law_1.01,0.6601203155517579
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,2,32,power_law_1.01,0.9247500610351562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,2,32,power_law_1.01,1.0876646423339844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,2,32,power_law_1.01,0.04864255905151367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,2,32,power_law_1.01,0.04859392166137695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,2,32,power_law_1.01,0.04839807987213135
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,2,32,power_law_1.01,0.05074048042297363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,2,32,power_law_1.01,0.04978687763214111
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,2,32,power_law_1.01,1.776976623535156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,2,32,power_law_1.01,0.05876224040985108
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,2,32,power_law_1.01,0.05774975776672363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,2,32,power_law_1.01,0.05751039981842041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,2,32,power_law_1.01,0.06240511894226074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,2,32,power_law_1.01,0.06515456199645996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,2,32,power_law_1.01,0.06821375846862793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,2,32,power_law_1.01,0.07352831840515137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,2,32,power_law_1.01,0.0748364782333374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,2,32,power_law_1.01,0.07188223838806153
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,2,32,power_law_1.01,0.07613440036773682
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,2,32,power_law_1.01,0.07968895912170411
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,2,32,power_law_1.01,0.09284095764160157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,2,32,power_law_1.01,0.09969152450561523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,2,32,power_law_1.01,2.8009652709960937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,2,32,power_law_1.01,0.13233280181884766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,2,32,power_law_1.01,0.16382207870483398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,2,32,power_law_1.01,0.22461568832397463
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,2,32,power_law_1.01,0.05969664096832276
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,2,32,power_law_1.01,0.3162227249145508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,2,32,power_law_1.01,0.3890316772460937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,2,32,power_law_1.01,0.022632958889007567
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,32,2,power_law_1.2,1.075188446044922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,2,32,power_law_1.01,0.02349695920944214
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,2,32,power_law_1.01,0.534382095336914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,2,32,power_law_1.01,0.02304896116256714
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,2,32,balanced,0.07797887802124023
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,2,32,power_law_1.01,0.023366398811340332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,2,32,power_law_1.01,0.0235263991355896
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,2,32,power_law_1.01,0.024126720428466798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,2,32,power_law_1.01,0.8287449645996094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,2,32,power_law_1.01,0.024154880046844483
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,2,32,power_law_1.01,0.023957760334014894
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,2,32,power_law_1.01,0.024796159267425538
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,2,32,power_law_1.01,0.024375040531158448
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,2,32,power_law_1.01,1.095198745727539
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,2,32,power_law_1.01,0.02434943914413452
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,2,32,power_law_1.01,0.024369919300079347
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,2,32,power_law_1.01,0.025588479042053226
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,2,32,power_law_1.01,0.025372159481048585
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,2,32,power_law_1.01,0.03508352041244507
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,2,32,power_law_1.01,0.024104959964752197
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,2,32,balanced,0.09485440254211426
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,2,32,power_law_1.01,0.02357759952545166
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,2,32,power_law_1.01,0.02857856035232544
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,2,32,power_law_1.01,0.0555788803100586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,2,32,power_law_1.01,0.04658559799194336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,2,32,power_law_1.01,0.07612671852111816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,2,32,power_law_1.2,0.07498112201690674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,2,32,power_law_1.2,0.07560448169708252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,2,32,power_law_1.2,0.07620223999023437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,2,32,power_law_1.2,0.07144320011138916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,2,32,power_law_1.01,0.03943039894104004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,2,32,power_law_1.2,0.12274304389953614
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,2,32,power_law_1.2,0.12385408401489258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,2,32,power_law_1.01,0.08007807731628418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,2,32,power_law_1.2,0.12853887557983398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,2,32,power_law_1.2,0.1293452835083008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,2,32,power_law_1.2,0.13242752075195313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,2,32,power_law_1.2,0.1385638427734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,2,32,power_law_1.2,0.14197248458862305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,2,32,power_law_1.2,0.1457356834411621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,2,32,power_law_1.2,0.15165696144104004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,2,32,power_law_1.2,0.14903552055358887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,2,32,power_law_1.2,0.16603391647338867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,2,32,power_law_1.2,0.1779225540161133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,2,32,power_law_1.2,0.12415231704711913
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,2,32,power_law_1.2,0.2126156806945801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,2,32,power_law_1.2,0.23417728424072265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,2,32,power_law_1.2,0.33888126373291017
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,2,32,power_law_1.2,0.4512704086303711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,2,32,power_law_1.2,0.6827865600585937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,2,32,power_law_1.01,0.11836031913757324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,2,32,power_law_1.2,0.8906240081787109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,2,32,power_law_1.2,0.04847231864929199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,2,32,power_law_1.2,0.04876416206359863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,2,32,power_law_1.2,1.205687026977539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,2,32,power_law_1.2,0.05016960144042969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,2,32,power_law_1.2,0.047260160446166995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,2,32,power_law_1.2,0.052487678527832024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,2,32,power_law_1.2,0.05510655879974365
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,2,32,power_law_1.2,0.06066431999206543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,2,32,power_law_1.2,0.06146815776824951
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,2,32,power_law_1.2,0.06171904087066651
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,2,32,power_law_1.2,0.06291200160980225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,2,32,power_law_1.2,0.06788224220275879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,2,32,power_law_1.2,0.07181824207305908
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,2,32,power_law_1.2,0.07399936199188233
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,2,32,power_law_1.2,0.06882175922393799
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,2,32,power_law_1.2,0.07302400112152099
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,2,32,power_law_1.2,2.5807577514648434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,2,32,power_law_1.2,0.0788812780380249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,2,32,power_law_1.2,0.08428544044494629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,2,32,power_law_1.2,0.1020479965209961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,2,32,power_law_1.2,1.817084197998047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,2,32,power_law_1.2,0.11592320442199706
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,2,32,power_law_1.2,0.18626815795898438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,2,32,power_law_1.2,0.19228031158447265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,2,32,power_law_1.2,0.3125337600708008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,2,32,power_law_1.2,0.38525054931640623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,2,32,power_law_1.2,0.5348441696166992
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,2,32,power_law_1.2,0.02345472097396851
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,2,32,power_law_1.2,0.6711014556884766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,2,32,power_law_1.2,0.0235481595993042
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,2,32,power_law_1.2,0.023185920715332032
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,2,32,power_law_1.2,0.022664959430694583
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,2,32,power_law_1.2,0.023348479270935057
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,2,32,power_law_1.2,0.023873279094696047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,2,32,power_law_1.2,0.024764161109924316
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,2,32,power_law_1.2,0.02402559995651245
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,2,32,power_law_1.2,0.02401279926300049
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,2,32,power_law_1.2,0.024510719776153565
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,2,32,power_law_1.2,0.024994559288024902
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,2,32,power_law_1.2,0.02428031921386719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,2,32,power_law_1.2,0.024418559074401856
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,2,32,power_law_1.2,0.024377601146697996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,2,32,power_law_1.2,0.024761600494384767
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,2,32,power_law_1.2,0.030475521087646486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,2,32,power_law_1.2,1.0412313842773437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,2,32,power_law_1.2,0.03900928020477295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,2,32,power_law_1.2,3.0188479614257813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,2,32,power_law_1.2,1.5777728271484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,2,32,power_law_1.01,0.20447616577148436
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,2,32,power_law_1.01,0.14524031639099122
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,2,32,power_law_1.01,0.2929664039611816
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,2,32,power_law_1.2,0.027969279289245606
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,2,32,power_law_1.2,0.046694397926330566
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,2,32,power_law_1.2,0.06937471866607667
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,2,32,power_law_1.2,0.06087423801422119
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,4,16,balanced,0.03418623924255371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,4,16,balanced,0.03344383955001831
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,4,16,balanced,0.05584383964538574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,4,16,balanced,0.07176064014434815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,4,16,balanced,0.12192768096923827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,4,16,balanced,0.12414079666137696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,4,16,balanced,0.03331968069076538
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,4,16,balanced,0.12697088241577148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,4,16,balanced,0.12803327560424804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,4,16,balanced,0.12927359580993653
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,4,16,balanced,0.1365824031829834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,4,16,balanced,0.1398259162902832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,4,16,balanced,0.1427558422088623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,4,16,balanced,0.1353382396697998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,4,16,balanced,0.14042240142822265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,4,16,balanced,0.14173952102661133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,4,16,balanced,0.1483839988708496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,4,16,balanced,0.16468351364135742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,4,16,balanced,0.1827123260498047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,4,16,balanced,0.21620096206665038
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,4,16,balanced,0.2442483139038086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,4,16,balanced,0.3042380714416504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,4,16,balanced,0.3749504089355469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,4,16,balanced,0.5272345733642578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,4,16,balanced,0.687409896850586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,4,16,balanced,0.9634073638916016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,4,16,balanced,0.03255935907363892
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,4,16,balanced,0.03293567895889282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,4,16,balanced,1.258245086669922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,4,16,balanced,0.03500927925109863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,4,16,balanced,0.04779007911682129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,4,16,balanced,0.06002175807952881
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,4,16,balanced,0.06170752048492432
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,4,16,balanced,0.0628275203704834
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,2,32,power_law_1.2,0.08922240257263184
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,4,16,balanced,0.03257728099822998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,4,16,balanced,0.06399231910705566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,4,16,balanced,0.06708223819732666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,4,16,balanced,0.06932991981506348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,4,16,balanced,0.07234687805175781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,4,16,balanced,0.07460991859436035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,4,16,balanced,0.06884096145629884
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,4,16,balanced,0.0702233600616455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,4,16,balanced,0.0742412805557251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,4,16,balanced,0.07521535873413085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,4,16,balanced,0.08412672042846679
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,4,16,balanced,0.09517696380615234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,4,16,balanced,0.11107839584350584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,4,16,balanced,0.13025535583496092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,4,16,balanced,0.17529983520507814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,4,16,balanced,0.21192319869995116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,4,16,balanced,0.3090905570983887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,4,16,balanced,0.38761215209960936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,4,16,balanced,0.5315046310424805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,4,16,balanced,0.70110595703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,2,32,power_law_1.2,0.1111398410797119
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,2,32,power_law_1.2,0.2181644821166992
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,2,32,power_law_1.2,0.16272640228271484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,2,32,power_law_1.2,0.35383041381835934
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,2,32,power_law_1.2,0.49862529754638674
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,4,16,balanced,0.019845119714736938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,4,16,balanced,0.027957758903503417
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,4,16,balanced,0.028215041160583498
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,4,16,balanced,0.021401600837707518
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,4,16,balanced,0.027975680828094485
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,4,16,balanced,0.028368639945983886
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,4,16,balanced,0.028520960807800293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,4,16,balanced,0.019646719694137574
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,4,16,balanced,0.028175361156463623
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,4,16,balanced,0.02825344085693359
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,4,16,balanced,0.028908801078796387
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,4,16,balanced,0.02945663928985596
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,4,16,balanced,0.03056256055831909
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,4,16,balanced,0.03115904092788696
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,4,16,balanced,0.03287039995193482
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,4,16,balanced,0.034663679599761965
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,4,16,balanced,0.02027008056640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,4,16,balanced,0.024284160137176512
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,4,16,balanced,0.027901439666748046
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,4,16,balanced,0.027630081176757814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,4,16,power_law_1.01,0.0698956823348999
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,4,16,power_law_1.01,0.07607552051544189
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,4,16,power_law_1.01,0.07432447910308838
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,4,16,power_law_1.01,0.07834112167358398
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,4,16,balanced,0.028044800758361816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,4,16,power_law_1.01,0.1228544044494629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,4,16,power_law_1.01,0.12023039817810059
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,4,16,power_law_1.01,0.1322214412689209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,4,16,power_law_1.01,0.12603648185729982
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,4,16,power_law_1.01,0.12919808387756349
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,4,16,power_law_1.01,0.1301542377471924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,4,16,power_law_1.01,0.13599488258361817
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,4,16,power_law_1.01,0.14015999794006348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,4,16,power_law_1.01,0.1464780807495117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,4,16,power_law_1.01,0.14202239990234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,4,16,power_law_1.01,0.14518400192260744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,4,16,power_law_1.01,0.15408512115478518
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,4,16,power_law_1.01,0.1644339179992676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,4,16,power_law_1.01,0.1915763282775879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,4,16,power_law_1.01,0.2054911994934082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,4,16,power_law_1.01,0.26180864334106446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,4,16,power_law_1.01,0.3197695922851563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,4,16,power_law_1.01,0.4822540664672852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,4,16,power_law_1.01,0.5713459014892578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,4,16,power_law_1.01,0.8472025299072266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,4,16,power_law_1.01,1.0815129852294922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,4,16,power_law_1.01,1.5220889282226562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,4,16,power_law_1.01,0.04417280197143555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,4,16,power_law_1.01,2.270707244873047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,4,16,power_law_1.01,0.05172607898712158
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,4,16,power_law_1.01,0.0491648006439209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,4,16,power_law_1.01,0.05303679943084717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,4,16,power_law_1.01,0.0575167989730835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,4,16,power_law_1.01,0.06381696224212646
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,4,16,power_law_1.01,0.060419840812683104
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,4,16,power_law_1.01,0.06342144012451172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,4,16,power_law_1.01,0.06490240097045899
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,4,16,power_law_1.01,0.0680140781402588
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,4,16,power_law_1.01,0.07104512214660644
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,4,16,balanced,0.07169536113739014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,4,16,power_law_1.01,0.073918719291687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,4,16,power_law_1.01,0.07677696228027345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,4,16,power_law_1.01,0.0706496000289917
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,4,16,power_law_1.01,0.07561600208282471
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,4,16,power_law_1.01,0.0774118423461914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,4,16,power_law_1.01,0.08254719734191894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,4,16,power_law_1.01,0.0935859203338623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,4,16,power_law_1.01,0.11400447845458986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,4,16,power_law_1.01,0.1297599983215332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,4,16,balanced,0.052021760940551755
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,4,16,power_law_1.01,0.17069055557250976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,4,16,power_law_1.01,0.25880960464477537
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,4,16,power_law_1.01,0.29153024673461914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,4,16,power_law_1.01,0.44140033721923827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,4,16,power_law_1.01,0.5557414245605469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,4,16,balanced,0.10626943588256836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,4,16,power_law_1.01,0.02446592092514038
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,4,16,power_law_1.01,0.027637760639190673
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,4,16,power_law_1.01,0.026568961143493653
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,4,16,power_law_1.01,0.7794137573242187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,4,16,power_law_1.01,0.027243518829345705
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,4,16,power_law_1.01,0.027302401065826414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,4,16,power_law_1.01,0.027534079551696778
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,4,16,power_law_1.01,0.02712575912475586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,4,16,power_law_1.01,0.028257279396057128
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,4,16,power_law_1.01,0.028055040836334227
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,4,16,power_law_1.01,0.028450560569763184
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,4,16,balanced,0.13406335830688476
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,4,16,balanced,0.08341119766235353
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,4,16,power_law_1.01,0.028866560459136964
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,4,16,power_law_1.01,0.029232640266418457
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,4,16,power_law_1.01,0.030439679622650144
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,4,16,power_law_1.01,0.030987520217895508
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,4,16,power_law_1.01,0.028407039642333987
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,4,16,balanced,0.05588096141815185
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,4,16,power_law_1.01,1.1461670684814453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,4,16,power_law_1.01,0.02940416097640991
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,4,16,power_law_1.01,0.034114561080932616
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,4,16,power_law_1.01,0.04470272064208984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,4,16,power_law_1.01,0.038293759822845455
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,4,16,power_law_1.01,0.08862079620361328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,4,16,power_law_1.2,0.06877439975738525
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,4,16,power_law_1.01,0.12651391983032229
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,4,16,power_law_1.01,0.0567526388168335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,4,16,power_law_1.2,0.07535103797912598
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,4,16,power_law_1.2,0.07449215888977051
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,4,16,power_law_1.01,0.06670464038848876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,4,16,power_law_1.2,0.07760128021240234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,4,16,power_law_1.2,0.11561856269836426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,4,16,power_law_1.2,0.1200972843170166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,4,16,power_law_1.2,0.12477312088012696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,4,16,power_law_1.2,0.12653823852539064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,4,16,power_law_1.2,0.12863360404968263
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,4,16,power_law_1.2,0.12882944107055666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,4,16,power_law_1.2,0.1381503963470459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,4,16,power_law_1.2,0.13924351692199707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,4,16,power_law_1.2,0.1425228786468506
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,4,16,power_law_1.2,0.1466534423828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,4,16,power_law_1.2,0.15673727989196778
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,4,16,power_law_1.2,0.16894975662231446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,4,16,power_law_1.2,0.20008575439453127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,4,16,power_law_1.2,0.21253887176513672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,4,16,power_law_1.2,0.27942399978637694
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,4,16,power_law_1.2,0.34675582885742184
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,4,16,power_law_1.01,0.1000755214691162
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,4,16,power_law_1.2,0.1433843231201172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,4,16,power_law_1.2,0.5603238296508789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,4,16,power_law_1.2,0.6925068664550781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,4,16,power_law_1.2,0.042453761100769045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,4,16,power_law_1.2,0.0502348804473877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,4,16,power_law_1.2,1.0617945861816405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,4,16,power_law_1.2,0.05100031852722168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,4,16,power_law_1.2,1.3889138793945313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,4,16,power_law_1.2,0.05187712192535401
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,4,16,power_law_1.2,0.05348351955413818
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,4,16,power_law_1.2,0.059399681091308595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,4,16,power_law_1.2,0.0643059206008911
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,4,16,power_law_1.2,0.06492544174194335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,4,16,power_law_1.2,0.06584959983825683
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,4,16,power_law_1.2,0.06609792232513427
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,4,16,power_law_1.2,0.07157120227813721
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,4,16,power_law_1.2,0.0765721607208252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,4,16,power_law_1.2,0.07310336112976075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,4,16,power_law_1.2,0.07214975833892823
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,4,16,power_law_1.2,0.07483263969421386
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,4,16,power_law_1.2,0.07975039958953858
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,4,16,power_law_1.2,0.08500096321105957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,4,16,power_law_1.2,0.1061081600189209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,4,16,power_law_1.2,0.11970303535461427
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,4,16,power_law_1.2,0.14822272300720213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,4,16,power_law_1.2,0.2014348793029785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,4,16,power_law_1.2,2.4778509521484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,4,16,power_law_1.2,0.36527103424072266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,4,16,power_law_1.2,1.9969293212890626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,4,16,power_law_1.2,0.5395187377929688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,4,16,power_law_1.2,0.5662284851074219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,4,16,power_law_1.2,0.2823142433166504
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,4,16,power_law_1.2,0.024375040531158448
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,4,16,power_law_1.2,0.027139840126037595
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,4,16,power_law_1.2,0.0275929594039917
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,4,16,power_law_1.2,0.026992640495300292
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,4,16,power_law_1.2,0.02611072063446045
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,4,16,power_law_1.2,0.02673536062240601
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,4,16,power_law_1.2,0.028047358989715575
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,4,16,power_law_1.2,0.027631359100341795
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,4,16,power_law_1.2,0.027489280700683592
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,4,16,power_law_1.2,0.028032000064849856
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,4,16,power_law_1.2,0.028801279067993162
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,4,16,power_law_1.2,0.029105920791625977
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,4,16,power_law_1.2,0.02919552087783813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,4,16,power_law_1.2,0.029675519466400145
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,4,16,power_law_1.2,0.03631488084793091
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,4,16,power_law_1.2,1.0207218933105469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,4,16,power_law_1.2,1.3555775451660157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,4,16,power_law_1.01,0.18122751235961915
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,4,16,power_law_1.01,0.254648323059082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,4,16,power_law_1.01,0.42740608215332027
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,4,16,power_law_1.2,0.03209088087081909
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,4,16,power_law_1.2,0.033958399295806886
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,4,16,power_law_1.2,0.04348159790039062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,4,16,power_law_1.2,0.05007359981536865
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,4,16,power_law_1.2,0.06290815830230713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,8,8,balanced,0.030021119117736816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,8,8,balanced,0.031001598834991456
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,8,8,balanced,0.03335551977157593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,8,8,balanced,0.058305277824401855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,8,8,balanced,0.0749836778640747
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,8,8,balanced,0.12698880195617676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,8,8,balanced,0.13018367767333985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,8,8,balanced,0.13484288215637208
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,8,8,balanced,0.13642623901367187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,8,8,balanced,0.13982975959777832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,8,8,balanced,0.1338764762878418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,8,8,balanced,0.13828607559204104
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,8,8,balanced,0.14115839958190918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,8,8,balanced,0.14691712379455565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,8,8,balanced,0.15094271659851075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,8,8,balanced,0.1573311996459961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,8,8,balanced,0.1722060775756836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,8,8,balanced,0.19617023468017578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,8,8,balanced,0.2239014434814453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,8,8,balanced,0.2816217613220215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,8,8,balanced,0.33295745849609376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,8,8,balanced,0.4440755081176758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,8,8,balanced,0.5550822448730469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,4,16,power_law_1.2,0.08120063781738282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,8,8,balanced,0.8048844909667968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,4,16,power_law_1.2,0.11601792335510255
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,4,16,power_law_1.2,0.14852992057800293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,8,8,balanced,1.0634969329833983
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,8,8,balanced,0.034124801158905024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,8,8,balanced,0.03747711896896362
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,8,8,balanced,0.04209407806396485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,8,8,balanced,1.5234393310546874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,8,8,balanced,0.058302721977233886
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,8,8,balanced,0.07426688194274902
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,8,8,balanced,0.07545983791351318
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,8,8,balanced,0.03247359991073608
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,8,8,balanced,0.07866367816925049
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,8,8,balanced,0.08082559585571289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,8,8,balanced,0.08309247970581055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,8,8,balanced,0.07550079822540283
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,8,8,balanced,2.0137126159667966
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,8,8,balanced,0.0790502405166626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,8,8,balanced,0.08079232215881348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,8,8,balanced,0.08691200256347656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,8,8,balanced,0.08668543815612792
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,8,8,balanced,0.09835519790649414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,8,8,balanced,0.11405440330505372
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,8,8,balanced,0.1326643180847168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,8,8,balanced,0.16631807327270506
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,8,8,balanced,0.20001152038574216
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,8,8,balanced,0.28323328018188476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,8,8,balanced,0.3492736053466797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,8,8,balanced,0.509040641784668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,8,8,balanced,0.09012607574462891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,8,8,balanced,0.6527629089355469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,8,8,balanced,0.9416166687011719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,8,8,balanced,1.2223385620117189
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,4,16,power_law_1.2,0.17561471939086915
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,4,16,power_law_1.2,0.27214975357055665
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,4,16,power_law_1.2,0.403870735168457
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,4,16,power_law_1.2,0.45082878112792973
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,8,8,balanced,0.021130239963531493
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,8,8,balanced,0.03707263946533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,8,8,balanced,0.02160640001296997
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,8,8,balanced,0.037556478977203364
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,8,8,balanced,0.0370739197731018
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,8,8,balanced,0.037943038940429684
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,8,8,balanced,0.0381606388092041
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,8,8,balanced,0.02265984058380127
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,8,8,balanced,0.024317440986633302
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,8,8,balanced,0.038638079166412355
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,8,8,balanced,0.037962241172790526
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,8,8,balanced,0.03828991889953613
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,8,8,balanced,0.04008959770202637
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,8,8,balanced,0.041251840591430666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,8,8,balanced,0.04246016025543213
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,8,8,balanced,0.04558847904205322
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,8,8,balanced,0.04927231788635254
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,8,8,balanced,0.03092351913452148
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,8,8,balanced,0.036831998825073244
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,8,8,balanced,0.03659775972366333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,8,8,power_law_1.01,0.05478400230407715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,8,8,power_law_1.01,0.06779136180877685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,8,8,power_law_1.01,0.06509568214416503
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,8,8,power_law_1.01,0.07214719772338866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,8,8,power_law_1.01,0.07524991989135742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,8,8,power_law_1.01,0.1238745594024658
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,8,8,power_law_1.01,0.12454527854919432
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,8,8,power_law_1.01,0.12872960090637206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,8,8,power_law_1.01,0.1309107208251953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,8,8,power_law_1.01,0.135164155960083
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,8,8,power_law_1.01,0.14083200454711914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,8,8,power_law_1.01,0.14074111938476563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,8,8,power_law_1.01,0.1363212776184082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,8,8,power_law_1.01,0.14134655952453615
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,8,8,power_law_1.01,0.14845952033996582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,8,8,power_law_1.01,0.15273471832275393
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,8,8,power_law_1.01,0.1658086395263672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,8,8,power_law_1.01,0.1895667266845703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,8,8,power_law_1.01,0.21761920928955075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,8,8,power_law_1.01,0.2642444801330567
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,8,8,power_law_1.01,0.32583553314208985
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,8,8,balanced,0.036693758964538574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,8,8,power_law_1.01,0.4632588958740234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,8,8,power_law_1.01,0.5698380661010742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,8,8,power_law_1.01,0.7644032287597656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,8,8,power_law_1.01,1.0181811523437498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,8,8,power_law_1.01,1.54289794921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,8,8,power_law_1.01,0.03750272035598755
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,8,8,power_law_1.01,0.048983039855957033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,8,8,power_law_1.01,0.04765696048736572
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,8,8,power_law_1.01,0.054109439849853516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,8,8,power_law_1.01,2.25623291015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,8,8,power_law_1.01,0.06066175937652588
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,8,8,power_law_1.01,0.06650368213653565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,8,8,power_law_1.01,0.06946047782897949
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,8,8,power_law_1.01,0.0727180814743042
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,8,8,power_law_1.01,0.07756415843963624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,8,8,power_law_1.01,0.08488448143005371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,8,8,power_law_1.01,0.08028415679931641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,8,8,power_law_1.01,0.07650559902191162
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,8,8,power_law_1.01,0.07631999969482421
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,8,8,power_law_1.01,0.08042624473571777
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,8,8,power_law_1.01,0.08430848121643067
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,8,8,power_law_1.01,0.0870259189605713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,8,8,power_law_1.01,0.09364607810974121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,8,8,power_law_1.01,0.11389568328857422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,8,8,power_law_1.01,0.12182271957397461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,8,8,power_law_1.01,0.1683148765563965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,8,8,power_law_1.01,0.19696640014648437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,8,8,power_law_1.01,0.2830246353149414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,8,8,balanced,0.06881792068481445
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,8,8,balanced,0.075000319480896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,8,8,power_law_1.01,0.34399230957031246
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,8,8,balanced,0.10075903892517091
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,8,8,power_law_1.01,0.44414207458496097
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,8,8,power_law_1.01,0.024313600063323976
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,8,8,power_law_1.01,0.02980736017227173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,8,8,power_law_1.01,0.6656140899658203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,8,8,power_law_1.01,0.02884223937988281
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,8,8,power_law_1.01,0.031343359947204594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,8,8,power_law_1.01,0.8363533020019531
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,8,8,power_law_1.01,0.03379584074020386
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,8,8,power_law_1.01,0.03356544017791748
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,8,8,power_law_1.01,0.03511296033859253
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,8,8,power_law_1.01,0.035723519325256345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,8,8,power_law_1.01,1.2332275390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,8,8,power_law_1.01,0.03519232034683227
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,8,8,power_law_1.01,0.03599488019943237
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,8,8,power_law_1.01,0.036805119514465336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,8,8,power_law_1.01,0.03779072046279907
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,8,8,power_law_1.01,0.03952640056610107
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,8,8,power_law_1.01,0.03995136022567749
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,8,8,power_law_1.01,0.04139135837554932
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,8,8,power_law_1.01,0.045722880363464356
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,8,8,balanced,0.12115839958190917
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,8,8,power_law_1.01,0.03428992033004761
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,8,8,balanced,0.15955967903137208
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,8,8,balanced,0.20824447631835938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,8,8,power_law_1.01,0.08900863647460937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,8,8,power_law_1.01,0.08210304260253906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,8,8,power_law_1.01,0.14562432289123534
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,8,8,power_law_1.01,0.05294591903686523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,8,8,power_law_1.2,0.05517695903778076
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,8,8,power_law_1.2,0.06038527965545655
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,8,8,power_law_1.01,0.058534398078918456
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,8,8,power_law_1.2,0.06700160026550292
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,8,8,power_law_1.2,0.06935935974121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,8,8,power_law_1.2,0.07125887870788575
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,8,8,power_law_1.2,0.12421631813049316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,8,8,power_law_1.2,0.1251852798461914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,8,8,power_law_1.01,0.10875904083251953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,8,8,power_law_1.2,0.1284160041809082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,8,8,power_law_1.2,0.13265664100646973
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,8,8,power_law_1.2,0.13651455879211424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,8,8,power_law_1.2,0.13648768424987795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,8,8,power_law_1.2,0.13856639862060546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,8,8,power_law_1.01,0.21814655303955077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,8,8,power_law_1.2,0.14584320068359374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,8,8,power_law_1.2,0.15046784400939942
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,8,8,power_law_1.2,0.1531328010559082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,8,8,power_law_1.2,0.16511743545532226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,8,8,power_law_1.2,0.19396352767944336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,8,8,power_law_1.2,0.2204863929748535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,8,8,power_law_1.2,0.14257023811340333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,8,8,power_law_1.2,0.2864486312866211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,8,8,power_law_1.2,0.3569740676879883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,8,8,power_law_1.2,0.4879206466674805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,8,8,power_law_1.01,0.24510591506958007
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,8,8,power_law_1.2,0.6552358245849609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,8,8,power_law_1.2,0.0374015998840332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,8,8,power_law_1.2,0.04479231834411621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,8,8,power_law_1.2,0.04786816120147705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,8,8,power_law_1.2,0.05221375942230224
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,8,8,power_law_1.2,1.217470703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,8,8,power_law_1.2,0.05556608200073242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,8,8,power_law_1.2,0.07160064220428466
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,8,8,power_law_1.2,0.0693887996673584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,8,8,power_law_1.2,0.07363200187683105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,8,8,power_law_1.2,0.0787494421005249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,8,8,power_law_1.2,0.0823846435546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,8,8,power_law_1.2,0.08584320068359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,8,8,power_law_1.2,0.07569664001464844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,8,8,power_law_1.2,0.8821952056884765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,8,8,power_law_1.2,0.07878784179687501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,8,8,power_law_1.2,0.08454655647277833
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,8,8,power_law_1.2,0.08555264472961426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,8,8,power_law_1.2,0.08975232124328614
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,8,8,power_law_1.2,0.0963584041595459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,8,8,power_law_1.2,0.11344767570495606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,8,8,power_law_1.2,0.1302886390686035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,8,8,power_law_1.2,0.16044160842895508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,8,8,power_law_1.2,0.2084543991088867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,8,8,power_law_1.2,0.28941696166992187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,8,8,power_law_1.2,0.3600870513916016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,8,8,power_law_1.2,0.498260498046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,8,8,power_law_1.2,0.024312319755554198
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,8,8,power_law_1.2,1.2938636779785155
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,8,8,power_law_1.2,0.02707200050354004
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,8,8,power_law_1.2,0.029053440093994142
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,8,8,power_law_1.2,0.031667199134826664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,8,8,power_law_1.2,0.030389759540557858
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,8,8,power_law_1.2,0.03237760066986084
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,8,8,power_law_1.2,0.03343744039535522
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,8,8,power_law_1.2,0.03511935949325561
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,8,8,power_law_1.2,1.8104243469238281
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,8,8,power_law_1.2,0.03544447898864746
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,8,8,power_law_1.2,0.03616895914077759
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,8,8,power_law_1.2,1.050799331665039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,8,8,power_law_1.2,0.03682559967041016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,8,8,power_law_1.2,0.038133759498596195
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,8,8,power_law_1.2,0.03851392030715942
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,8,8,power_law_1.2,0.040733442306518555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,8,8,power_law_1.2,0.042652158737182616
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,8,8,power_law_1.2,0.7170342254638672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,8,8,power_law_1.2,2.3692723083496094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,8,8,power_law_1.01,0.3651776123046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,8,8,power_law_1.01,0.5273907089233398
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,8,8,power_law_1.2,0.06935296058654786
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,8,8,power_law_1.2,0.05803904056549072
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,8,8,power_law_1.2,0.047087359428405764
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,8,8,power_law_1.2,0.04478079795837402
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,8,8,power_law_1.2,0.08204928398132325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,16,4,balanced,0.031343359947204594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,16,4,balanced,0.03415168046951294
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,16,4,balanced,0.0435865592956543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,16,4,balanced,0.061712641716003415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,16,4,balanced,0.07819136142730712
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,16,4,balanced,0.09059583663940429
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,16,4,balanced,0.13765631675720216
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,16,4,balanced,0.1409267234802246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,16,4,balanced,0.14438400268554688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,16,4,balanced,0.14991616249084472
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,16,4,balanced,0.1426252841949463
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,16,4,balanced,0.14484992027282714
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,16,4,balanced,0.14969856262207032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,16,4,balanced,0.1540544033050537
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,16,4,balanced,0.15758079528808594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,16,4,balanced,0.1635327911376953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,16,4,balanced,0.17773696899414063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,16,4,balanced,0.20241792678833007
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,8,8,power_law_1.2,0.10784000396728514
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,16,4,balanced,0.22794879913330077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,16,4,balanced,0.28485631942749023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,16,4,balanced,0.33601280212402346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,16,4,balanced,0.4436633682250976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,16,4,balanced,0.5536396789550782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,16,4,balanced,0.8001395416259764
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,16,4,balanced,1.0549900817871094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,16,4,balanced,1.5080677795410158
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,16,4,balanced,0.03340800046920776
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,16,4,balanced,0.03626240015029907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,16,4,balanced,0.04245888233184815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,16,4,balanced,0.05027840137481689
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,16,4,balanced,1.9853517150878908
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,16,4,balanced,0.06525055885314941
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,16,4,balanced,0.07949696063995362
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,16,4,balanced,0.09107583999633789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,16,4,balanced,0.09241727828979493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,16,4,balanced,0.0958028793334961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,16,4,balanced,0.09962752342224121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,16,4,balanced,0.09150591850280762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,16,4,balanced,0.09478272438049316
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,8,8,power_law_1.2,0.15546624183654784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,16,4,balanced,0.09729408264160157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,16,4,balanced,0.10218367576599122
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,16,4,balanced,0.1024345588684082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,16,4,balanced,0.10561152458190919
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,16,4,balanced,0.11400064468383789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,16,4,balanced,0.1299443244934082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,16,4,balanced,0.1436838436126709
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,16,4,balanced,0.17607040405273439
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,16,4,balanced,0.21101440429687499
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,8,8,power_law_1.2,0.1972991943359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,16,4,balanced,0.31923583984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,16,4,balanced,0.38264575958251956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,16,4,balanced,0.5335321426391602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,16,4,balanced,0.7032537841796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,16,4,balanced,0.9807027435302734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,16,4,balanced,1.2978970336914064
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,8,8,power_law_1.2,0.31405824661254883
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,8,8,power_law_1.2,0.37206401824951174
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,8,8,power_law_1.2,0.9003097534179687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,8,8,power_law_1.2,0.6013875198364258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,16,4,balanced,0.02514688014984131
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,16,4,balanced,0.05334144115447999
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,16,4,balanced,0.053368320465087896
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,16,4,balanced,0.05387904167175293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,16,4,balanced,0.026521599292755126
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,16,4,balanced,0.027601919174194335
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,16,4,balanced,0.05375616073608398
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,16,4,balanced,0.054583039283752445
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,16,4,balanced,0.05664639949798585
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,16,4,balanced,0.05805696010589599
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,16,4,balanced,0.057734398841857916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,16,4,balanced,0.062403841018676756
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,16,4,balanced,0.06360191822052001
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,16,4,balanced,0.03301503896713257
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,16,4,balanced,0.07055615901947021
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,16,4,balanced,0.07390975952148438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,16,4,balanced,0.056294398307800295
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,16,4,balanced,0.047238402366638184
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,16,4,balanced,0.05516543865203858
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,16,4,balanced,0.05488383769989014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,16,4,power_law_1.01,0.03504640102386475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,16,4,power_law_1.01,0.058871040344238276
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,16,4,power_law_1.01,0.06163968086242676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,16,4,power_law_1.01,0.06900864124298096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,16,4,power_law_1.01,0.0768064022064209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,16,4,power_law_1.01,0.08941951751708985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,16,4,power_law_1.01,0.13501184463500976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,16,4,power_law_1.01,0.13948800086975097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,16,4,power_law_1.01,0.1414681625366211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,16,4,power_law_1.01,0.14399871826171876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,16,4,power_law_1.01,0.1485542392730713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,16,4,power_law_1.01,0.14269824028015138
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,16,4,balanced,0.05419136047363281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,16,4,power_law_1.01,0.14582271575927735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,16,4,power_law_1.01,0.1531443214416504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,16,4,power_law_1.01,0.15772031784057616
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,16,4,power_law_1.01,0.16355840682983397
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,16,4,power_law_1.01,0.1785651206970215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,16,4,power_law_1.01,0.2022425651550293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,16,4,power_law_1.01,0.23157119750976562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,16,4,power_law_1.01,0.28166784286499025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,16,4,power_law_1.01,0.33676158905029296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,16,4,power_law_1.01,0.42877822875976557
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,16,4,power_law_1.01,0.556234245300293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,16,4,power_law_1.01,0.8048947143554688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,16,4,power_law_1.01,1.0607552337646484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,16,4,power_law_1.01,1.5274560546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,16,4,power_law_1.01,0.0407206392288208
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,16,4,power_law_1.01,0.04358272075653076
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,16,4,power_law_1.01,0.048069119453430176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,16,4,power_law_1.01,2.1055705261230466
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,16,4,power_law_1.01,0.05759871959686279
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,16,4,power_law_1.01,0.06589695930480957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,16,4,power_law_1.01,0.07393407821655273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,16,4,power_law_1.01,0.0865766429901123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,16,4,power_law_1.01,0.0900607967376709
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,16,4,power_law_1.01,0.09325311660766603
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,16,4,power_law_1.01,0.09481216430664062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,16,4,power_law_1.01,0.10270976066589355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,16,4,power_law_1.01,0.09303296089172364
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,16,4,power_law_1.01,0.0951142406463623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,16,4,power_law_1.01,0.10103039741516114
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,16,4,power_law_1.01,0.10161791801452638
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,16,4,power_law_1.01,0.10510080337524415
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,16,4,balanced,0.10997887611389161
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,16,4,balanced,0.11997183799743652
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,16,4,balanced,0.1701478385925293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,16,4,power_law_1.01,0.1116966438293457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,16,4,power_law_1.01,0.13118720054626465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,16,4,power_law_1.01,0.1508582401275635
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,16,4,power_law_1.01,0.18564096450805664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,16,4,power_law_1.01,0.22215808868408202
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,16,4,balanced,0.2144704055786133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,16,4,power_law_1.01,0.30713600158691406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,16,4,power_law_1.01,0.396940803527832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,16,4,power_law_1.01,0.5140454483032226
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,16,4,power_law_1.01,0.028058879375457764
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,16,4,power_law_1.01,0.032833280563354494
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,16,4,power_law_1.01,0.033679358959198
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,16,4,power_law_1.01,0.7166912078857421
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,16,4,power_law_1.01,0.0372326397895813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,16,4,power_law_1.01,0.043683838844299314
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,16,4,power_law_1.01,0.05123199939727783
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,16,4,power_law_1.01,0.05123712062835693
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,16,4,power_law_1.01,1.0683097839355469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,16,4,power_law_1.01,0.051975679397583005
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,16,4,power_law_1.01,0.05239935874938965
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,16,4,power_law_1.01,0.05410560131072998
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,16,4,power_law_1.01,0.05670911788940429
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,16,4,power_law_1.01,0.05493760108947754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,16,4,power_law_1.01,1.342598419189453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,16,4,power_law_1.01,0.05894527912139893
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,16,4,power_law_1.01,0.06054272174835205
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,16,4,power_law_1.01,0.061041917800903324
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,16,4,power_law_1.01,0.06667776107788086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,16,4,power_law_1.01,0.058277120590209966
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,16,4,balanced,0.28099584579467773
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,16,4,balanced,0.37071231842041014
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,16,4,power_law_1.01,0.08070015907287598
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,16,4,power_law_1.01,0.09522944450378418
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,16,4,power_law_1.01,0.14952192306518555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,16,4,power_law_1.2,0.03465856075286865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,16,4,power_law_1.2,0.05733759880065918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,16,4,power_law_1.2,0.06164735794067383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,16,4,power_law_1.2,0.06486656188964843
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,16,4,power_law_1.01,0.2415872001647949
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,16,4,power_law_1.01,0.2048806381225586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,16,4,power_law_1.2,0.07189248085021974
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,16,4,power_law_1.2,0.0833792018890381
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,16,4,power_law_1.2,0.1354534435272217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,16,4,power_law_1.2,0.139685115814209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,16,4,power_law_1.2,0.14263168334960938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,16,4,power_law_1.2,0.14331007957458497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,16,4,power_law_1.2,0.1469760036468506
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,16,4,power_law_1.2,0.1434598445892334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,16,4,power_law_1.2,0.14664959907531738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,16,4,power_law_1.2,0.15393919944763185
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,16,4,power_law_1.2,0.1571776008605957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,16,4,power_law_1.2,0.16376192092895508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,16,4,power_law_1.2,0.1790540885925293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,16,4,power_law_1.2,0.20422271728515623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,16,4,power_law_1.2,0.23223167419433594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,16,4,power_law_1.2,0.28139392852783207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,16,4,power_law_1.2,0.34512767791748045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,16,4,power_law_1.2,0.4630259323120117
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,16,4,power_law_1.01,0.3529753494262695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,16,4,power_law_1.2,0.6183552169799805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,16,4,power_law_1.2,0.854529266357422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,16,4,power_law_1.2,0.04119423866271973
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,16,4,power_law_1.01,0.44030849456787113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,16,4,power_law_1.2,0.04094719886779785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,16,4,power_law_1.2,0.04812032222747803
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,16,4,power_law_1.2,0.055820798873901366
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,16,4,power_law_1.2,0.0602291202545166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,16,4,power_law_1.2,0.07253632068634033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,16,4,power_law_1.2,1.089486083984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,16,4,power_law_1.2,0.08681856155395508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,16,4,power_law_1.2,0.09071999549865722
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,16,4,power_law_1.2,0.09246335983276367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,16,4,power_law_1.2,0.09547136306762696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,16,4,power_law_1.2,0.09669759750366211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,16,4,power_law_1.2,0.09353983879089356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,16,4,power_law_1.2,1.633667907714844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,16,4,power_law_1.2,0.09656448364257812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,16,4,power_law_1.2,0.10225024223327636
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,16,4,power_law_1.2,0.10187135696411134
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,16,4,power_law_1.2,0.10613375663757325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,16,4,power_law_1.2,0.11636480331420898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,16,4,power_law_1.2,0.13676032066345215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,16,4,power_law_1.2,0.15243519783020018
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,16,4,power_law_1.2,0.18838911056518554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,16,4,power_law_1.2,0.2382348823547363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,16,4,power_law_1.2,0.32877056121826176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,16,4,power_law_1.2,0.4001574325561523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,16,4,power_law_1.2,2.1854092407226564
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,16,4,power_law_1.2,0.02785151958465576
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,16,4,power_law_1.2,0.029295361042022704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,16,4,power_law_1.2,0.5797580718994141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,16,4,power_law_1.2,0.7285555267333985
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,16,4,power_law_1.2,0.0341375994682312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,16,4,power_law_1.2,0.03559168100357056
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,16,4,power_law_1.2,0.03962879896163941
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,16,4,power_law_1.2,0.04973696231842041
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,16,4,power_law_1.2,0.05075712203979492
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,16,4,power_law_1.2,0.05145855903625488
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,16,4,power_law_1.2,0.05369344234466553
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,16,4,power_law_1.2,0.051787519454956056
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,16,4,power_law_1.2,0.055440640449523924
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,16,4,power_law_1.2,0.05547264099121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,16,4,power_law_1.2,0.05895936012268066
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,16,4,power_law_1.2,0.06033023834228516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,16,4,power_law_1.2,0.06541056156158447
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,16,4,power_law_1.2,1.119983367919922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,16,4,power_law_1.2,0.07558015823364259
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,16,4,power_law_1.2,0.12982144355773925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,16,4,power_law_1.2,1.4094309997558594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,16,4,power_law_1.01,0.6565593719482422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,16,4,power_law_1.01,0.790535659790039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,16,4,power_law_1.2,0.06974976062774658
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,16,4,power_law_1.2,0.1134233570098877
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,16,4,power_law_1.2,0.18680192947387694
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,16,4,power_law_1.2,0.09408639907836915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,32,2,balanced,0.03588223934173584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,32,2,balanced,0.03372544050216675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,32,2,balanced,0.0425216007232666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,32,2,balanced,0.06735487937927245
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,32,2,balanced,0.086429443359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,32,2,balanced,0.09867263793945312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,32,2,balanced,0.09906944274902343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,32,2,balanced,0.10035455703735352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,32,2,balanced,0.10184448242187501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,32,2,balanced,0.15805312156677248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,32,2,balanced,0.16100992202758788
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,32,2,balanced,0.1651737594604492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,32,2,balanced,0.16882047653198243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,32,2,balanced,0.1758451271057129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,32,2,balanced,0.17891328811645507
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,32,2,balanced,0.18481664657592772
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,32,2,balanced,0.19866880416870117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,32,2,balanced,0.22251136779785155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,32,2,balanced,0.24892032623291013
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,32,2,balanced,0.3085670471191406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,32,2,balanced,0.3583206558227539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,32,2,balanced,0.4688652801513672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,16,4,power_law_1.2,0.2479782485961914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,32,2,balanced,0.6069196701049805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,32,2,balanced,0.8637875366210939
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,32,2,balanced,1.0908467102050783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,32,2,balanced,1.5997030639648437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,32,2,balanced,0.032497920989990235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,32,2,balanced,0.036618239879608154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,32,2,balanced,0.0448473596572876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,32,2,balanced,0.06124032020568848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,32,2,balanced,0.08702207565307617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,32,2,balanced,2.0713267517089844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,32,2,balanced,0.10503040313720705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,32,2,balanced,0.10135295867919922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,32,2,balanced,0.10193280220031738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,32,2,balanced,0.10333696365356446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,32,2,balanced,0.11851391792297364
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,32,2,balanced,0.12078847885131835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,32,2,balanced,0.12419327735900879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,32,2,balanced,0.12674431800842284
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,32,2,balanced,0.13095808029174805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,32,2,balanced,0.12989695549011232
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,32,2,balanced,0.13238271713256836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,32,2,balanced,0.14034432411193848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,32,2,balanced,0.157128963470459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,32,2,balanced,0.1711334419250488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,32,2,balanced,0.20183423995971678
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,32,2,balanced,0.23173887252807618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,32,2,balanced,0.36213375091552735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,32,2,balanced,0.42113918304443365
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,16,4,power_law_1.2,0.29918848037719725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,32,2,balanced,0.5893324661254883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,32,2,balanced,0.778424301147461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,32,2,balanced,1.0737100982666017
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,16,4,power_law_1.2,0.4851609420776367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,32,2,power_law_1.01,0.03783551931381225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,32,2,power_law_1.01,0.03676287889480591
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,32,2,power_law_1.01,0.06822912216186523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,32,2,power_law_1.01,0.05955711841583252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,32,2,power_law_1.01,0.07622144222259522
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,32,2,power_law_1.01,0.09055871963500976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,32,2,balanced,1.43046142578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,32,2,power_law_1.01,0.09332223892211913
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,32,2,power_law_1.01,0.09660160064697265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,32,2,power_law_1.01,0.09784704208374023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,32,2,power_law_1.01,0.15638784408569334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,32,2,power_law_1.01,0.16067712783813476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,32,2,power_law_1.01,0.1683839988708496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,32,2,power_law_1.01,0.17544191360473632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,32,2,power_law_1.01,0.18000511169433592
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,32,2,power_law_1.01,0.18627071380615234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,32,2,power_law_1.01,0.19889408111572265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,32,2,power_law_1.01,0.25513471603393556
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,32,2,power_law_1.01,0.31099008560180663
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,32,2,power_law_1.01,0.3642828750610351
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,32,2,power_law_1.01,0.16611839294433595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,32,2,power_law_1.01,0.47853950500488285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,32,2,power_law_1.01,0.6169023895263672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,32,2,power_law_1.01,0.8736742401123048
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,32,2,power_law_1.01,0.2280985641479492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,32,2,power_law_1.01,1.1052313232421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,32,2,power_law_1.01,0.03913856029510498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,32,2,power_law_1.01,0.04109824180603027
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,32,2,power_law_1.01,0.04775551795959473
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,32,2,power_law_1.01,1.621702423095703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,32,2,power_law_1.01,0.0641971206665039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,32,2,power_law_1.01,0.07698944091796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,32,2,power_law_1.01,0.09291647911071778
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,32,2,power_law_1.01,0.09258496284484863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,32,2,power_law_1.01,0.09603839874267578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,32,2,power_law_1.01,0.0971673583984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,32,2,power_law_1.01,0.11331199645996093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,32,2,power_law_1.01,0.11520000457763672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,32,2,power_law_1.01,0.12059647560119628
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,32,2,power_law_1.01,2.128889617919922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,32,2,power_law_1.01,0.12281472206115722
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,32,2,power_law_1.01,0.12955519676208496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,32,2,power_law_1.01,0.12767871856689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,32,2,power_law_1.01,0.13172096252441406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,32,2,power_law_1.01,0.14043904304504395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,32,2,power_law_1.01,0.1642099189758301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,32,2,power_law_1.01,0.17965696334838868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,32,2,power_law_1.01,0.22148607254028319
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,32,2,power_law_1.01,0.26569215774536137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,32,2,power_law_1.01,0.36495998382568356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,32,2,power_law_1.01,0.44535934448242187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,32,2,power_law_1.01,0.6060031890869141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,32,2,power_law_1.01,0.7979942321777344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,32,2,power_law_1.2,0.035022079944610596
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,32,2,power_law_1.2,0.053566718101501466
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,32,2,power_law_1.2,0.06000512123107911
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,32,2,power_law_1.01,1.1460800170898438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,16,4,power_law_1.2,0.6926105499267579
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,32,2,power_law_1.2,0.07375103950500488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,32,2,power_law_1.2,0.08941311836242676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,32,2,power_law_1.2,0.09648127555847168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,32,2,power_law_1.2,0.09292927742004395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,32,2,power_law_1.01,1.4454170227050782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,32,2,power_law_1.2,0.09943807601928711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,32,2,power_law_1.2,0.15656064033508302
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,32,2,power_law_1.2,0.16213632583618165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,32,2,power_law_1.2,0.16542848587036132
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,32,2,power_law_1.2,0.17674751281738282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,32,2,power_law_1.2,0.18118911743164062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,32,2,power_law_1.2,0.1866048049926758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,32,2,power_law_1.2,0.06610688209533691
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,32,2,power_law_1.2,0.20234495162963867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,32,2,power_law_1.2,0.22917760848999022
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,32,2,power_law_1.2,0.2553651237487793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,32,2,power_law_1.2,0.3144140815734863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,32,2,power_law_1.2,0.36704895019531253
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,32,2,power_law_1.2,0.4819430541992188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,32,2,power_law_1.2,0.6383385467529297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,16,4,power_law_1.2,1.1605951690673828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,32,2,power_law_1.2,0.03578111886978149
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,32,2,power_law_1.2,0.03947776079177857
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,32,2,power_law_1.2,0.8825894165039063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,32,2,power_law_1.2,0.16971519470214844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,32,2,power_law_1.2,0.04978303909301758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,32,2,power_law_1.2,0.06265471935272217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,32,2,power_law_1.2,0.07280511856079101
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,32,2,power_law_1.2,0.0871833610534668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,32,2,power_law_1.2,0.0913548755645752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,32,2,power_law_1.2,0.09447296142578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,32,2,power_law_1.2,0.09857151985168458
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,32,2,power_law_1.2,0.11306112289428712
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,32,2,power_law_1.2,0.11537919998168947
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,32,2,power_law_1.2,0.1208409595489502
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,32,2,power_law_1.2,0.12353280067443848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,32,2,power_law_1.2,1.1213568115234376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,32,2,power_law_1.2,0.12914175987243653
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,32,2,power_law_1.2,0.1290342426300049
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,32,2,power_law_1.2,0.13314047813415528
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,32,2,power_law_1.2,0.1474880027770996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,32,2,power_law_1.2,0.16509056091308594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,32,2,power_law_1.2,0.18634239196777344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,32,2,power_law_1.2,0.22179712295532225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,32,2,power_law_1.2,0.2704089546203613
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,32,2,power_law_1.2,0.3691455841064453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,32,2,power_law_1.2,0.4543487930297852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,32,2,power_law_1.2,0.6187148666381835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,32,2,power_law_1.2,0.8054425811767578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,32,2,power_law_1.2,2.160364837646484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,64,balanced,0.06846464157104493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,64,balanced,0.06878464221954346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,64,balanced,0.07850751876831055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,32,2,power_law_1.2,1.6389132690429686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,64,balanced,0.09530112266540527
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,64,balanced,0.06896255970001221
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,64,balanced,0.0672883176803589
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,64,balanced,0.11958527565002441
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,64,balanced,0.1670400047302246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,64,balanced,0.17420288085937502
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,64,balanced,0.1754150390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,64,balanced,0.18068351745605468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,32,2,power_law_1.2,1.480334014892578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,64,balanced,0.17819904327392577
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,32,2,power_law_1.2,1.1700057220458984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,64,balanced,0.18239871978759764
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,64,balanced,0.17940479278564453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,64,balanced,0.18504064559936523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,64,balanced,0.16649728775024414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,64,balanced,0.1942848014831543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,64,balanced,0.19411327362060546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,64,balanced,0.16637439727783204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,64,balanced,0.20989183425903318
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,64,balanced,0.22254335403442385
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,64,balanced,0.23278079986572267
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,64,balanced,0.29559167861938473
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,64,balanced,0.34853118896484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,64,balanced,0.03904128074645996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,64,balanced,0.46641918182373043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,64,balanced,0.03828736066818238
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,64,balanced,0.03875839948654175
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,64,balanced,0.5622732925415039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,64,balanced,0.041356801986694336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,64,balanced,0.047325439453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,16,4,power_law_1.2,1.4048870849609376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,64,balanced,0.05991551876068115
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,64,balanced,0.753222427368164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,64,balanced,0.07673344135284424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,64,balanced,0.08156928062438965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,64,balanced,0.08253824234008789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,64,balanced,0.0858176040649414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,64,balanced,0.09423616409301758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,64,balanced,0.08554368019104004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,64,balanced,0.08733568191528321
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,64,balanced,0.08906111717224122
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,64,balanced,0.09239680290222167
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,64,balanced,0.09756799697875976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,64,balanced,0.09109760284423828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,64,balanced,0.09769599914550782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,64,balanced,0.10213248252868652
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,64,balanced,0.115480318069458
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,64,balanced,0.12880767822265624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,64,balanced,0.15745920181274414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,64,balanced,0.21303167343139648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,64,balanced,0.25964544296264647
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,64,balanced,0.32524929046630857
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,64,balanced,0.08172672271728516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,64,balanced,0.08617600440979004
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,1,64,balanced,0.03393663883209229
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,1,64,balanced,0.021067519187927247
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,1,64,balanced,0.0476697587966919
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,1,64,balanced,0.04778880119323731
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,1,64,balanced,0.048079361915588384
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,1,64,balanced,0.04788095951080322
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,1,64,balanced,0.04793471813201904
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,1,64,balanced,0.0477785587310791
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,1,64,balanced,0.02106112003326416
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,1,64,balanced,0.04805759906768799
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,1,64,balanced,0.048195838928222656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,1,64,balanced,0.04819327831268311
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,1,64,balanced,0.047697920799255375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,1,64,balanced,0.048195838928222656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,1,64,balanced,0.048035840988159176
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,1,64,balanced,0.049286398887634274
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,1,64,balanced,0.04919424057006836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,1,64,balanced,0.05091839790344238
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,1,64,balanced,0.021374719142913817
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,1,64,balanced,0.051280641555786134
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,1,64,balanced,0.05296256065368652
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,64,power_law_1.01,0.11426431655883788
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,64,power_law_1.01,0.11455871582031249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,64,power_law_1.01,0.11681407928466796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,64,power_law_1.01,0.11936767578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,1,64,balanced,0.02187648057937622
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,64,power_law_1.01,0.16177280426025392
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,64,power_law_1.01,0.15882368087768556
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,1,64,balanced,0.021056001186370847
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,64,power_law_1.01,0.1637708854675293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,64,power_law_1.01,0.16024703979492189
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,64,power_law_1.01,0.17291135787963868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,64,power_law_1.01,0.17611007690429686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,64,power_law_1.01,0.1736307144165039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,64,power_law_1.01,0.17932672500610353
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,64,power_law_1.01,0.18243200302124024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,64,power_law_1.01,0.18621952056884766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,64,power_law_1.01,0.18915199279785155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,64,power_law_1.01,0.2017292785644531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,64,power_law_1.01,0.22062463760375978
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,64,power_law_1.01,0.25265535354614255
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,64,power_law_1.01,0.17187967300415039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,64,power_law_1.01,0.30201599121093753
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,64,power_law_1.01,0.3670463943481445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,64,power_law_1.01,0.4947558212280273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,64,power_law_1.01,0.5929484939575195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,64,power_law_1.01,0.8714790344238281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,64,power_law_1.01,0.07466879844665528
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,64,power_law_1.01,1.142946548461914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,64,power_law_1.01,0.07508607864379882
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,64,power_law_1.01,0.0768230390548706
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,64,power_law_1.01,0.07506432056427001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,64,power_law_1.01,1.4643699645996093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,64,power_law_1.01,0.08157695770263672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,64,power_law_1.01,0.07220223903656006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,64,power_law_1.01,0.0686195182800293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,64,power_law_1.01,0.07586175918579101
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,64,power_law_1.01,2.144085693359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,64,power_law_1.01,0.07706111907958985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,64,power_law_1.01,0.08056960105895997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,64,power_law_1.01,0.07982975959777831
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,64,power_law_1.01,0.08033791542053223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,64,power_law_1.01,0.0916006374359131
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,64,power_law_1.01,0.09701888084411621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,64,power_law_1.01,0.09521280288696289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,64,power_law_1.01,0.10396544456481933
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,64,power_law_1.01,0.08509311676025391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,64,power_law_1.01,0.11400447845458986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,64,power_law_1.01,0.13154687881469726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,64,power_law_1.01,0.15114496231079103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,64,power_law_1.01,0.2062886428833008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,64,power_law_1.01,0.2591948890686035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,64,power_law_1.01,0.36593536376953123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,64,power_law_1.01,0.4873062515258789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,64,power_law_1.01,0.09063039779663086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,64,power_law_1.01,0.6354367828369141
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,1,64,power_law_1.01,0.04749951839447021
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,1,64,power_law_1.01,0.047482881546020504
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,1,64,power_law_1.01,0.04770431995391845
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,1,64,power_law_1.01,0.047680001258850094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,1,64,power_law_1.01,0.042536959648132325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,64,power_law_1.01,0.9073881530761719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,1,64,power_law_1.01,0.04460800170898437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,1,64,power_law_1.01,0.0443071985244751
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,1,64,power_law_1.01,0.04683008193969727
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,1,64,power_law_1.01,0.04699903964996338
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,1,64,power_law_1.01,0.04568064212799072
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,1,64,power_law_1.01,0.04725503921508789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,1,64,power_law_1.01,0.0470681619644165
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,1,64,power_law_1.01,0.0477785587310791
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,1,64,power_law_1.01,0.0470527982711792
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,1,64,power_law_1.01,0.04762239933013916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,1,64,power_law_1.01,0.04806272029876709
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,1,64,power_law_1.01,0.04853631973266602
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,1,64,power_law_1.01,0.04844543933868408
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,1,64,balanced,0.06604544162750245
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,1,64,balanced,0.06969471931457519
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,1,64,balanced,0.046579198837280275
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,1,64,balanced,0.08662143707275391
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,1,64,power_law_1.01,0.05099775791168213
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,1,64,power_law_1.01,0.059438080787658686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,64,power_law_1.2,0.11414527893066406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,64,power_law_1.2,0.11431679725646973
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,1,64,power_law_1.01,0.05369472026824951
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,64,power_law_1.2,0.11765376091003418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,64,power_law_1.2,0.12036735534667968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,64,power_law_1.2,0.16460927963256836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,1,64,power_law_1.01,0.07429887771606444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,64,power_law_1.2,0.15725055694580078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,64,power_law_1.2,0.1618150329589844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,64,power_law_1.2,0.1628326416015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,64,power_law_1.2,0.16649343490600585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,64,power_law_1.2,0.17282175064086913
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,64,power_law_1.2,0.17184640884399413
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,64,power_law_1.2,0.1776896095275879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,64,power_law_1.2,0.17989248275756836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,64,power_law_1.2,0.18683391571044922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,64,power_law_1.2,0.1953446388244629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,64,power_law_1.2,0.19708032608032228
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,64,power_law_1.2,0.21525760650634768
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,64,power_law_1.2,0.2439219284057617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,64,power_law_1.2,0.2821286392211914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,64,power_law_1.2,0.37724033355712894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,64,power_law_1.2,0.4431500625610352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,64,power_law_1.2,0.5738240051269531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,64,power_law_1.2,0.7180147552490235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,64,power_law_1.2,1.0922598266601562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,64,power_law_1.2,1.5775885009765624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,64,power_law_1.2,0.07451648235321044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,64,power_law_1.2,0.07563648223876954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,64,power_law_1.2,0.07714943885803223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,64,power_law_1.2,2.3259007263183595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,64,power_law_1.2,0.07862016201019287
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,64,power_law_1.2,0.07747712135314941
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,64,power_law_1.2,0.07006336212158203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,64,power_law_1.2,0.07174015998840332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,64,power_law_1.2,0.08248831748962401
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,64,power_law_1.2,0.07640448093414307
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,64,power_law_1.2,3.282762145996094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,64,power_law_1.2,0.08645759582519531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,64,power_law_1.2,0.07680255889892577
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,64,power_law_1.2,0.0828940773010254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,64,power_law_1.2,0.08063488006591797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,64,power_law_1.2,0.08889087677001953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,64,power_law_1.2,0.09363327980041504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,64,power_law_1.2,0.09439616203308106
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,64,power_law_1.2,0.09757056236267089
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,64,power_law_1.2,0.10906880378723145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,64,power_law_1.2,0.12290431976318358
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,64,power_law_1.2,0.1520089626312256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,64,power_law_1.2,0.17590656280517578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,64,power_law_1.2,0.24081792831420898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,64,power_law_1.2,0.2910016059875488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,64,power_law_1.2,0.47965824127197265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,64,power_law_1.2,0.6377561569213868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,64,power_law_1.2,0.9651455688476563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,1,64,power_law_1.2,0.04756735801696778
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,1,64,power_law_1.2,0.04781824111938476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,64,power_law_1.2,1.3195724487304688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,1,64,power_law_1.2,0.047607040405273436
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,1,64,power_law_1.2,0.04803199768066406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,1,64,power_law_1.2,0.04763264179229736
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,1,64,power_law_1.2,0.04386303901672363
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,1,64,power_law_1.2,0.04353536128997802
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,1,64,power_law_1.2,0.04533631801605224
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,1,64,power_law_1.2,0.04723199844360352
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,1,64,power_law_1.2,0.04707327842712402
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,1,64,power_law_1.2,0.046316800117492674
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,1,64,power_law_1.2,0.04724480152130127
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,1,64,power_law_1.2,0.047562241554260254
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,1,64,power_law_1.2,0.04450687885284424
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,1,64,power_law_1.2,0.047540478706359864
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,1,64,power_law_1.2,0.0474406385421753
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,1,64,power_law_1.2,0.05052288055419922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,1,64,power_law_1.2,0.04762495994567871
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,1,64,power_law_1.01,0.11480319976806641
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,1,64,power_law_1.01,0.08511360168457031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,1,64,power_law_1.01,0.19489408493041993
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,1,64,power_law_1.01,0.14145407676696778
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,1,64,power_law_1.01,0.2511180877685547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,1,64,power_law_1.2,0.05263360023498535
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,1,64,power_law_1.2,0.0697318410873413
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,32,balanced,0.03783040046691895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,32,balanced,0.03820672035217285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,32,balanced,0.03778815984725952
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,32,balanced,0.06260096073150635
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,1,64,power_law_1.2,0.05963007926940918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,32,balanced,0.0748364782333374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,32,balanced,0.09842687606811523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,32,balanced,0.12355968475341797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,32,balanced,0.17299583435058594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,32,balanced,0.17857791900634765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,32,balanced,0.1735820770263672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,32,balanced,0.17324159622192384
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,32,balanced,0.1786419105529785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,32,balanced,0.17600128173828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,32,balanced,0.1830681610107422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,32,balanced,0.18745855331420896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,32,balanced,0.19174528121948242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,32,balanced,0.18544000625610352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,32,balanced,0.19370880126953124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,32,balanced,0.20131711959838866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,32,balanced,0.21887231826782227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,32,balanced,0.236812801361084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,32,balanced,0.27064832687377927
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,1,64,power_law_1.2,0.08945664405822754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,32,balanced,0.32866046905517576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,32,balanced,0.3906419372558594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,1,64,power_law_1.2,0.11440640449523927
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,32,balanced,0.5605619049072266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,32,balanced,0.035791358947753905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,32,balanced,0.7103910064697265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,32,balanced,0.03536256074905396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,32,balanced,0.03752063989639282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,32,balanced,0.0447705602645874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,32,balanced,0.06460671901702882
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,32,balanced,0.9590707397460937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,32,balanced,0.07746175765991212
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,32,balanced,0.09389439582824707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,32,balanced,0.08798336029052735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,32,balanced,0.08593279838562011
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,32,balanced,0.08630016326904297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,32,balanced,0.08828800201416016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,32,balanced,0.0889254379272461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,32,balanced,0.09438336372375489
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,32,balanced,0.09455231666564942
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,32,balanced,0.09797887802124024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,32,balanced,0.03568000078201294
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,1,64,power_law_1.2,0.1882636833190918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,32,balanced,0.09194368362426758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,32,balanced,0.09750656127929688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,1,64,power_law_1.2,0.1550374412536621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,32,balanced,0.09963647842407226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,32,balanced,0.1112012767791748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,32,balanced,0.14598400115966798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,32,balanced,0.16892799377441406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,32,balanced,0.2084992027282715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,32,balanced,0.2857804870605469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,32,balanced,0.3667264175415039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,32,balanced,0.46928127288818355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,32,balanced,0.12454272270202638
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,1,64,power_law_1.2,0.30153343200683597
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,1,64,power_law_1.2,0.40124160766601563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,2,32,balanced,0.021274878978729247
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,2,32,balanced,0.02109312057495117
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,2,32,balanced,0.04870143890380859
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,2,32,balanced,0.0486297607421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,2,32,balanced,0.04929152011871338
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,2,32,balanced,0.04972799777984619
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,2,32,balanced,0.021062400341033936
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,2,32,balanced,0.049767680168151855
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,2,32,balanced,0.049326081275939945
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,2,32,balanced,0.020997118949890137
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,2,32,balanced,0.049546241760253906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,2,32,balanced,0.05019775867462158
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,2,32,balanced,0.04973440170288086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,2,32,balanced,0.04987135887145996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,2,32,balanced,0.049862399101257324
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,2,32,balanced,0.05132927894592285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,2,32,balanced,0.05223552227020264
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,2,32,balanced,0.05414400100708008
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,2,32,balanced,0.05606783866882324
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,2,32,balanced,0.058620162010192864
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,2,32,balanced,0.024944639205932616
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,32,power_law_1.01,0.09692416191101075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,32,power_law_1.01,0.11837056159973144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,32,power_law_1.01,0.1191104030609131
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,32,power_law_1.01,0.12223872184753418
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,2,32,balanced,0.034517760276794436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,32,power_law_1.01,0.17091840744018555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,32,power_law_1.01,0.1623731231689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,32,power_law_1.01,0.16273408889770508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,32,power_law_1.01,0.1742297554016113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,32,power_law_1.01,0.16339071273803712
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,32,power_law_1.01,0.16574335098266602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,32,power_law_1.01,0.1714137649536133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,32,power_law_1.01,0.179366397857666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,32,power_law_1.01,0.18731008529663087
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,32,power_law_1.01,0.18431615829467773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,32,power_law_1.01,0.18183040618896484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,32,power_law_1.01,0.19237632751464845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,32,power_law_1.01,0.20052608489990234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,32,power_law_1.01,0.23483135223388674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,32,power_law_1.01,0.2642931175231934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,32,power_law_1.01,0.32630401611328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,32,power_law_1.01,0.4531110382080078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,32,power_law_1.01,0.5490803146362305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,32,power_law_1.01,0.7905702209472656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,2,32,balanced,0.04842112064361572
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,32,power_law_1.01,0.9763340759277345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,32,power_law_1.01,0.060577278137207036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,32,power_law_1.01,0.07479040145874023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,32,power_law_1.01,1.5360357666015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,32,power_law_1.01,0.07629695892333985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,32,power_law_1.01,0.07588736057281495
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,32,power_law_1.01,0.08094464302062988
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,32,power_law_1.01,0.08158335685729981
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,32,power_law_1.01,0.08062335968017578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,32,power_law_1.01,0.08197759628295899
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,32,power_law_1.01,0.08927743911743165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,32,power_law_1.01,0.08185471534729004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,32,power_law_1.01,1.8269938659667968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,32,power_law_1.01,0.08472576141357421
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,32,power_law_1.01,0.08609279632568359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,32,power_law_1.01,0.09666943550109863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,32,power_law_1.01,0.08939647674560547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,32,power_law_1.01,0.08991104125976562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,32,power_law_1.01,0.09784064292907715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,32,power_law_1.01,0.09935104370117187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,32,power_law_1.01,0.11248127937316894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,32,power_law_1.01,0.14182016372680664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,32,power_law_1.01,0.16303487777709963
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,32,power_law_1.01,0.08856448173522949
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,32,power_law_1.01,0.22057600021362306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,32,power_law_1.01,0.25717376708984374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,32,power_law_1.01,0.40084991455078123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,32,power_law_1.01,0.46597759246826176
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,2,32,power_law_1.01,0.036468479633331295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,32,power_law_1.01,0.6391820907592773
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,2,32,power_law_1.01,0.04875648021697998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,32,power_law_1.01,0.16249343872070313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,2,32,power_law_1.01,0.0485427188873291
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,2,32,power_law_1.01,0.04860544204711914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,2,32,power_law_1.01,0.04863359928131104
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,2,32,power_law_1.01,0.044432640075683594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,2,32,power_law_1.01,0.040166401863098146
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,2,32,power_law_1.01,0.044326400756835936
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,2,32,power_law_1.01,0.04426496028900147
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,2,32,power_law_1.01,0.04443391799926758
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,2,32,power_law_1.01,0.04678143978118897
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,2,32,power_law_1.01,0.04679423809051514
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,2,32,power_law_1.01,0.046856961250305175
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,2,32,power_law_1.01,0.048611841201782226
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,2,32,power_law_1.01,0.04712831974029541
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,2,32,power_law_1.01,0.04928256034851074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,32,power_law_1.01,0.8761433410644532
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,2,32,power_law_1.01,0.04837247848510742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,2,32,power_law_1.01,0.049340162277221676
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,2,32,balanced,0.048808960914611815
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,2,32,balanced,0.08261247634887695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,2,32,balanced,0.07617919921875001
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,2,32,balanced,0.10358271598815919
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,2,32,power_law_1.01,0.06328576087951661
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,2,32,power_law_1.01,0.05414144039154053
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,2,32,power_law_1.01,0.0576204776763916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,2,32,power_law_1.01,0.07768703937530518
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,32,power_law_1.2,0.09642880439758302
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,32,power_law_1.2,0.11753215789794921
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,32,power_law_1.2,0.12217087745666504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,32,power_law_1.2,0.11652223587036133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,32,power_law_1.2,0.16545024871826172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,32,power_law_1.2,0.16178304672241212
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,32,power_law_1.2,0.15264512062072755
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,32,power_law_1.2,0.15979007720947264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,32,power_law_1.2,0.16326143264770507
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,32,power_law_1.2,0.16807552337646486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,32,power_law_1.2,0.17236223220825195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,32,power_law_1.2,0.17832960128784178
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,32,power_law_1.2,0.17332223892211912
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,32,power_law_1.2,0.18371583938598632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,32,power_law_1.2,0.18772863388061523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,32,power_law_1.2,0.1947929573059082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,32,power_law_1.2,0.17991424560546876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,32,power_law_1.2,0.22107648849487305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,32,power_law_1.2,0.24494976043701172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,32,power_law_1.2,0.31126399993896486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,32,power_law_1.2,0.4066022491455078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,32,power_law_1.2,0.5469350433349609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,32,power_law_1.2,0.6192204666137695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,32,power_law_1.2,0.9229631805419922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,32,power_law_1.2,1.254814682006836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,32,power_law_1.2,0.059837441444396976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,32,power_law_1.2,0.07430399894714355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,32,power_law_1.2,1.8377357482910157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,32,power_law_1.2,0.07487616062164307
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,32,power_law_1.2,0.07605760097503662
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,32,power_law_1.2,0.08220416069030761
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,32,power_law_1.2,0.07715968132019044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,32,power_law_1.2,2.3086834716796876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,2,32,power_law_1.01,0.09211392402648926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,32,power_law_1.2,0.07254144191741943
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,32,power_law_1.2,0.0813708782196045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,32,power_law_1.2,0.08842368125915527
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,32,power_law_1.2,0.08249600410461426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,32,power_law_1.2,0.08034432411193848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,32,power_law_1.2,0.08600319862365723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,32,power_law_1.2,0.08726400375366211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,32,power_law_1.2,0.09609343528747558
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,32,power_law_1.2,0.09251456260681153
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,32,power_law_1.2,0.09067647933959962
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,32,power_law_1.2,0.0963584041595459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,32,power_law_1.2,0.10534272193908692
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,32,power_law_1.2,0.12210687637329101
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,32,power_law_1.2,0.14923775672912598
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,32,power_law_1.2,0.18052223205566406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,32,power_law_1.2,0.22824832916259769
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,32,power_law_1.2,0.3274739074707031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,32,power_law_1.2,0.4185087966918945
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,2,32,power_law_1.2,0.03626496076583862
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,32,power_law_1.2,0.6154662322998047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,2,32,power_law_1.2,0.04871039867401123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,32,power_law_1.2,0.8137075042724609
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,2,32,power_law_1.2,0.048706560134887694
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,2,32,power_law_1.2,0.04824063777923584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,2,32,power_law_1.2,0.04899968147277832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,2,32,power_law_1.2,0.03579648017883301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,32,power_law_1.2,1.1094195556640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,2,32,power_law_1.2,0.03478271961212158
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,2,32,power_law_1.2,0.04393983840942382
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,2,32,power_law_1.2,0.04452991962432861
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,2,32,power_law_1.2,0.04717440128326416
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,2,32,power_law_1.2,0.0440883207321167
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,2,32,power_law_1.2,0.043979520797729495
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,2,32,power_law_1.2,0.0440934419631958
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,2,32,power_law_1.2,0.04757760047912597
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,2,32,power_law_1.2,0.04679296016693115
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,2,32,power_law_1.2,0.04934783935546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,2,32,power_law_1.01,0.11800319671630859
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,2,32,power_law_1.2,0.04717951774597168
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,2,32,power_law_1.01,0.14222975730895998
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,2,32,power_law_1.01,0.1972275161743164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,2,32,power_law_1.01,0.24863616943359376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,2,32,power_law_1.2,0.20290048599243163
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,2,32,power_law_1.2,0.05361536026000977
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,2,32,power_law_1.2,0.0629747200012207
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,2,32,power_law_1.2,0.056801280975341796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,4,16,balanced,0.03617151975631714
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,16,balanced,0.03337471961975098
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,16,balanced,0.03888767957687378
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,16,balanced,0.061666560173034665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,16,balanced,0.0785971212387085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,16,balanced,0.09614591598510742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,16,balanced,0.12906880378723146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,16,balanced,0.1675276756286621
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,2,32,power_law_1.2,0.07332992076873779
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,16,balanced,0.1713395118713379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,16,balanced,0.1705638313293457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,16,balanced,0.17668479919433594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,16,balanced,0.18166271209716797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,16,balanced,0.18492544174194336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,16,balanced,0.1791334342956543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,16,balanced,0.18439424514770508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,16,balanced,0.18768255233764647
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,16,balanced,0.19296127319335937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,16,balanced,0.20951808929443358
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,16,balanced,0.2252992057800293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,16,balanced,0.25397504806518556
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,16,balanced,0.287321605682373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,16,balanced,0.34787456512451176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,16,balanced,0.427454719543457
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,2,32,power_law_1.2,0.11687680244445801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,16,balanced,0.5481523132324219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,2,32,power_law_1.2,0.09383296012878418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,16,balanced,0.03495424032211304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,16,balanced,0.034979839324951176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,16,balanced,0.7597325134277344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,16,balanced,0.037849600315094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,16,balanced,0.04243328094482422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,16,balanced,0.05521535873413086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,2,32,power_law_1.2,0.1490764808654785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,16,balanced,1.0319308471679687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,16,balanced,0.06897664070129395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,16,balanced,0.09537792205810547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,16,balanced,0.09605119705200196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,16,balanced,0.0976473617553711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,16,balanced,0.09714176177978515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,16,balanced,0.10099200248718261
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,16,balanced,0.10485631942749023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,16,balanced,0.10896767616271971
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,16,balanced,0.09980799674987792
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,16,balanced,1.3715225219726563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,16,balanced,0.10330240249633789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,16,balanced,0.10422911643981933
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,16,balanced,0.10766336441040039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,16,balanced,0.11704959869384765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,16,balanced,0.1294976043701172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,16,balanced,0.14822400093078614
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,16,balanced,0.1688934326171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,16,balanced,0.20462207794189452
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,16,balanced,0.24487295150756835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,16,balanced,0.3195891189575195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,16,balanced,0.43803775787353516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,16,balanced,0.5794944000244141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,16,balanced,0.7558003234863281
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,2,32,power_law_1.2,0.29632383346557617
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,2,32,power_law_1.2,0.39139713287353517
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,4,16,balanced,0.022364161014556884
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,4,16,balanced,0.0558835220336914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,4,16,balanced,0.05566207885742187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,4,16,balanced,0.022720000743865966
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,4,16,balanced,0.0561356782913208
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,4,16,balanced,0.05658495903015137
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,4,16,balanced,0.0567961597442627
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,4,16,balanced,0.056652798652648925
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,4,16,balanced,0.022328319549560546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,4,16,balanced,0.05684864044189453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,4,16,balanced,0.05635839939117432
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,4,16,balanced,0.028003840446472167
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,4,16,balanced,0.05746431827545166
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,4,16,balanced,0.059230718612670895
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,4,16,balanced,0.06023168087005615
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,4,16,balanced,0.061966080665588376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,4,16,balanced,0.06514560222625733
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,4,16,balanced,0.06882431983947754
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,4,16,balanced,0.07403520107269287
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,4,16,balanced,0.02462719917297363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,4,16,power_law_1.01,0.07453440189361572
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,4,16,balanced,0.04119040012359619
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,16,power_law_1.01,0.09356672286987304
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,4,16,balanced,0.05712512016296387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,16,power_law_1.01,0.11440511703491212
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,16,power_law_1.01,0.11438591957092285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,16,power_law_1.01,0.11737983703613282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,16,power_law_1.01,0.11996288299560547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,16,power_law_1.01,0.15927040100097656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,16,power_law_1.01,0.15484543800354006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,16,power_law_1.01,0.1571993637084961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,16,power_law_1.01,0.16354175567626955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,16,power_law_1.01,0.16533119201660157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,16,power_law_1.01,0.17202816009521485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,16,power_law_1.01,0.18076927185058594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,16,power_law_1.01,0.17192703247070312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,16,power_law_1.01,0.18006528854370116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,16,power_law_1.01,0.18324607849121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,16,power_law_1.01,0.1885683250427246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,16,power_law_1.01,0.21411712646484374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,16,power_law_1.01,0.23848960876464842
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,16,power_law_1.01,0.2762713623046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,16,power_law_1.01,0.3322060775756836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,16,power_law_1.01,0.43073406219482424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,16,power_law_1.01,0.604592628479004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,16,power_law_1.01,0.7847500610351562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,16,power_law_1.01,1.0235084533691405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,16,power_law_1.01,0.04571135997772217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,16,power_law_1.01,1.4524032592773437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,16,power_law_1.01,0.06632319927215577
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,16,power_law_1.01,0.08157183647155761
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,16,power_law_1.01,0.08040831565856935
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,16,power_law_1.01,0.08619263648986816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,16,power_law_1.01,1.8296293640136718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,16,power_law_1.01,0.09481216430664062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,16,power_law_1.01,0.08592896461486817
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,16,power_law_1.01,0.08895232200622558
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,16,power_law_1.01,0.09060352325439454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,16,power_law_1.01,0.0954470443725586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,16,power_law_1.01,0.09791104316711426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,16,power_law_1.01,0.09320704460144043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,16,power_law_1.01,0.09528448104858397
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,16,power_law_1.01,0.10121855735778809
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,16,power_law_1.01,0.10136960029602052
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,16,power_law_1.01,0.10423040390014648
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,4,16,balanced,0.05549824237823486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,16,power_law_1.01,0.11585280418395996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,16,power_law_1.01,0.13170047760009768
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,16,power_law_1.01,0.16058111190795898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,16,power_law_1.01,0.181209602355957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,16,power_law_1.01,0.23535743713378907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,16,power_law_1.01,0.1017574405670166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,16,power_law_1.01,0.31058559417724607
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,16,power_law_1.01,0.4218393707275391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,16,power_law_1.01,0.510190086364746
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,4,16,power_law_1.01,0.027023360729217526
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,16,power_law_1.01,0.7143462371826172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,4,16,power_law_1.01,0.04240255832672119
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,4,16,power_law_1.01,0.05165952205657959
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,4,16,power_law_1.01,0.0554201602935791
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,16,power_law_1.01,0.98906494140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,4,16,power_law_1.01,0.05540224075317383
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,4,16,power_law_1.01,0.05499392032623292
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,4,16,power_law_1.01,0.0526310396194458
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,4,16,power_law_1.01,0.05331711769104004
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,4,16,power_law_1.01,0.05049088001251221
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,4,16,power_law_1.01,0.05289984226226807
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,4,16,power_law_1.01,0.05085184097290039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,4,16,power_law_1.01,0.054448637962341305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,4,16,power_law_1.01,0.05521152019500732
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,4,16,power_law_1.01,0.05569024085998535
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,4,16,power_law_1.01,0.05334527969360352
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,4,16,power_law_1.01,0.05347072124481202
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,4,16,power_law_1.01,0.057066240310668946
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,4,16,power_law_1.01,0.060267519950866696
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,4,16,balanced,0.05468416213989258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,4,16,balanced,0.09688447952270508
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,4,16,balanced,0.1405452823638916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,4,16,balanced,0.10904191970825194
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,4,16,power_law_1.01,0.06491648197174073
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,4,16,power_law_1.01,0.08045056343078613
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,4,16,power_law_1.01,0.0710591983795166
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,4,16,power_law_1.01,0.09987968444824219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,4,16,power_law_1.2,0.07449984073638916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,16,power_law_1.2,0.09391615867614747
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,16,power_law_1.2,0.1087667179107666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,16,power_law_1.2,0.1129702377319336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,16,power_law_1.2,0.12002176284790038
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,16,power_law_1.2,0.16336639404296877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,16,power_law_1.2,0.15683199882507323
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,16,power_law_1.2,0.15262847900390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,16,power_law_1.2,0.15675264358520508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,16,power_law_1.2,0.1588684844970703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,16,power_law_1.2,0.1678553581237793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,16,power_law_1.2,0.1698828887939453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,16,power_law_1.2,0.1757734489440918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,16,power_law_1.2,0.17557504653930664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,16,power_law_1.2,0.18569215774536132
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,16,power_law_1.2,0.18537599563598633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,16,power_law_1.2,0.2052083206176758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,16,power_law_1.2,0.21710847854614257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,16,power_law_1.2,0.25291135787963864
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,16,power_law_1.2,0.295413761138916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,16,power_law_1.2,0.3694630432128906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,16,power_law_1.2,0.5248371124267578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,16,power_law_1.2,0.6193318557739258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,16,power_law_1.2,0.8576396942138672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,16,power_law_1.2,1.160705261230469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,16,power_law_1.2,0.04334591865539551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,16,power_law_1.2,0.06659711837768555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,16,power_law_1.2,1.8524263000488284
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,16,power_law_1.2,0.08065024375915528
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,16,power_law_1.2,0.0815436840057373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,16,power_law_1.2,0.08402048110961914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,16,power_law_1.2,0.09767168045043946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,16,power_law_1.2,0.08146944046020507
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,16,power_law_1.2,2.5996786499023434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,16,power_law_1.2,0.08548864364624023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,16,power_law_1.2,0.09110527992248535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,16,power_law_1.2,0.09291135787963867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,16,power_law_1.2,0.0887667179107666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,16,power_law_1.2,0.09996288299560548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,16,power_law_1.2,0.10348031997680665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,16,power_law_1.2,0.09736960411071778
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,16,power_law_1.2,0.1019315242767334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,16,power_law_1.2,0.10811008453369139
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,16,power_law_1.2,0.12317440032958984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,16,power_law_1.2,0.13766528129577638
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,4,16,power_law_1.01,0.11404416084289551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,16,power_law_1.2,0.17281919479370117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,16,power_law_1.2,0.20075008392333987
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,16,power_law_1.2,0.26919551849365236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,16,power_law_1.2,0.10295807838439941
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,16,power_law_1.2,0.3540300750732422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,16,power_law_1.2,0.45296897888183596
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,4,16,power_law_1.2,0.027194879055023193
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,16,power_law_1.2,0.6481497955322266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,4,16,power_law_1.2,0.04224127769470215
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,4,16,power_law_1.2,0.05484543800354004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,16,power_law_1.2,0.9441843414306641
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,4,16,power_law_1.2,0.05274367809295655
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,4,16,power_law_1.2,0.055404801368713376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,4,16,power_law_1.2,0.05592063903808594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,16,power_law_1.2,1.2357965087890626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,4,16,power_law_1.2,0.048565759658813476
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,4,16,power_law_1.2,0.0531276798248291
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,4,16,power_law_1.2,0.05340415954589843
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,4,16,power_law_1.2,0.05306496143341064
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,4,16,power_law_1.2,0.047642879486083985
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,4,16,power_law_1.2,0.05394176006317138
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,4,16,power_law_1.2,0.052876801490783686
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,4,16,power_law_1.2,0.05290239810943603
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,4,16,power_law_1.2,0.05581696033477783
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,4,16,power_law_1.2,0.057324800491333004
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,4,16,power_law_1.2,0.05909887790679932
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,4,16,power_law_1.01,0.14306816101074218
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,4,16,power_law_1.01,0.17785472869873048
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,4,16,power_law_1.01,0.26234367370605466
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,4,16,power_law_1.01,0.33836673736572265
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,4,16,power_law_1.2,0.06406527996063233
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,4,16,power_law_1.2,0.06815360069274903
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,4,16,power_law_1.2,0.07874688148498535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,8,8,balanced,0.03914880037307739
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,8,8,balanced,0.03545727968215943
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,8,8,balanced,0.047331838607788085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,8,8,balanced,0.06754687786102295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,8,8,balanced,0.0810700798034668
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,4,16,power_law_1.2,0.09305472373962402
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,8,8,balanced,0.10520832061767578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,8,8,balanced,0.12766079902648925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,8,8,balanced,0.1815488052368164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,4,16,power_law_1.2,0.12016639709472657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,8,8,balanced,0.18485887527465822
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,8,8,balanced,0.18717311859130858
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,8,8,balanced,0.18223615646362307
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,8,8,balanced,0.18749696731567383
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,4,16,power_law_1.2,0.2169932746887207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,8,8,balanced,0.2006412887573242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,8,8,balanced,0.21236095428466797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,8,8,balanced,0.22309247970581056
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,4,16,power_law_1.2,0.15661055564880372
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,8,8,balanced,0.2530892753601074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,8,8,balanced,0.28251775741577145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,8,8,balanced,0.33937793731689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,8,8,balanced,0.39992576599121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,8,8,balanced,0.19134719848632814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,8,8,balanced,0.20301439285278322
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,8,8,balanced,0.5186777496337891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,8,8,balanced,0.6455014038085938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,8,8,balanced,0.03618304014205932
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,8,8,balanced,0.8817561340332031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,8,8,balanced,0.03870464086532593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,8,8,balanced,0.0456601619720459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,8,8,balanced,0.05528448104858399
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,8,8,balanced,1.2035481262207033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,8,8,balanced,0.0662886381149292
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,8,8,balanced,0.09299072265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,8,8,balanced,0.10874752044677734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,8,8,balanced,0.12140416145324706
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,8,8,balanced,0.12309632301330567
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,8,8,balanced,0.12521471977233886
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,8,8,balanced,0.11913472175598146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,8,8,balanced,0.12249088287353516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,8,8,balanced,1.6969139099121093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,8,8,balanced,0.12528767585754394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,8,8,balanced,0.13067520141601563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,8,8,balanced,0.1315174388885498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,8,8,balanced,0.13688063621520996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,8,8,balanced,0.14656895637512207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,8,8,balanced,0.1653862380981445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,8,8,balanced,0.18439168930053712
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,8,8,balanced,0.221026554107666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,8,8,balanced,2.3118182373046876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,8,8,balanced,0.25374847412109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,8,8,balanced,0.32539646148681645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,8,8,balanced,0.39921409606933594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,8,8,balanced,0.5385689544677734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,8,8,balanced,0.7344703674316406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,8,8,balanced,1.0168102264404297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,8,8,balanced,1.3321139526367187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,4,16,power_law_1.2,0.294069766998291
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,4,16,power_law_1.2,0.4248627090454102
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,4,16,power_law_1.2,0.49198078155517583
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,8,8,balanced,0.027578880786895753
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,8,8,balanced,0.06989439964294433
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,8,8,balanced,0.028491520881652833
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,8,8,balanced,0.07006207942962647
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,8,8,balanced,0.07026944160461426
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,8,8,balanced,0.07004159927368164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,8,8,balanced,0.07034111976623535
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,8,8,balanced,0.07087615966796876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,8,8,balanced,0.026759679317474368
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,8,8,balanced,0.07099647998809815
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,8,8,balanced,0.071943678855896
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,8,8,balanced,0.07256959915161133
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,8,8,balanced,0.07423488140106202
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,8,8,balanced,0.07594367980957031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,8,8,balanced,0.07860608100891113
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,8,8,balanced,0.08098688125610351
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,8,8,balanced,0.0373964810371399
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,8,8,balanced,0.0870041561126709
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,8,8,balanced,0.09331711769104004
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,8,8,balanced,0.03048703908920288
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,8,8,balanced,0.10456192016601562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,8,8,power_law_1.01,0.06381184101104737
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,8,8,power_law_1.01,0.0771942377090454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,8,8,power_law_1.01,0.09580672264099122
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,8,8,power_law_1.01,0.0962880039215088
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,8,8,power_law_1.01,0.10242048263549805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,8,8,power_law_1.01,0.12690048217773436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,8,8,power_law_1.01,0.12029312133789063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,8,8,power_law_1.01,0.17828992843627928
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,8,8,power_law_1.01,0.17332223892211912
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,8,8,power_law_1.01,0.17018623352050782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,8,8,power_law_1.01,0.17634687423706055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,8,8,power_law_1.01,0.17691392898559571
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,8,8,power_law_1.01,0.1769126319885254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,8,8,power_law_1.01,0.1855936050415039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,8,8,power_law_1.01,0.1911244773864746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,8,8,power_law_1.01,0.1926041603088379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,8,8,power_law_1.01,0.20266624450683596
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,8,8,power_law_1.01,0.22808959960937497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,8,8,power_law_1.01,0.26047744750976565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,8,8,power_law_1.01,0.3051507186889649
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,8,8,balanced,0.05390079975128174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,8,8,power_law_1.01,0.37180927276611325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,8,8,power_law_1.01,0.4679385757446289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,8,8,power_law_1.01,0.6196403121948242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,8,8,power_law_1.01,0.8873804473876954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,8,8,power_law_1.01,1.134213104248047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,8,8,power_law_1.01,0.04214655876159668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,8,8,power_law_1.01,0.054017281532287596
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,8,8,power_law_1.01,1.557100830078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,8,8,power_law_1.01,0.07941887855529786
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,8,8,power_law_1.01,0.07842304229736327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,8,8,power_law_1.01,1.9349133300781252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,8,8,power_law_1.01,0.09490303993225098
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,8,8,power_law_1.01,0.1006284809112549
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,8,8,power_law_1.01,0.09694080352783203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,8,8,power_law_1.01,0.1129139232635498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,8,8,power_law_1.01,0.11537023544311524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,8,8,power_law_1.01,0.12030847549438477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,8,8,power_law_1.01,0.12125184059143065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,8,8,power_law_1.01,0.1216051197052002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,8,8,power_law_1.01,0.11660287857055665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,8,8,power_law_1.01,0.11870207786560058
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,8,8,power_law_1.01,0.12279295921325684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,8,8,power_law_1.01,0.12185600280761719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,8,8,power_law_1.01,0.12880000114440918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,8,8,power_law_1.01,0.14884608268737792
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,8,8,power_law_1.01,0.163372802734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,8,8,power_law_1.01,0.2010995292663574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,8,8,power_law_1.01,0.22605823516845702
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,8,8,power_law_1.01,0.3004096031188965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,8,8,power_law_1.01,0.3798233413696289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,8,8,power_law_1.01,0.5245849609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,8,8,power_law_1.01,0.6360831832885742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,8,8,power_law_1.01,0.030110719203948977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,8,8,power_law_1.01,0.8713766479492188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,8,8,power_law_1.01,0.03661312103271484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,8,8,power_law_1.01,0.05391615867614746
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,8,8,power_law_1.01,0.050114560127258304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,8,8,power_law_1.01,1.2700415802001952
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,8,8,power_law_1.01,0.057272319793701176
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,8,8,power_law_1.01,0.06868480205535889
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,8,8,power_law_1.01,0.06793727874755859
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,8,8,power_law_1.01,0.0694975996017456
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,8,8,power_law_1.01,0.06140799999237061
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,8,8,power_law_1.01,0.06168575763702393
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,8,8,power_law_1.01,0.060096001625061034
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,8,8,power_law_1.01,0.06467072010040284
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,8,8,power_law_1.01,0.06517632007598877
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,8,8,power_law_1.01,0.06894336223602296
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,8,8,power_law_1.01,0.0664473581314087
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,8,8,power_law_1.01,0.07578112125396728
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,8,8,power_law_1.01,0.07525504112243653
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,8,8,power_law_1.01,0.060710401535034174
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,8,8,balanced,0.0696614408493042
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,8,8,balanced,0.06978047847747802
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,8,8,balanced,0.14390144348144532
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,8,8,balanced,0.17092863082885743
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,8,8,balanced,0.22202239990234376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,8,8,power_law_1.01,0.0897049617767334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,8,8,power_law_1.01,0.11506431579589844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,8,8,power_law_1.2,0.0606169605255127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,8,8,power_law_1.2,0.07710080146789551
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,8,8,power_law_1.01,0.09890687942504883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,8,8,power_law_1.2,0.08563712120056152
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,8,8,power_law_1.2,0.08932352066040039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,8,8,power_law_1.2,0.10939264297485352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,8,8,power_law_1.2,0.11906944274902345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,8,8,power_law_1.2,0.11997695922851563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,8,8,power_law_1.2,0.17218303680419922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,8,8,power_law_1.2,0.17535743713378907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,8,8,power_law_1.2,0.1717900848388672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,8,8,power_law_1.2,0.17676416397094727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,8,8,power_law_1.2,0.17785472869873048
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,8,8,power_law_1.2,0.17516672134399414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,8,8,power_law_1.2,0.18618879318237305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,8,8,power_law_1.01,0.13798144340515137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,8,8,power_law_1.2,0.18994176864624024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,8,8,power_law_1.2,0.1909209632873535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,8,8,power_law_1.2,0.20883712768554688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,8,8,power_law_1.2,0.22956415176391604
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,8,8,power_law_1.2,0.2606028747558594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,8,8,power_law_1.2,0.31220863342285154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,8,8,power_law_1.2,0.37400833129882816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,8,8,power_law_1.2,0.4925503921508789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,8,8,power_law_1.2,0.6445171356201171
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,8,8,power_law_1.2,0.8485145568847656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,8,8,power_law_1.01,0.17998079299926756
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,8,8,power_law_1.2,1.13683837890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,8,8,power_law_1.2,0.04536960124969482
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,8,8,power_law_1.2,0.05371776103973389
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,8,8,power_law_1.2,0.06831999778747558
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,8,8,power_law_1.2,1.7945689392089843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,8,8,power_law_1.2,0.07746560096740722
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,8,8,power_law_1.2,0.09241600036621093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,8,8,power_law_1.2,0.10114815711975098
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,8,8,power_law_1.2,0.0991487979888916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,8,8,power_law_1.2,2.1761740112304686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,8,8,power_law_1.2,0.11338239669799806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,8,8,power_law_1.2,0.11430015563964843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,8,8,power_law_1.2,0.11758975982666016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,8,8,power_law_1.2,0.1130777645111084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,8,8,power_law_1.2,0.11903231620788575
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,8,8,power_law_1.2,0.1124185562133789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,8,8,power_law_1.2,0.11880703926086426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,8,8,power_law_1.2,0.1203878402709961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,8,8,power_law_1.2,0.12011008262634278
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,8,8,power_law_1.2,0.1295680046081543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,8,8,power_law_1.2,0.1543628787994385
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,8,8,power_law_1.2,0.16554367065429687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,8,8,power_law_1.2,0.2052556800842285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,8,8,power_law_1.2,0.24369024276733398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,8,8,power_law_1.2,0.32582271575927735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,8,8,power_law_1.2,0.3968217468261719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,8,8,power_law_1.2,0.5546047973632813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,8,8,power_law_1.2,0.030103039741516114
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,8,8,power_law_1.2,0.7034162902832032
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,8,8,power_law_1.2,0.0381439995765686
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,8,8,power_law_1.2,0.04730495929718018
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,8,8,power_law_1.2,1.042414093017578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,8,8,power_law_1.2,0.04900864124298095
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,8,8,power_law_1.2,0.05885056018829345
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,8,8,power_law_1.2,0.06680575847625733
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,8,8,power_law_1.2,0.06821631908416748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,8,8,power_law_1.2,1.3186904907226562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,8,8,power_law_1.2,0.06900352001190185
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,8,8,power_law_1.2,0.05779712200164795
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,8,8,power_law_1.01,0.23829504013061525
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,8,8,power_law_1.2,0.05826303958892822
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,8,8,power_law_1.01,0.27869056701660155
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,8,8,power_law_1.2,0.06554624080657959
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,8,8,power_law_1.2,0.05730815887451172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,8,8,power_law_1.2,0.060762882232666016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,8,8,power_law_1.2,0.0680076789855957
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,8,8,power_law_1.2,0.07751935958862305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,8,8,power_law_1.2,0.09481344223022461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,8,8,power_law_1.2,0.06687359809875489
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,8,8,power_law_1.2,0.06615935802459717
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,8,8,power_law_1.01,0.45034240722656255
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,8,8,power_law_1.01,0.5626828765869141
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,8,8,power_law_1.2,0.12175616264343261
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,8,8,power_law_1.2,0.08410495758056641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,16,4,balanced,0.035879681110382075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,16,4,balanced,0.03371520042419433
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,16,4,balanced,0.0373145604133606
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,8,8,power_law_1.2,0.13209856033325196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,16,4,balanced,0.0682099199295044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,16,4,balanced,0.0849779224395752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,16,4,balanced,0.10922880172729492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,16,4,balanced,0.132675838470459
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,8,8,power_law_1.2,0.19906816482543946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,16,4,balanced,0.13376511573791502
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,16,4,balanced,0.13554431915283202
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,16,4,balanced,0.13649024009704588
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,16,4,balanced,0.21306112289428708
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,16,4,balanced,0.21500415802001954
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,8,8,power_law_1.2,0.3084940719604492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,16,4,balanced,0.22076160430908204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,16,4,balanced,0.22649599075317384
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,16,4,balanced,0.2300953674316406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,16,4,balanced,0.23825664520263673
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,16,4,balanced,0.2516313552856445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,16,4,balanced,0.2812377548217774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,16,4,balanced,0.30927999496459957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,16,4,balanced,0.364400634765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,16,4,balanced,0.4206771087646485
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,8,8,power_law_1.2,0.4433267211914062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,8,8,power_law_1.2,0.2508761596679688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,16,4,balanced,0.6649842834472656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,16,4,balanced,0.03455744028091431
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,16,4,balanced,0.8926579284667969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,16,4,balanced,0.03784447908401489
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,16,4,balanced,0.04521984100341797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,16,4,balanced,0.05880447864532471
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,16,4,balanced,0.08374655723571778
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,16,4,balanced,1.21346435546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,16,4,balanced,0.11088512420654298
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,16,4,balanced,0.535855369567871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,16,4,balanced,0.13590527534484864
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,16,4,balanced,0.13743616104125977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,16,4,balanced,0.1395353603363037
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,16,4,balanced,0.14073984146118163
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,16,4,balanced,0.15796992301940918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,16,4,balanced,0.16131584167480467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,16,4,balanced,0.16405887603759767
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,16,4,balanced,1.7053312683105468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,16,4,balanced,0.1696294403076172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,16,4,balanced,0.16916608810424805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,16,4,balanced,0.1763007926940918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,16,4,balanced,0.18440063476562502
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,16,4,balanced,0.20323711395263672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,16,4,balanced,0.22008192062377932
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,16,4,balanced,0.2540608024597168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,16,4,balanced,0.2899839973449707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,16,4,balanced,2.309282531738281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,16,4,balanced,0.35838462829589846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,16,4,balanced,0.42656127929687504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,16,4,balanced,0.5614451217651367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,16,4,balanced,0.7852185821533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,16,4,balanced,1.056348190307617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,16,4,balanced,1.3956787109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,8,8,power_law_1.2,0.7081241607666016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,8,8,power_law_1.2,0.882598419189453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,16,4,balanced,0.037852160930633545
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,16,4,balanced,0.0999232006072998
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,16,4,balanced,0.10041343688964843
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,16,4,balanced,0.10008959770202637
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,16,4,balanced,0.10048383712768554
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,16,4,balanced,0.03903615951538086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,16,4,balanced,0.04054143905639648
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,16,4,balanced,0.1006553554534912
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,16,4,balanced,0.04534783840179443
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,16,4,balanced,0.10184703826904298
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,16,4,balanced,0.10334207534790038
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,16,4,balanced,0.10260736465454101
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,16,4,balanced,0.10606847763061525
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,16,4,balanced,0.1071718406677246
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,16,4,balanced,0.11187071800231933
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,16,4,balanced,0.116428804397583
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,16,4,balanced,0.12247039794921874
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,16,4,balanced,0.13233152389526368
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,16,4,balanced,0.058063359260559076
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,16,4,balanced,0.14429951667785645
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,16,4,balanced,0.16774656295776366
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,16,4,power_law_1.01,0.03544575929641723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,16,4,power_law_1.01,0.06040832042694092
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,16,4,balanced,0.07671807765960693
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,16,4,power_law_1.01,0.07746431827545167
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,16,4,power_law_1.01,0.08183039665222168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,16,4,power_law_1.01,0.09652992248535155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,16,4,power_law_1.01,0.1055948829650879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,16,4,power_law_1.01,0.10854528427124024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,16,4,power_law_1.01,0.12001152038574218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,16,4,power_law_1.01,0.12277119636535645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,16,4,power_law_1.01,0.12630528450012207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,16,4,power_law_1.01,0.2002726364135742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,16,4,power_law_1.01,0.20659839630126955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,16,4,power_law_1.01,0.21295103073120117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,16,4,power_law_1.01,0.22000383377075194
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,16,4,power_law_1.01,0.22613759994506838
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,16,4,power_law_1.01,0.22955904006958008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,16,4,power_law_1.01,0.24540927886962888
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,16,4,power_law_1.01,0.2758387184143066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,16,4,power_law_1.01,0.30244863510131836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,16,4,power_law_1.01,0.35822593688964843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,16,4,power_law_1.01,0.41100288391113277
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,16,4,power_law_1.01,0.5371072006225586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,16,4,power_law_1.01,0.673096923828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,16,4,power_law_1.01,0.9645005035400391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,16,4,power_law_1.01,1.200189437866211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,16,4,power_law_1.01,0.038696959018707275
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,16,4,power_law_1.01,0.04606592178344727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,16,4,power_law_1.01,1.728038330078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,16,4,power_law_1.01,0.060500478744506835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,16,4,power_law_1.01,0.07789311885833741
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,16,4,power_law_1.01,0.09861632347106933
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,16,4,power_law_1.01,0.11542143821716308
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,16,4,power_law_1.01,0.11347840309143067
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,16,4,power_law_1.01,2.2426226806640623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,16,4,power_law_1.01,0.12598527908325194
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,16,4,power_law_1.01,0.12881279945373536
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,16,4,power_law_1.01,0.1299404811859131
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,16,4,power_law_1.01,0.14594176292419433
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,16,4,power_law_1.01,0.14846464157104494
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,16,4,power_law_1.01,0.14814208030700685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,16,4,power_law_1.01,0.15635328292846679
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,16,4,balanced,0.09970047950744629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,16,4,power_law_1.01,0.15810303688049315
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,16,4,power_law_1.01,0.16018943786621093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,16,4,power_law_1.01,0.17021184921264648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,16,4,power_law_1.01,0.18727167129516603
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,16,4,power_law_1.01,0.20613759994506836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,16,4,power_law_1.01,0.244005126953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,16,4,power_law_1.01,0.28295551300048827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,16,4,power_law_1.01,0.3686195373535156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,16,4,power_law_1.01,0.4500428771972656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,16,4,power_law_1.01,0.6276275253295898
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,16,4,power_law_1.01,0.04112127780914306
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,16,4,power_law_1.01,0.04544384002685547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,16,4,power_law_1.01,1.1116684722900392
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,16,4,power_law_1.01,0.05902719974517823
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,16,4,power_law_1.01,0.059415040016174314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,16,4,power_law_1.01,1.4500428771972655
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,16,4,power_law_1.01,0.06597631931304931
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,16,4,power_law_1.01,0.7666226959228515
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,16,4,power_law_1.01,0.06777984142303467
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,16,4,power_law_1.01,0.07868159770965576
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,16,4,power_law_1.01,0.09240832328796386
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,16,4,power_law_1.01,0.08035967826843263
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,16,4,power_law_1.01,0.08309375762939453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,16,4,power_law_1.01,0.08545280456542968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,16,4,power_law_1.01,0.0905395221710205
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,16,4,power_law_1.01,0.09368191719055176
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,16,4,power_law_1.01,0.07863296031951904
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,16,4,power_law_1.01,0.09852800369262696
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,16,4,power_law_1.01,0.1034278392791748
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,16,4,power_law_1.01,0.10660863876342774
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,16,4,power_law_1.01,0.11649408340454101
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,16,4,balanced,0.09969663619995117
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,16,4,balanced,0.23904640197753907
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,16,4,balanced,0.3800422286987305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,16,4,balanced,0.2856064033508301
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,16,4,power_law_1.01,0.13675007820129395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,16,4,power_law_1.01,0.16578943252563477
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,16,4,power_law_1.01,0.2292211151123047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,16,4,power_law_1.2,0.035470080375671384
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,16,4,power_law_1.01,0.17522432327270507
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,16,4,power_law_1.2,0.07147136211395264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,16,4,power_law_1.2,0.060526080131530755
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,16,4,power_law_1.2,0.08161919593811036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,16,4,power_law_1.2,0.09438591957092285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,16,4,power_law_1.2,0.1078451156616211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,16,4,power_law_1.2,0.10994560241699218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,16,4,power_law_1.2,0.12102399826049806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,16,4,power_law_1.2,0.12392448425292968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,16,4,power_law_1.2,0.12658687591552736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,16,4,power_law_1.2,0.20174591064453126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,16,4,power_law_1.2,0.20475135803222658
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,16,4,power_law_1.2,0.21222143173217772
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,16,4,power_law_1.2,0.22530815124511722
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,16,4,power_law_1.2,0.23147775650024416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,16,4,power_law_1.2,0.24507904052734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,16,4,power_law_1.2,0.2762700843811035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,16,4,power_law_1.2,0.30480255126953126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,16,4,power_law_1.2,0.36616958618164064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,16,4,power_law_1.2,0.2215897560119629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,16,4,power_law_1.2,0.42133632659912107
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,16,4,power_law_1.2,0.5586099243164062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,16,4,power_law_1.2,0.6912255859374999
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,16,4,power_law_1.2,0.9928755187988282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,16,4,power_law_1.2,1.263146209716797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,16,4,power_law_1.2,0.03913216114044189
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,16,4,power_law_1.2,0.04601727962493897
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,16,4,power_law_1.2,1.7434841918945314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,16,4,power_law_1.2,0.05831808090209961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,16,4,power_law_1.2,0.0765337610244751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,16,4,power_law_1.2,0.0971878433227539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,16,4,power_law_1.2,0.10924032211303711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,16,4,power_law_1.01,0.30674560546874996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,16,4,power_law_1.2,2.3259852600097655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,16,4,power_law_1.2,0.11394816398620607
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,16,4,power_law_1.2,0.12766592025756834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,16,4,power_law_1.2,0.12357760429382325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,16,4,power_law_1.2,0.12996479988098145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,16,4,power_law_1.2,0.14319104194641114
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,16,4,power_law_1.2,0.14879743576049803
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,16,4,power_law_1.2,0.1460761642456055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,16,4,power_law_1.2,0.1532966423034668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,16,4,power_law_1.2,0.1544921588897705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,16,4,power_law_1.2,0.1625075149536133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,16,4,power_law_1.2,0.1689344024658203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,16,4,power_law_1.2,0.18790655136108397
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,16,4,power_law_1.2,0.2109324836730957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,16,4,power_law_1.2,0.2532403182983399
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,16,4,power_law_1.2,0.2912384033203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,16,4,power_law_1.2,0.37034496307373044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,16,4,power_law_1.2,0.4689971160888672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,16,4,power_law_1.2,0.6399526214599609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,16,4,power_law_1.2,0.7765414428710937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,16,4,power_law_1.2,0.041149439811706545
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,16,4,power_law_1.2,0.04460031986236572
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,16,4,power_law_1.2,0.054120960235595705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,16,4,power_law_1.2,1.100549087524414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,16,4,power_law_1.2,0.055125761032104495
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,16,4,power_law_1.2,0.06401663780212402
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,16,4,power_law_1.2,0.0759500789642334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,16,4,power_law_1.2,1.4360333251953126
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,16,4,power_law_1.2,0.06590335845947266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,16,4,power_law_1.2,0.07819776058197021
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,16,4,power_law_1.2,0.07943039894104004
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,16,4,power_law_1.2,0.09376128196716309
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,16,4,power_law_1.2,0.08130047798156739
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,16,4,power_law_1.2,0.08474240303039551
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,16,4,power_law_1.2,0.09020031929016113
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,16,4,power_law_1.2,0.09249919891357422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,16,4,power_law_1.2,0.09856255531311035
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,16,4,power_law_1.2,0.1107750415802002
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,16,4,power_law_1.2,0.10465023994445802
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,16,4,power_law_1.01,0.40447105407714845
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,16,4,power_law_1.01,0.49616897583007813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,16,4,power_law_1.01,0.7187712097167969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,16,4,power_law_1.01,0.917084197998047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,16,4,power_law_1.2,0.7600064086914062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,16,4,power_law_1.2,0.15231743812561035
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,16,4,power_law_1.2,0.19650304794311524
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,16,4,power_law_1.2,0.13028223991394045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,32,2,balanced,0.03305599927902222
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,32,2,balanced,0.03542655944824219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,32,2,balanced,0.04248320102691651
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,32,2,balanced,0.07089920043945312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,16,4,power_law_1.2,0.31518976211547856
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,16,4,power_law_1.2,0.23689727783203124
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,16,4,power_law_1.2,0.3629094314575195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,32,2,balanced,0.1317619228363037
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,32,2,balanced,0.1614131164550781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,32,2,balanced,0.16308992385864257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,32,2,balanced,0.1639232063293457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,32,2,balanced,0.16550655364990235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,32,2,balanced,0.16806400299072266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,32,2,balanced,0.1722483253479004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,32,2,balanced,0.1755276870727539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,32,2,balanced,0.2901158332824707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,32,2,balanced,0.2935308837890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,32,2,balanced,0.09680000305175782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,32,2,balanced,0.3006233596801758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,32,2,balanced,0.3144025611877441
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,32,2,balanced,0.34206336975097656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,32,2,balanced,0.369436149597168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,32,2,balanced,0.4237222290039062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,32,2,balanced,0.48096767425537107
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,32,2,balanced,0.592305908203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,32,2,balanced,0.7269926452636719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,16,4,power_law_1.2,0.5924761581420899
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,32,2,balanced,0.035348479747772214
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,32,2,balanced,0.9469055938720704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,32,2,balanced,0.04028927803039551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,32,2,balanced,0.04821631908416748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,32,2,balanced,0.06611584186553955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,32,2,balanced,0.10760191917419433
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,32,2,balanced,1.2894706726074219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,32,2,balanced,0.16972415924072265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,32,2,balanced,0.2077030372619629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,32,2,balanced,0.20966400146484374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,32,2,balanced,0.21110015869140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,32,2,balanced,0.2131391906738281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,32,2,balanced,0.21624319076538087
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,32,2,balanced,0.2197068786621094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,32,2,balanced,0.22262271881103515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,32,2,balanced,0.28100608825683593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,32,2,balanced,1.7745677185058593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,32,2,balanced,0.28065536499023436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,32,2,balanced,0.286484489440918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,32,2,balanced,0.29387008666992187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,32,2,balanced,0.30965887069702147
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,32,2,balanced,0.324403190612793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,32,2,balanced,0.3599103927612305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,32,2,balanced,0.3926259231567383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,32,2,balanced,0.4610956954956055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,32,2,balanced,2.405827789306641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,32,2,balanced,0.5239334487915039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,32,2,balanced,0.6563597106933594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,32,2,balanced,0.9503743743896484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,32,2,balanced,1.2121395111083983
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,32,2,balanced,1.6424870300292969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,16,4,power_law_1.2,1.2834982299804687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,16,4,power_law_1.2,1.737830352783203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,32,2,balanced,0.0599948787689209
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,32,2,balanced,0.06352896213531493
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,32,2,balanced,0.15856639862060545
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,32,2,balanced,0.1588684844970703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,32,2,balanced,0.05874815940856933
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,32,2,balanced,0.07328767776489258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,32,2,balanced,0.15932415962219237
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,32,2,balanced,0.16098432540893554
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,32,2,balanced,0.16018815994262697
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,32,2,balanced,0.1625152015686035
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,32,2,balanced,0.16394752502441406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,32,2,balanced,0.16611328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,32,2,balanced,0.16949375152587892
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,32,2,balanced,0.1742246437072754
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,32,2,balanced,0.1809561538696289
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,32,2,balanced,0.19159679412841796
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,32,2,balanced,0.2033011245727539
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,32,2,balanced,0.226507511138916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,32,2,balanced,0.2520895957946777
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,32,2,balanced,0.09024640083312988
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,32,2,balanced,0.301136646270752
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,32,2,balanced,0.1235801601409912
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,32,2,power_law_1.01,0.04078080177307129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,32,2,power_law_1.01,0.038984959125518796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,32,2,power_law_1.01,0.07885056018829346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,32,2,power_law_1.01,0.06634880065917968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,32,2,power_law_1.01,0.09667327880859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,32,2,power_law_1.01,0.11757951736450196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,32,2,power_law_1.01,0.12397055625915528
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,32,2,power_law_1.01,0.1384614372253418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,32,2,power_law_1.01,0.1406067180633545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,32,2,power_law_1.01,0.14626815795898437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,32,2,power_law_1.01,0.14977791786193848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,32,2,power_law_1.01,0.1567360019683838
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,32,2,power_law_1.01,0.1615705680847168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,32,2,power_law_1.01,0.2823091125488281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,32,2,power_law_1.01,0.28516992568969723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,32,2,power_law_1.01,0.29190143585205075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,32,2,power_law_1.01,0.30820991516113283
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,32,2,power_law_1.01,0.33848831176757815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,32,2,power_law_1.01,0.36739200592041016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,32,2,power_law_1.01,0.4306841659545898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,32,2,power_law_1.01,0.4900697708129883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,32,2,power_law_1.01,0.608039665222168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,32,2,power_law_1.01,0.7378201293945312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,32,2,power_law_1.01,1.0260671997070312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,32,2,power_law_1.01,1.3102770996093749
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,32,2,power_law_1.01,0.03824896097183227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,32,2,power_law_1.01,0.046012158393859866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,32,2,power_law_1.01,0.05757311820983887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,32,2,power_law_1.01,1.8343154907226562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,32,2,power_law_1.01,0.07781504154205322
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,32,2,power_law_1.01,0.11195008277893066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,32,2,power_law_1.01,0.1487936019897461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,32,2,power_law_1.01,0.1561638355255127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,32,2,power_law_1.01,0.17775360107421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,32,2,power_law_1.01,2.3868389892578126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,32,2,power_law_1.01,0.18257663726806642
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,32,2,balanced,0.15819392204284669
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,32,2,power_law_1.01,0.18603519439697264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,32,2,power_law_1.01,0.18952831268310547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,32,2,power_law_1.01,0.19697919845581055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,32,2,power_law_1.01,0.19724288940429688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,32,2,power_law_1.01,0.24147071838378906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,32,2,power_law_1.01,0.24680831909179685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,32,2,power_law_1.01,0.2622515106201172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,32,2,power_law_1.01,0.27029888153076176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,32,2,power_law_1.01,0.2892940711975098
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,32,2,power_law_1.01,0.30937728881835935
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,32,2,power_law_1.01,0.3569919967651367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,32,2,power_law_1.01,0.4067379379272461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,32,2,power_law_1.01,0.4980223846435547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,32,2,power_law_1.01,0.5832575988769532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,32,2,power_law_1.01,0.7908863830566406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,32,2,power_law_1.01,0.06115200042724609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,32,2,power_law_1.01,0.9567334747314453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,32,2,power_law_1.01,0.0641971206665039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,32,2,power_law_1.01,0.07529727935791015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,32,2,power_law_1.01,1.3445939636230468
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,32,2,power_law_1.01,0.09220352172851562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,32,2,power_law_1.01,0.10577280044555663
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,32,2,power_law_1.01,1.7055194091796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,32,2,power_law_1.01,0.11237248420715332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,32,2,power_law_1.01,0.12422911643981933
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,32,2,power_law_1.01,0.13362303733825684
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,32,2,power_law_1.01,0.1312012767791748
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,32,2,power_law_1.01,0.08049407958984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,32,2,power_law_1.01,0.13477888107299804
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,32,2,power_law_1.01,0.14093695640563966
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,32,2,power_law_1.01,0.1435763168334961
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,32,2,power_law_1.01,0.15221247673034669
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,32,2,power_law_1.01,0.15687935829162597
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,32,2,power_law_1.01,0.16125951766967775
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,32,2,power_law_1.01,0.17221376419067383
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,32,2,power_law_1.01,0.19510271072387694
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,32,2,balanced,0.1576524829864502
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,32,2,balanced,0.5423295974731446
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,32,2,balanced,0.7322585296630859
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,32,2,balanced,0.44504959106445313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,32,2,power_law_1.01,0.22304384231567384
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,32,2,power_law_1.01,0.3362188720703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,32,2,power_law_1.01,0.413752326965332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,32,2,power_law_1.01,0.25703935623168944
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,32,2,power_law_1.2,0.03422847986221313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,32,2,power_law_1.2,0.03905663967132568
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,32,2,power_law_1.2,0.06787968158721924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,32,2,power_law_1.2,0.07844096183776855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,32,2,power_law_1.2,0.09487615585327148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,32,2,power_law_1.2,0.11479807853698731
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,32,2,power_law_1.2,0.12336128234863282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,32,2,power_law_1.2,0.13787391662597656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,32,2,power_law_1.2,0.1418278408050537
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,32,2,power_law_1.2,0.1453593635559082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,32,2,power_law_1.2,0.15149567604064942
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,32,2,power_law_1.2,0.15796992301940918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,32,2,power_law_1.2,0.16219392776489258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,32,2,power_law_1.2,0.2813183975219727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,32,2,power_law_1.2,0.2848652839660645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,32,2,power_law_1.2,0.29277183532714846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,32,2,power_law_1.2,0.30863231658935547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,32,2,power_law_1.2,0.3386188888549805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,32,2,power_law_1.2,0.36967937469482426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,32,2,power_law_1.2,0.4317299270629883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,32,2,power_law_1.2,0.4923968124389648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,32,2,power_law_1.2,0.6085452651977539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,32,2,power_law_1.2,0.7621414184570312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,32,2,power_law_1.2,1.0333952331542968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,32,2,power_law_1.2,1.3125836181640627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,32,2,power_law_1.2,0.03812096118927002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,32,2,power_law_1.2,0.04563712120056153
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,32,2,power_law_1.01,0.5131776046752929
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,32,2,power_law_1.2,0.05465856075286866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,32,2,power_law_1.2,1.854017333984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,32,2,power_law_1.2,0.08023296356201172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,32,2,power_law_1.2,0.10719743728637696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,32,2,power_law_1.2,0.14600192070007326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,32,2,power_law_1.2,0.1502617645263672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,32,2,power_law_1.2,0.17516159057617187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,32,2,power_law_1.2,0.18230527877807617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,32,2,power_law_1.2,0.1831167984008789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,32,2,power_law_1.2,2.409743347167969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,32,2,power_law_1.2,0.18745599746704103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,32,2,power_law_1.2,0.19651327133178711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,32,2,power_law_1.2,0.1988287925720215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,32,2,power_law_1.2,0.23875839233398438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,32,2,power_law_1.2,0.2403046417236328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,32,2,power_law_1.2,0.2610534477233887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,32,2,power_law_1.2,0.26891904830932617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,32,2,power_law_1.2,0.2916454315185547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,32,2,power_law_1.2,0.31469312667846683
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,32,2,power_law_1.2,0.36531585693359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,32,2,power_law_1.2,0.4102425765991211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,32,2,power_law_1.2,0.5096230316162109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,32,2,power_law_1.2,0.5944601440429688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,32,2,power_law_1.2,0.8091558074951172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,32,2,power_law_1.2,0.061148161888122554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,32,2,power_law_1.2,0.9699903869628906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,32,2,power_law_1.2,0.06446847915649415
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,32,2,power_law_1.2,0.07305727958679199
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,32,2,power_law_1.2,0.07965695858001709
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,32,2,power_law_1.2,1.3744895935058594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,32,2,power_law_1.2,0.09021696090698242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,32,2,power_law_1.2,0.10018560409545899
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,32,2,power_law_1.2,0.10775168418884278
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,32,2,power_law_1.2,0.12325247764587402
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,32,2,power_law_1.2,0.12591232299804686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,32,2,power_law_1.2,1.7458738708496093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,32,2,power_law_1.2,0.12950016021728517
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,32,2,power_law_1.2,0.13283712387084962
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,32,2,power_law_1.2,0.13701631546020507
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,32,2,power_law_1.2,0.14412287712097166
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,32,2,power_law_1.01,0.7046489715576172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,32,2,power_law_1.2,0.16190975189208984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,32,2,power_law_1.2,0.15422592163085938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,32,2,power_law_1.2,0.1679372787475586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,32,2,power_law_1.2,0.18461952209472657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,32,2,power_law_1.2,0.22590591430664064
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,32,2,power_law_1.2,0.24737535476684572
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,32,2,power_law_1.2,0.31686656951904296
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,32,2,power_law_1.01,0.9536486053466797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,32,2,power_law_1.01,1.4438758850097657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,32,2,power_law_1.01,1.842488250732422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,64,balanced,0.026777598857879635
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,64,balanced,0.025955839157104494
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,64,balanced,0.026371839046478274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,64,balanced,0.026365439891815185
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,64,balanced,0.0283187198638916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,64,balanced,0.04871424198150635
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,64,balanced,0.07723904132843018
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,64,balanced,0.07729407787322998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,64,balanced,0.07859968185424805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,64,balanced,0.07853439807891846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,64,balanced,0.07926400184631348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,64,balanced,0.08007935523986817
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,64,balanced,0.0811571216583252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,64,balanced,0.08208255767822266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,64,balanced,0.08396544456481933
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,64,balanced,0.08645631790161132
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,64,balanced,0.0901427173614502
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,64,balanced,0.0961894416809082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,64,balanced,0.08954239845275878
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,64,balanced,0.09660032272338867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,64,balanced,0.10152192115783691
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,64,balanced,0.11370368003845215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,64,balanced,0.12463232040405274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,64,balanced,0.1498316764831543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,64,balanced,0.1779033660888672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,64,balanced,0.2587379264831543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,64,balanced,0.32213630676269533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,64,balanced,0.028433918952941895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,64,balanced,0.028471040725708007
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,64,balanced,0.028494079113006592
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,64,balanced,0.02857599973678589
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,64,balanced,0.029272320270538332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,64,balanced,0.0293120002746582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,64,balanced,0.03561984062194824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,64,balanced,0.03581824064254761
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,64,balanced,0.03611263990402221
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,64,balanced,0.03615231990814209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,64,balanced,0.036607999801635746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,64,balanced,0.037063679695129394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,64,balanced,0.037427198886871335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,64,balanced,0.03790335893630982
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,64,balanced,0.03837951898574829
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,64,balanced,0.03906559944152832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,64,balanced,0.040285439491271974
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,64,balanced,0.04322432041168213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,64,balanced,0.041208319664001465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,64,balanced,0.04837376117706299
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,64,balanced,0.05110015869140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,64,balanced,0.05869823932647705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,64,balanced,0.06537727832794189
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,64,balanced,0.08321279525756836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,32,2,power_law_1.2,0.40335617065429685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,64,balanced,0.10055551528930665
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,32,2,power_law_1.2,0.5328217697143555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,64,balanced,0.13791487693786622
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,64,balanced,0.17458816528320314
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,32,2,power_law_1.2,0.7177458953857422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,32,2,power_law_1.2,1.4525311279296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,32,2,power_law_1.2,1.015805435180664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,32,2,power_law_1.2,2.1053900146484374
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,32,2,power_law_1.2,2.936490173339844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,1,64,balanced,0.016582399606704712
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,1,64,balanced,0.017639679908752443
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,1,64,balanced,0.017015039920806885
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,1,64,balanced,0.017424639463424683
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,1,64,balanced,0.017446399927139283
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,1,64,balanced,0.017771519422531128
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,1,64,balanced,0.016948479413986205
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,1,64,balanced,0.01742591977119446
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,1,64,balanced,0.01739135980606079
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,1,64,balanced,0.01773568034172058
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,1,64,balanced,0.017404160499572753
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,1,64,balanced,0.01780351996421814
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,1,64,balanced,0.01815808057785034
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,1,64,balanced,0.018156800270080566
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,1,64,balanced,0.018227200508117675
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,1,64,balanced,0.018350080251693726
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,1,64,balanced,0.019032319784164427
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,1,64,balanced,0.019399679899215698
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,1,64,balanced,0.016930559873580935
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,1,64,balanced,0.01736575961112976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,64,power_law_1.01,0.04995967864990235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,64,power_law_1.01,0.04882559776306152
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,1,64,balanced,0.017697279453277585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,64,power_law_1.01,0.048544001579284665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,64,power_law_1.01,0.04898687839508057
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,64,power_law_1.01,0.07393919944763183
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,64,power_law_1.01,0.07654399871826172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,64,power_law_1.01,0.07804160118103028
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,64,power_law_1.01,0.07923840045928955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,64,power_law_1.01,0.08051456451416016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,64,power_law_1.01,0.08160767555236817
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,1,64,balanced,0.01740288019180298
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,64,power_law_1.01,0.08295680046081542
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,64,power_law_1.01,0.08645119667053222
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,64,power_law_1.01,0.08959615707397461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,64,power_law_1.01,0.09302656173706055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,64,power_law_1.01,0.09644800186157226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,64,power_law_1.01,0.09828607559204101
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,64,power_law_1.01,0.09683584213256836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,64,power_law_1.01,0.10345727920532226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,64,power_law_1.01,0.11294207572937012
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,64,power_law_1.01,0.13930496215820312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,64,power_law_1.01,0.16458623886108398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,64,power_law_1.01,0.24652416229248048
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,64,power_law_1.01,0.25125120162963865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,64,power_law_1.01,0.39784191131591795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,64,power_law_1.01,0.5533721542358399
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,1,64,balanced,0.017390079498291015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,64,power_law_1.01,0.02892672061920166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,64,power_law_1.01,0.8524543762207031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,64,power_law_1.01,0.029300479888916015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,64,power_law_1.01,0.029822719097137452
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,64,power_law_1.01,1.1366528320312501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,64,power_law_1.01,0.03543168067932129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,64,power_law_1.01,0.03626240015029907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,64,power_law_1.01,0.036634879112243654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,64,power_law_1.01,0.03664383888244629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,64,power_law_1.01,0.037411839962005616
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,64,power_law_1.01,0.03839359998703003
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,64,power_law_1.01,0.03912447929382325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,64,power_law_1.01,0.037946879863739014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,64,power_law_1.01,0.04032896041870117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,64,power_law_1.01,0.042414078712463374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,64,power_law_1.01,0.029317119121551515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,64,power_law_1.01,0.04527743816375732
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,64,power_law_1.01,0.046108160018920895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,64,power_law_1.01,0.05419904232025147
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,64,power_law_1.01,0.06368768215179443
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,64,power_law_1.01,0.07373311996459961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,64,power_law_1.01,0.08537983894348145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,64,power_law_1.01,0.1386470413208008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,64,power_law_1.01,0.14723584175109863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,64,power_law_1.01,0.24183423995971678
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,64,power_law_1.01,0.2970572853088379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,64,power_law_1.01,0.04203775882720947
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,1,64,power_law_1.01,0.017415679693222046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,64,power_law_1.01,0.42403839111328123
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,1,64,power_law_1.01,0.017396479845046997
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,1,64,power_law_1.01,0.0173363196849823
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,1,64,power_law_1.01,0.017768959999084472
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,1,64,power_law_1.01,0.01744256019592285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,1,64,power_law_1.01,0.01737727999687195
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,1,64,power_law_1.01,0.017429759502410887
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,1,64,power_law_1.01,0.017396479845046997
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,1,64,power_law_1.01,0.017405439615249634
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,1,64,power_law_1.01,0.017745920419692994
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,1,64,power_law_1.01,0.017812479734420777
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,1,64,power_law_1.01,0.017804800271987914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,1,64,power_law_1.01,0.017822719812393188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,1,64,power_law_1.01,0.01786496043205261
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,1,64,power_law_1.01,0.018169599771499633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,1,64,power_law_1.01,0.0184934401512146
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,1,64,power_law_1.01,0.018622720241546632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,64,power_law_1.01,0.638476791381836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,1,64,balanced,0.03044735908508301
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,1,64,balanced,0.02670464038848877
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,1,64,balanced,0.025848319530487062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,1,64,balanced,0.039109120368957515
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,1,64,power_law_1.01,0.02517503976821899
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,1,64,power_law_1.01,0.02021631956100464
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,1,64,power_law_1.01,0.021962239742279052
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,64,power_law_1.2,0.050599679946899415
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,1,64,power_law_1.01,0.028815360069274903
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,64,power_law_1.2,0.04928256034851074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,64,power_law_1.2,0.04870783805847168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,64,power_law_1.2,0.07467008113861084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,64,power_law_1.2,0.07366655826568604
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,64,power_law_1.2,0.07520512104034424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,64,power_law_1.2,0.0785088014602661
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,64,power_law_1.2,0.07973120212554932
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,64,power_law_1.2,0.0800870418548584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,64,power_law_1.2,0.08151040077209473
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,64,power_law_1.2,0.08408831596374512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,64,power_law_1.2,0.08902272224426269
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,64,power_law_1.2,0.09201151847839355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,64,power_law_1.2,0.09947392463684082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,64,power_law_1.2,0.09668736457824707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,64,power_law_1.2,0.09569791793823243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,64,power_law_1.2,0.0979967975616455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,64,power_law_1.2,0.11428223609924318
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,64,power_law_1.2,0.1293900775909424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,64,power_law_1.2,0.15970687866210936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,64,power_law_1.2,0.21526784896850587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,64,power_law_1.2,0.2707148742675781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,64,power_law_1.2,0.41673854827880863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,64,power_law_1.2,0.508988151550293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,64,power_law_1.2,0.7591705322265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,64,power_law_1.2,1.163787536621094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,64,power_law_1.2,0.030872321128845213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,64,power_law_1.2,0.029729280471801754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,64,power_law_1.2,0.029753599166870114
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,64,power_law_1.2,1.841930236816406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,64,power_law_1.2,0.03539072036743164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,64,power_law_1.2,0.036627199649810795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,64,power_law_1.2,0.03578880071640015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,64,power_law_1.2,0.03707648038864136
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,64,power_law_1.2,0.03789952039718628
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,64,power_law_1.2,0.037662720680236815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,64,power_law_1.2,0.03831680059432983
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,64,power_law_1.2,0.038396160602569576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,64,power_law_1.2,0.03939712047576904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,64,power_law_1.2,0.04071040153503418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,64,power_law_1.2,0.044107518196105956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,64,power_law_1.2,0.04865024089813232
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,64,power_law_1.2,0.05263999938964844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,64,power_law_1.2,0.06271999835968017
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,1,64,power_law_1.01,0.037911040782928465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,64,power_law_1.2,0.045250558853149415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,64,power_law_1.2,0.06535808086395264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,64,power_law_1.2,0.08865280151367187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,64,power_law_1.2,0.10284543991088868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,64,power_law_1.2,0.15927552223205566
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,1,64,power_law_1.01,0.05418240070343018
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,64,power_law_1.2,0.190513916015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,64,power_law_1.2,0.32927745819091797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,64,power_law_1.2,0.38804481506347654
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,1,64,power_law_1.2,0.017395199537277223
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,1,64,power_law_1.2,0.01742591977119446
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,1,64,power_law_1.01,0.04054272174835205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,64,power_law_1.2,0.6010585784912109
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,1,64,power_law_1.2,0.01761023998260498
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,1,64,power_law_1.2,0.017388800382614138
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,1,64,power_law_1.2,0.017461760044097902
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,64,power_law_1.2,0.8757606506347656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,1,64,power_law_1.2,0.017418240308761594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,1,64,power_law_1.2,0.017405439615249634
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,1,64,power_law_1.2,0.017824000120162962
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,1,64,power_law_1.2,0.017847039699554444
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,1,64,power_law_1.2,0.018145279884338377
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,1,64,power_law_1.2,0.017779200077056884
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,1,64,power_law_1.2,0.018173439502716066
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,1,64,power_law_1.2,0.018236160278320312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,1,64,power_law_1.2,0.01821568012237549
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,1,64,power_law_1.2,0.019627519845962525
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,1,64,power_law_1.2,0.017242879867553712
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,1,64,power_law_1.2,0.025437440872192386
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,1,64,power_law_1.2,0.01739776015281677
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,1,64,power_law_1.01,0.07197824001312256
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,1,64,power_law_1.01,0.09081343650817872
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,1,64,power_law_1.01,0.12904191970825196
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,1,64,power_law_1.2,0.02141439914703369
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,1,64,power_law_1.2,0.04417535781860352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,2,32,balanced,0.025195519924163818
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,2,32,balanced,0.02405503988265991
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,1,64,power_law_1.2,0.054401278495788574
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,1,64,power_law_1.2,0.029798400402069092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,2,32,balanced,0.02470400094985962
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,2,32,balanced,0.023960320949554442
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,2,32,balanced,0.026220800876617434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,2,32,balanced,0.04591487884521485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,2,32,balanced,0.07060736179351806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,2,32,balanced,0.07205503940582275
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,2,32,balanced,0.07668735980987548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,2,32,balanced,0.07255167961120605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,2,32,balanced,0.07670783996582031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,2,32,balanced,0.07477375984191895
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,1,64,power_law_1.2,0.07841536045074463
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,2,32,balanced,0.0775487995147705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,2,32,balanced,0.08178175926208496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,2,32,balanced,0.08402560234069825
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,2,32,balanced,0.08873215675354004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,2,32,balanced,0.07941504001617432
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,2,32,balanced,0.08580608367919922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,2,32,balanced,0.09079551696777344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,2,32,balanced,0.09920127868652344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,2,32,balanced,0.10659071922302246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,2,32,balanced,0.12902400016784668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,2,32,balanced,0.14569727897644041
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,1,64,power_law_1.2,0.03495167970657349
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,2,32,balanced,0.18816640853881836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,2,32,balanced,0.23113088607788085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,2,32,balanced,0.02776832103729248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,2,32,balanced,0.33707008361816404
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,2,32,balanced,0.02600831985473633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,2,32,balanced,0.028023040294647215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,2,32,balanced,0.42582145690917966
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,2,32,balanced,0.026832640171051025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,2,32,balanced,0.027650558948516847
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,2,32,balanced,0.029789440631866455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,2,32,balanced,0.035022079944610596
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,2,32,balanced,0.03658879995346069
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,2,32,balanced,0.03546367883682251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,2,32,balanced,0.03580415964126587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,2,32,balanced,0.03824127912521362
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,2,32,balanced,0.036680960655212404
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,2,32,balanced,0.03906559944152832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,2,32,balanced,0.04045055866241455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,2,32,balanced,0.04273920059204102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,2,32,balanced,0.03870208024978637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,2,32,balanced,0.04281087875366211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,2,32,balanced,0.0468287992477417
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,2,32,balanced,0.0534335994720459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,2,32,balanced,0.05953407764434815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,2,32,balanced,0.07190656185150146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,2,32,balanced,0.08408063888549805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,2,32,balanced,0.1129190444946289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,2,32,balanced,0.137642240524292
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,2,32,balanced,0.1929216003417969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,2,32,balanced,0.25560703277587893
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,2,32,balanced,0.035427839756011964
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,1,64,power_law_1.2,0.09220224380493164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,1,64,power_law_1.2,0.13714816093444823
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,1,64,power_law_1.2,0.2141017532348633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,2,32,balanced,0.017036800384521482
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,2,32,balanced,0.017401599884033205
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,2,32,balanced,0.016943360567092898
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,2,32,balanced,0.019048960208892823
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,2,32,balanced,0.019040000438690186
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,2,32,balanced,0.01904512047767639
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,2,32,balanced,0.018881280422210694
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,2,32,balanced,0.019031039476394652
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,2,32,balanced,0.01900287985801697
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,2,32,balanced,0.017363200187683107
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,2,32,balanced,0.017077759504318238
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,2,32,balanced,0.01904255986213684
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,2,32,balanced,0.019374079704284668
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,2,32,balanced,0.019392000436782835
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,2,32,balanced,0.019466240406036377
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,2,32,balanced,0.019722239971160887
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,2,32,balanced,0.020217599868774413
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,2,32,balanced,0.021267199516296388
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,2,32,balanced,0.02186624050140381
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,2,32,balanced,0.01980288028717041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,2,32,power_law_1.01,0.0445849609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,2,32,power_law_1.01,0.044843521118164066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,2,32,power_law_1.01,0.044834561347961426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,2,32,power_law_1.01,0.04646399974822998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,2,32,power_law_1.01,0.043669757843017576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,2,32,power_law_1.01,0.06894720077514649
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,2,32,balanced,0.019097599983215332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,2,32,power_law_1.01,0.0731059217453003
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,2,32,balanced,0.018969600200653074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,2,32,power_law_1.01,0.07314815998077392
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,2,32,power_law_1.01,0.07741695880889893
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,2,32,power_law_1.01,0.07597184181213379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,2,32,power_law_1.01,0.07718656063079835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,2,32,power_law_1.01,0.08163328170776367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,2,32,power_law_1.01,0.0829196834564209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,2,32,power_law_1.01,0.0903923225402832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,2,32,power_law_1.01,0.0828876781463623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,2,32,power_law_1.01,0.08383872032165526
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,2,32,power_law_1.01,0.08881152153015137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,2,32,power_law_1.01,0.09677951812744141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,2,32,power_law_1.01,0.11017344474792481
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,2,32,power_law_1.01,0.12731391906738282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,2,32,power_law_1.01,0.15255807876586913
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,2,32,power_law_1.01,0.178155517578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,2,32,power_law_1.01,0.26871551513671876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,2,32,power_law_1.01,0.33003265380859376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,2,32,power_law_1.01,0.5125708770751953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,2,32,power_law_1.01,0.029230079650878905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,2,32,power_law_1.01,0.7202073669433593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,2,32,power_law_1.01,0.029109759330749514
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,2,32,power_law_1.01,0.881786880493164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,2,32,power_law_1.01,0.02930304050445557
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,2,32,power_law_1.01,0.02939775943756103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,2,32,power_law_1.01,0.030127360820770267
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,2,32,power_law_1.01,0.034236159324646
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,2,32,power_law_1.01,0.03508991956710815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,2,32,power_law_1.01,0.03498752117156982
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,2,32,power_law_1.01,0.03592063903808594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,2,32,power_law_1.01,0.036204800605773926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,2,32,power_law_1.01,0.03789567947387695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,2,32,power_law_1.01,0.039121921062469485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,2,32,power_law_1.01,0.04115456104278564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,2,32,power_law_1.01,0.044834561347961426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,2,32,power_law_1.01,0.04260096073150635
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,2,32,power_law_1.01,0.04113535881042481
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,2,32,power_law_1.01,0.04688896179199219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,2,32,power_law_1.01,0.05394944190979004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,2,32,power_law_1.01,0.058743038177490235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,2,32,power_law_1.01,0.0720524787902832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,2,32,power_law_1.01,0.09043456077575683
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,2,32,power_law_1.01,0.11281791687011719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,2,32,power_law_1.01,0.13385215759277344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,2,32,power_law_1.01,0.2027737617492676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,2,32,power_law_1.01,0.2787827110290527
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,2,32,power_law_1.01,0.018978559970855714
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,2,32,power_law_1.01,0.3882624053955078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,2,32,power_law_1.01,0.01881343960762024
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,2,32,power_law_1.01,0.019048960208892823
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,2,32,power_law_1.01,0.5307968139648438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,2,32,power_law_1.01,0.01896191954612732
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,2,32,power_law_1.01,0.018627840280532836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,2,32,power_law_1.01,0.018216960430145264
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,2,32,power_law_1.01,0.018621439933776858
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,2,32,power_law_1.01,0.01901311993598938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,2,32,power_law_1.01,0.019069440364837646
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,2,32,power_law_1.01,0.01903615951538086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,2,32,power_law_1.01,0.019037439823150634
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,2,32,power_law_1.01,0.019475200176239015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,2,32,power_law_1.01,0.01913856029510498
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,2,32,power_law_1.01,0.0196070396900177
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,2,32,power_law_1.01,0.019486720561981204
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,2,32,power_law_1.01,0.019695359468460082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,2,32,power_law_1.01,0.019848959445953367
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,2,32,balanced,0.018999040126800537
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,2,32,balanced,0.03048959970474243
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,2,32,balanced,0.039441919326782225
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,2,32,balanced,0.030103039741516114
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,2,32,balanced,0.04978176116943359
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,2,32,power_law_1.01,0.022361600399017335
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,2,32,power_law_1.01,0.02155519962310791
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,2,32,power_law_1.01,0.02558079957962036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,2,32,power_law_1.2,0.043878397941589355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,2,32,power_law_1.2,0.044107518196105956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,2,32,power_law_1.2,0.04402175903320312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,2,32,power_law_1.2,0.04424064159393311
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,2,32,power_law_1.2,0.042954239845275875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,2,32,power_law_1.2,0.0675225591659546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,2,32,power_law_1.2,0.07043200016021729
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,2,32,power_law_1.2,0.07204095840454101
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,2,32,power_law_1.2,0.07467008113861084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,2,32,power_law_1.2,0.07687808036804199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,2,32,power_law_1.2,0.07951104164123535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,2,32,power_law_1.2,0.08409855842590333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,2,32,power_law_1.2,0.0832908821105957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,2,32,power_law_1.2,0.08717823982238769
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,2,32,power_law_1.01,0.031470079421997074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,2,32,power_law_1.2,0.08260224342346192
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,2,32,power_law_1.2,0.08569600105285644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,2,32,power_law_1.2,0.0903769588470459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,2,32,power_law_1.2,0.1008396816253662
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,2,32,power_law_1.2,0.11108608245849609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,2,32,power_law_1.2,0.14639103889465332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,2,32,power_law_1.2,0.1742246437072754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,2,32,power_law_1.2,0.25657728195190427
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,2,32,power_law_1.2,0.3157478332519531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,2,32,power_law_1.2,0.4818611145019531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,2,32,power_law_1.2,0.6407180786132812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,2,32,power_law_1.2,0.9022553253173828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,2,32,power_law_1.2,1.2362534332275392
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,2,32,power_law_1.01,0.03518208026885986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,2,32,power_law_1.2,0.029227519035339357
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,2,32,power_law_1.2,0.02932224035263061
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,2,32,power_law_1.2,0.029617919921875003
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,2,32,power_law_1.2,0.0297216010093689
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,2,32,power_law_1.2,0.034612479209899905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,2,32,power_law_1.2,0.03524863958358765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,2,32,power_law_1.2,0.03635328054428101
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,2,32,power_law_1.2,0.03545216083526611
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,2,32,power_law_1.2,0.03824127912521362
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,2,32,power_law_1.01,0.0563263988494873
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,2,32,power_law_1.01,0.07525631904602051
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,2,32,power_law_1.2,0.04472576141357422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,2,32,power_law_1.2,0.04236032009124756
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,2,32,power_law_1.01,0.045265917778015134
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,2,32,power_law_1.2,0.04263296127319336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,2,32,power_law_1.2,0.046370558738708496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,2,32,power_law_1.2,0.04821504116058349
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,2,32,power_law_1.2,0.05689087867736816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,2,32,power_law_1.2,0.06097919940948486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,2,32,power_law_1.2,0.08430208206176758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,2,32,power_law_1.2,0.09979647636413574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,2,32,power_law_1.2,0.03375488042831421
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,2,32,power_law_1.2,0.15561216354370117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,2,32,power_law_1.2,0.19019647598266604
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,2,32,power_law_1.2,0.27003007888793945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,2,32,power_law_1.2,0.36688766479492185
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,2,32,power_law_1.2,0.019216639995574953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,2,32,power_law_1.2,0.5440268707275391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,2,32,power_law_1.2,0.03623552083969116
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,2,32,power_law_1.2,0.019006719589233396
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,2,32,power_law_1.2,0.018982399702072144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,2,32,power_law_1.2,0.040330238342285156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,2,32,power_law_1.2,0.7896857452392578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,2,32,power_law_1.2,0.01863168001174927
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,2,32,power_law_1.2,0.01824895977973938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,2,32,power_law_1.2,0.018602240085601806
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,2,32,power_law_1.2,0.018638080358505248
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,2,32,power_law_1.2,0.019056639671325683
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,2,32,power_law_1.2,0.01899392008781433
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,2,32,power_law_1.2,0.019224319458007812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,2,32,power_law_1.2,0.01944192051887512
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,2,32,power_law_1.2,0.018984960317611693
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,2,32,power_law_1.2,0.019512319564819337
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,2,32,power_law_1.2,0.01947648048400879
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,2,32,power_law_1.2,0.01984768033027649
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,2,32,power_law_1.2,0.021273601055145263
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,2,32,power_law_1.2,0.018862080574035645
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,2,32,power_law_1.01,0.10125951766967774
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,2,32,power_law_1.01,0.12615551948547363
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,2,32,power_law_1.2,0.02622335910797119
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,2,32,power_law_1.2,0.033342719078063965
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,2,32,power_law_1.2,0.023578879833221437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,2,32,power_law_1.2,0.054426879882812494
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,4,16,balanced,0.02317823886871338
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,4,16,balanced,0.020689918994903564
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,2,32,power_law_1.2,0.0476582384109497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,4,16,balanced,0.021848320960998535
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,2,32,power_law_1.2,0.034544639587402344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,4,16,balanced,0.030151679515838626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,4,16,balanced,0.04708096027374268
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,4,16,balanced,0.07364736080169679
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,4,16,balanced,0.07247104167938231
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,4,16,balanced,0.07550975799560547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,4,16,balanced,0.0756006383895874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,4,16,balanced,0.07662591934204102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,4,16,balanced,0.0818342399597168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,4,16,balanced,0.0848742389678955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,4,16,balanced,0.07535488128662109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,4,16,balanced,0.07918975830078126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,4,16,balanced,0.08114815711975097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,4,16,balanced,0.08531071662902831
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,4,16,balanced,0.09318016052246095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,4,16,balanced,0.09869312286376954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,4,16,balanced,0.11596416473388671
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,4,16,balanced,0.13154687881469726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,4,16,balanced,0.16839935302734374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,4,16,balanced,0.023930881023406982
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,4,16,balanced,0.19920768737792968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,4,16,balanced,0.2657036781311035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,4,16,balanced,0.33446144104003905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,4,16,balanced,0.02717695951461792
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,4,16,balanced,0.025414400100708008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,4,16,balanced,0.026760959625244142
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,4,16,balanced,0.02697983980178833
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,4,16,balanced,0.6369792175292969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,4,16,balanced,0.028910079002380372
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,4,16,balanced,0.03405951976776123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,4,16,balanced,0.04062719821929932
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,4,16,balanced,0.040785918235778804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,4,16,balanced,0.04119552135467529
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,4,16,balanced,0.0411084794998169
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,4,16,balanced,0.04250495910644531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,4,16,balanced,0.04356863975524902
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,4,16,balanced,0.04512896060943604
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,4,16,balanced,0.04092415809631348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,4,16,balanced,0.04279551982879638
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,4,16,balanced,0.04490752220153808
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,4,16,balanced,0.047251200675964354
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,4,16,balanced,0.05128960132598877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,4,16,balanced,0.05715583801269532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,4,16,balanced,0.07017600059509277
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,4,16,balanced,0.49691009521484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,4,16,balanced,0.08059904098510742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,4,16,balanced,0.10356479644775392
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,4,16,balanced,0.12363391876220704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,4,16,balanced,0.17583103179931642
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,4,16,balanced,0.2182041549682617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,4,16,balanced,0.3085312080383301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,4,16,balanced,0.4119884872436524
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,2,32,power_law_1.2,0.07912064075469971
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,2,32,power_law_1.2,0.09984895706176758
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,2,32,power_law_1.2,0.13963135719299316
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,2,32,power_law_1.2,0.19659263610839844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,4,16,balanced,0.018188799619674685
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,4,16,balanced,0.021858561038970947
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,4,16,balanced,0.021903359889984132
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,4,16,balanced,0.021853439807891846
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,4,16,balanced,0.022058238983154295
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,4,16,balanced,0.01859455943107605
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,4,16,balanced,0.021908481121063233
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,4,16,balanced,0.021928958892822266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,4,16,balanced,0.018160640001296996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,4,16,balanced,0.02231679916381836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,4,16,balanced,0.022330880165100098
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,4,16,balanced,0.022247679233551025
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,4,16,balanced,0.022726399898529055
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,4,16,balanced,0.023065600395202637
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,4,16,balanced,0.01901695966720581
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,4,16,balanced,0.0234879994392395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,4,16,balanced,0.024368638992309573
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,4,16,balanced,0.026027519702911377
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,4,16,balanced,0.027051520347595216
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,4,16,balanced,0.019813120365142822
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,4,16,balanced,0.02147968053817749
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,4,16,power_law_1.01,0.025506560802459714
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,4,16,power_law_1.01,0.04403327941894532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,4,16,power_law_1.01,0.04428800106048584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,4,16,power_law_1.01,0.044801278114318846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,4,16,power_law_1.01,0.0458406400680542
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,4,16,power_law_1.01,0.06668799877166748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,4,16,power_law_1.01,0.06803455829620361
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,4,16,power_law_1.01,0.07030911922454834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,4,16,power_law_1.01,0.07165696144104003
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,4,16,power_law_1.01,0.07383296012878418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,4,16,power_law_1.01,0.0763865613937378
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,4,16,power_law_1.01,0.08152704238891602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,4,16,power_law_1.01,0.08461952209472656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,4,16,power_law_1.01,0.07659008026123046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,4,16,power_law_1.01,0.0783846378326416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,4,16,power_law_1.01,0.08089471817016601
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,4,16,power_law_1.01,0.08538496017456054
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,4,16,power_law_1.01,0.0975590419769287
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,4,16,power_law_1.01,0.1083238410949707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,4,16,power_law_1.01,0.1329190444946289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,4,16,power_law_1.01,0.15764479637145995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,4,16,power_law_1.01,0.21004287719726564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,4,16,power_law_1.01,0.24934015274047852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,4,16,power_law_1.01,0.40068607330322265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,4,16,power_law_1.01,0.49812095642089843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,4,16,power_law_1.01,0.7798464202880859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,4,16,power_law_1.01,0.027589120864868165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,4,16,power_law_1.01,0.8510578918457032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,4,16,power_law_1.01,0.03133311986923218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,4,16,power_law_1.01,0.03210623979568482
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,4,16,power_law_1.01,0.03221632003784179
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,4,16,power_law_1.01,0.03216255903244018
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,4,16,power_law_1.01,0.036729600429534916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,4,16,power_law_1.01,0.03947520017623901
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,4,16,power_law_1.01,0.039635200500488285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,4,16,power_law_1.01,0.039942400455474855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,4,16,power_law_1.01,0.0415718412399292
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,4,16,power_law_1.01,0.044211201667785645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,4,16,power_law_1.01,0.044546561241149904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,4,16,power_law_1.01,0.04598400115966797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,4,16,power_law_1.01,0.04303487777709961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,4,16,power_law_1.01,0.044558081626892086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,4,16,power_law_1.01,0.047054080963134765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,4,16,power_law_1.01,0.05074816226959229
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,4,16,power_law_1.01,0.057136640548706055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,4,16,power_law_1.01,0.06521984100341796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,4,16,power_law_1.01,0.08407423973083497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,4,16,power_law_1.01,0.10068736076354981
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,4,16,power_law_1.01,0.12411904335021973
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,4,16,power_law_1.01,0.16402048110961914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,4,16,power_law_1.01,0.23727104187011722
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,4,16,power_law_1.01,0.2984486389160156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,4,16,power_law_1.01,0.44679039001464843
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,4,16,power_law_1.01,0.020254719257354736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,4,16,power_law_1.01,0.5630335998535156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,4,16,power_law_1.01,0.02157183885574341
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,4,16,power_law_1.01,0.021442561149597167
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,4,16,power_law_1.01,0.021427199840545655
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,4,16,power_law_1.01,0.021443839073181155
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,4,16,power_law_1.01,0.02157183885574341
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,4,16,power_law_1.01,0.02167680025100708
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,4,16,power_law_1.01,0.02164736032485962
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,4,16,power_law_1.01,0.021864960193634035
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,4,16,power_law_1.01,0.02147200107574463
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,4,16,power_law_1.01,0.021838080883026124
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,4,16,power_law_1.01,0.02192512035369873
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,4,16,power_law_1.01,0.02187391996383667
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,4,16,power_law_1.01,0.022304000854492186
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,4,16,power_law_1.01,0.022263040542602537
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,4,16,power_law_1.01,0.022464001178741456
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,4,16,power_law_1.01,0.02315903902053833
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,4,16,balanced,0.021934080123901366
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,4,16,balanced,0.02157183885574341
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,4,16,balanced,0.056418561935424806
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,4,16,balanced,0.04183807849884033
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,4,16,balanced,0.04419839859008789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,4,16,balanced,0.06901760101318359
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,4,16,power_law_1.01,0.03863679885864258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,4,16,power_law_1.01,0.026169600486755373
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,4,16,power_law_1.01,0.028056321144104
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,4,16,power_law_1.01,0.033515520095825195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,4,16,power_law_1.2,0.025542399883270263
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,4,16,power_law_1.2,0.04439040184020996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,4,16,power_law_1.2,0.04247551918029785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,4,16,power_law_1.2,0.04437376022338867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,4,16,power_law_1.2,0.0463321590423584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,4,16,power_law_1.2,0.06379392147064208
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,4,16,power_law_1.2,0.06859776020050048
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,4,16,power_law_1.2,0.07006336212158203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,4,16,power_law_1.2,0.07025152206420898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,4,16,power_law_1.2,0.07294335842132568
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,4,16,power_law_1.2,0.07509632110595703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,4,16,power_law_1.2,0.08252287864685058
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,4,16,power_law_1.2,0.07919487953186036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,4,16,power_law_1.2,0.0769702386856079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,4,16,power_law_1.2,0.08130816459655762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,4,16,power_law_1.2,0.0849407958984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,4,16,power_law_1.2,0.09127424240112306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,4,16,power_law_1.2,0.1002355194091797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,4,16,power_law_1.2,0.10585472106933594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,4,16,power_law_1.2,0.14911359786987305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,4,16,power_law_1.2,0.18411392211914063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,4,16,power_law_1.2,0.22484479904174806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,4,16,power_law_1.2,0.2712524795532226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,4,16,power_law_1.2,0.4300697708129883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,4,16,power_law_1.2,0.562946548461914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,4,16,power_law_1.01,0.04850560188293457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,4,16,power_law_1.2,0.027654399871826173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,4,16,power_law_1.2,0.8972531127929688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,4,16,power_law_1.2,0.031923201084136966
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,4,16,power_law_1.2,0.032072958946228025
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,4,16,power_law_1.01,0.09385343551635741
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,4,16,power_law_1.2,0.03163264036178589
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,4,16,power_law_1.2,0.032547841072082515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,4,16,power_law_1.2,1.0665984344482422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,4,16,power_law_1.2,0.04232192039489746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,4,16,power_law_1.2,0.039496960639953616
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,4,16,power_law_1.01,0.055068159103393556
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,4,16,power_law_1.2,0.03913856029510498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,4,16,power_law_1.2,0.0410316801071167
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,4,16,power_law_1.2,0.04113024234771728
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,4,16,power_law_1.2,0.04328447818756104
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,4,16,power_law_1.2,0.04560256004333496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,4,16,power_law_1.2,0.04635392189025879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,4,16,power_law_1.2,0.04439295768737793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,4,16,power_law_1.2,0.04626431941986084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,4,16,power_law_1.2,0.047535362243652346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,4,16,power_law_1.2,0.05321472167968751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,4,16,power_law_1.2,0.05719039916992188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,4,16,power_law_1.2,0.06150144100189209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,4,16,power_law_1.2,0.09424511909484863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,4,16,power_law_1.2,0.10290559768676757
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,4,16,power_law_1.2,0.13951871871948243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,4,16,power_law_1.2,0.17676671981811523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,4,16,power_law_1.2,0.24665344238281248
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,4,16,power_law_1.2,0.019842560291290286
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,4,16,power_law_1.2,0.3793766403198242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,4,16,power_law_1.2,0.02154112100601196
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,4,16,power_law_1.2,0.021155838966369626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,4,16,power_law_1.01,0.0798028802871704
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,4,16,power_law_1.2,0.021079039573669432
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,4,16,power_law_1.2,0.5776486587524414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,4,16,power_law_1.2,0.02150399923324585
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,4,16,power_law_1.2,0.02064768075942993
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,4,16,power_law_1.2,0.02183295965194702
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,4,16,power_law_1.2,0.021040639877319335
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,4,16,power_law_1.2,0.02120448112487793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,4,16,power_law_1.2,0.021427199840545655
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,4,16,power_law_1.2,0.021405439376831054
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,4,16,power_law_1.2,0.02182528018951416
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,4,16,power_law_1.2,0.022301440238952634
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,4,16,power_law_1.2,0.02267904043197632
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,4,16,power_law_1.2,0.02316416025161743
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,4,16,power_law_1.2,0.029024000167846682
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,4,16,power_law_1.2,0.6639078521728516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,4,16,power_law_1.2,0.021932799816131592
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,4,16,power_law_1.01,0.14610303878784178
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,4,16,power_law_1.01,0.1665177536010742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,4,16,power_law_1.2,0.09345791816711427
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,4,16,power_law_1.2,0.0261516809463501
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,4,16,power_law_1.2,0.03826816082000732
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,4,16,power_law_1.2,0.046812157630920406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,8,8,balanced,0.022663679122924805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,4,16,power_law_1.2,0.03371903896331787
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,8,8,balanced,0.02346496105194092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,8,8,balanced,0.022732799053192136
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,8,8,balanced,0.02636415958404541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,8,8,balanced,0.02968064069747925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,8,8,balanced,0.05281023979187012
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,8,8,balanced,0.07796864032745361
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,8,8,balanced,0.0805568027496338
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,8,8,balanced,0.08412544250488281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,8,8,balanced,0.08758399963378907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,8,8,balanced,0.07696512222290039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,8,8,balanced,0.08075263977050781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,8,8,balanced,0.08271360397338867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,8,8,balanced,0.08693120002746582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,8,8,balanced,0.09084927558898925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,8,8,balanced,0.09298432350158692
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,8,8,balanced,0.09889151573181151
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,8,8,balanced,0.11436544418334962
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,8,8,balanced,0.12870143890380858
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,8,8,balanced,0.15974783897399902
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,4,16,power_law_1.2,0.05917439937591553
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,8,8,balanced,0.2522764778137207
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,4,16,power_law_1.2,0.08026495933532715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,8,8,balanced,0.3133465576171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,8,8,balanced,0.4344524765014649
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,8,8,balanced,0.026033918857574462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,8,8,balanced,0.5589785766601563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,8,8,balanced,0.026777598857879635
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,8,8,balanced,0.8256076812744141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,8,8,balanced,0.028389120101928712
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,8,8,balanced,0.1894540786743164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,8,8,balanced,0.03420799970626831
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,8,8,balanced,0.03054464101791382
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,8,8,balanced,1.0788569641113281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,8,8,balanced,0.041626877784729004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,8,8,balanced,0.047828478813171385
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,8,8,balanced,0.04965888023376465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,8,8,balanced,0.05095295906066895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,8,8,balanced,0.04780159950256348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,8,8,balanced,0.04972544193267822
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,8,8,balanced,0.05061759948730469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,8,8,balanced,0.05395711898803711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,8,8,balanced,0.05404799938201904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,8,8,balanced,0.05712255954742431
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,8,8,balanced,0.06408448219299316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,8,8,balanced,0.07386367797851563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,8,8,balanced,0.08656384468078612
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,8,8,balanced,0.10602879524230957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,8,8,balanced,0.12600192070007324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,8,8,balanced,0.16793855667114257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,8,8,balanced,0.20617599487304689
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,8,8,balanced,0.30149887084960936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,8,8,balanced,0.053451519012451175
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,8,8,balanced,0.3808294296264648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,8,8,balanced,0.5576742553710938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,8,8,balanced,0.7246361541748046
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,4,16,power_law_1.2,0.12378623962402344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,4,16,power_law_1.2,0.194465274810791
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,4,16,power_law_1.2,0.2419635200500488
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,8,8,balanced,0.021043200492858887
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,8,8,balanced,0.02726272106170654
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,8,8,balanced,0.02723072052001953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,8,8,balanced,0.027242240905761717
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,8,8,balanced,0.021448960304260252
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,8,8,balanced,0.027658240795135496
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,8,8,balanced,0.028071680068969727
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,8,8,balanced,0.02803071975708008
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,8,8,balanced,0.028751358985900876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,8,8,balanced,0.02839168071746826
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,8,8,balanced,0.028833279609680174
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,8,8,balanced,0.022243840694427492
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,8,8,balanced,0.029667840003967282
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,8,8,balanced,0.031263999938964844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,8,8,balanced,0.032503039836883546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,8,8,balanced,0.023658239841461183
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,8,8,balanced,0.03719552040100098
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,8,8,balanced,0.04201216220855713
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,8,8,balanced,0.0337228798866272
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,8,8,balanced,0.020678400993347168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,8,8,power_law_1.01,0.023975679874420168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,8,8,power_law_1.01,0.028007678985595703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,8,8,power_law_1.01,0.039665920734405516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,8,8,power_law_1.01,0.03993983983993531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,8,8,power_law_1.01,0.04470399856567383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,8,8,power_law_1.01,0.04845695972442627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,8,8,power_law_1.01,0.07259520053863525
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,8,8,power_law_1.01,0.07369344234466553
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,8,8,power_law_1.01,0.07681024074554443
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,8,8,power_law_1.01,0.07810815811157226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,8,8,power_law_1.01,0.08120960235595703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,8,8,power_law_1.01,0.08309120178222656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,8,8,balanced,0.027216639518737797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,8,8,balanced,0.027210240364074705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,8,8,power_law_1.01,0.07596415996551513
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,8,8,power_law_1.01,0.07992959976196289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,8,8,power_law_1.01,0.08243840217590333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,8,8,power_law_1.01,0.08617983818054199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,8,8,power_law_1.01,0.09005824089050293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,8,8,power_law_1.01,0.10646783828735351
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,8,8,power_law_1.01,0.11960960388183595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,8,8,power_law_1.01,0.1425369644165039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,8,8,power_law_1.01,0.17347328186035155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,8,8,power_law_1.01,0.21960704803466796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,8,8,power_law_1.01,0.2737459182739258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,8,8,power_law_1.01,0.3993049621582031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,8,8,power_law_1.01,0.5122137451171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,8,8,power_law_1.01,0.027631359100341795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,8,8,power_law_1.01,0.777364501953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,8,8,power_law_1.01,0.03240576028823853
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,8,8,power_law_1.01,0.034556159973144526
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,8,8,power_law_1.01,0.03377664089202881
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,8,8,power_law_1.01,1.043799057006836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,8,8,power_law_1.01,0.038624000549316403
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,8,8,power_law_1.01,0.03862656116485595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,8,8,power_law_1.01,0.04608128070831299
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,8,8,power_law_1.01,0.04649472236633301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,8,8,power_law_1.01,0.04803455829620361
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,8,8,power_law_1.01,0.0493939208984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,8,8,power_law_1.01,0.05174784183502197
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,8,8,power_law_1.01,0.0501529598236084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,8,8,power_law_1.01,0.04932352066040039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,8,8,power_law_1.01,0.05219583988189698
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,8,8,power_law_1.01,0.054327039718627936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,8,8,power_law_1.01,0.05640448093414306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,8,8,power_law_1.01,0.05916543960571289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,8,8,power_law_1.01,0.06956416130065918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,8,8,power_law_1.01,0.07933311939239503
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,8,8,power_law_1.01,0.09952768325805664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,8,8,power_law_1.01,0.1161996841430664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,8,8,power_law_1.01,0.15953920364379884
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,8,8,power_law_1.01,0.19035776138305666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,8,8,power_law_1.01,0.25763456344604496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,8,8,power_law_1.01,0.35725566864013675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,8,8,power_law_1.01,0.4839129638671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,8,8,power_law_1.01,0.022452480792999267
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,8,8,power_law_1.01,0.6986316680908203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,8,8,power_law_1.01,0.02370431900024414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,8,8,power_law_1.01,0.02594304084777832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,8,8,power_law_1.01,0.026245119571685793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,8,8,power_law_1.01,0.02568576097488403
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,8,8,power_law_1.01,0.027180800437927245
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,8,8,power_law_1.01,0.026632959842681887
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,8,8,power_law_1.01,0.027043840885162353
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,8,8,power_law_1.01,0.028034560680389405
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,8,8,power_law_1.01,0.027045118808746337
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,8,8,power_law_1.01,0.027307519912719725
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,8,8,power_law_1.01,0.02764287948608398
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,8,8,power_law_1.01,0.02769407987594604
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,8,8,power_law_1.01,0.0278656005859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,8,8,power_law_1.01,0.029361920356750486
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,8,8,power_law_1.01,0.02969599962234497
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,8,8,power_law_1.01,0.031470079421997074
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,8,8,balanced,0.028456959724426266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,8,8,balanced,0.06576255798339845
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,8,8,balanced,0.06001152038574219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,8,8,balanced,0.08309760093688964
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,8,8,balanced,0.10356096267700195
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,8,8,power_law_1.01,0.036161279678344725
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,8,8,power_law_1.01,0.04788864135742187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,8,8,power_law_1.01,0.040331521034240723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,8,8,power_law_1.2,0.02437376022338867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,8,8,power_law_1.01,0.056352000236511234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,8,8,power_law_1.2,0.027617280483245847
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,8,8,power_law_1.2,0.03544447898864746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,8,8,power_law_1.2,0.03957887887954712
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,8,8,power_law_1.2,0.04771584033966064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,8,8,power_law_1.2,0.04591104030609131
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,8,8,power_law_1.2,0.07310080051422119
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,8,8,power_law_1.2,0.07401855945587157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,8,8,power_law_1.2,0.076212477684021
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,8,8,power_law_1.2,0.07882880210876465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,8,8,power_law_1.2,0.0857267189025879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,8,8,power_law_1.2,0.07990143775939942
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,8,8,power_law_1.2,0.07776383876800537
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,8,8,power_law_1.2,0.08010751724243163
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,8,8,power_law_1.2,0.08299519538879394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,8,8,power_law_1.2,0.08626815795898438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,8,8,power_law_1.2,0.09570176124572753
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,8,8,power_law_1.2,0.10797439575195314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,8,8,power_law_1.2,0.12401791572570801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,8,8,power_law_1.2,0.15204095840454102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,8,8,power_law_1.2,0.18214656829833983
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,8,8,power_law_1.2,0.22512384414672854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,8,8,power_law_1.2,0.28036224365234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,8,8,power_law_1.2,0.4364467239379882
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,8,8,power_law_1.2,0.5762675094604492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,8,8,power_law_1.2,0.8679475402832031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,8,8,power_law_1.2,0.027617280483245847
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,8,8,power_law_1.2,1.0581375885009767
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,8,8,power_law_1.2,0.031744000911712644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,8,8,power_law_1.2,0.03267328023910522
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,8,8,power_law_1.2,0.03375999927520752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,8,8,power_law_1.2,0.03952640056610107
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,8,8,power_law_1.2,0.037893118858337405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,8,8,power_law_1.2,0.04657663822174073
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,8,8,power_law_1.2,0.04548736095428467
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,8,8,power_law_1.01,0.07401599884033203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,8,8,power_law_1.2,0.04869631767272949
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,8,8,power_law_1.2,0.04829823970794678
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,8,8,power_law_1.2,0.0530675220489502
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,8,8,power_law_1.2,0.05221759796142578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,8,8,power_law_1.2,0.05190271854400634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,8,8,power_law_1.2,0.05177087783813476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,8,8,power_law_1.2,0.05468287944793702
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,8,8,power_law_1.2,0.056248321533203124
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,8,8,power_law_1.01,0.14348671913146974
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,8,8,power_law_1.2,0.06122879981994629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,8,8,power_law_1.2,0.06959231853485107
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,8,8,power_law_1.01,0.0843238353729248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,8,8,power_law_1.2,0.08323455810546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,8,8,power_law_1.2,0.10478976249694824
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,8,8,power_law_1.01,0.11886079788208008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,8,8,power_law_1.2,0.12268159866333009
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,8,8,power_law_1.2,0.17085439682006837
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,8,8,power_law_1.2,0.2164300727844238
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,8,8,power_law_1.2,0.02247040033340454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,8,8,power_law_1.2,0.31464960098266603
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,8,8,power_law_1.2,0.02356352090835571
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,8,8,power_law_1.2,0.02388607978820801
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,8,8,power_law_1.2,0.0245580792427063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,8,8,power_law_1.2,0.5552640151977539
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,8,8,power_law_1.2,0.026888959407806397
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,8,8,power_law_1.2,0.7443981170654297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,8,8,power_law_1.2,0.027971839904785155
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,8,8,power_law_1.2,0.02676991939544678
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,8,8,power_law_1.2,0.026781439781188965
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,8,8,power_law_1.2,0.02667648077011108
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,8,8,power_law_1.2,0.02759936094284058
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,8,8,power_law_1.2,0.028334081172943115
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,8,8,power_law_1.2,0.027654399871826173
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,8,8,power_law_1.2,0.028856320381164553
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,8,8,power_law_1.2,0.029692161083221435
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,8,8,power_law_1.2,0.030952959060668944
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,8,8,power_law_1.2,0.03572607994079589
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,8,8,power_law_1.2,0.3820518493652344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,8,8,power_law_1.2,0.026446080207824706
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,8,8,power_law_1.01,0.19912319183349608
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,8,8,power_law_1.01,0.278606071472168
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,8,8,power_law_1.2,0.07076352119445802
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,8,8,power_law_1.2,0.055404801368713376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,8,8,power_law_1.2,0.09157888412475586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,8,8,power_law_1.2,0.03965568065643311
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,8,8,power_law_1.2,0.0456063985824585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,16,4,balanced,0.022242560386657714
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,16,4,balanced,0.021954560279846193
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,16,4,balanced,0.02392064094543457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,16,4,balanced,0.02926464080810547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,16,4,balanced,0.03206399917602539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,16,4,balanced,0.05321856021881104
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,16,4,balanced,0.05336575984954835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,16,4,balanced,0.05492735862731933
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,16,4,balanced,0.08163455963134766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,16,4,balanced,0.08286463737487793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,16,4,balanced,0.08453503608703614
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,16,4,balanced,0.08744576454162598
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,16,4,balanced,0.0897049617767334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,16,4,balanced,0.09183103561401366
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,16,4,balanced,0.09450112342834473
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,16,4,balanced,0.09816448211669922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,16,4,balanced,0.10529151916503907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,16,4,balanced,0.12042752265930176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,16,4,balanced,0.13591551780700684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,16,4,balanced,0.16482175827026366
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,16,4,balanced,0.1928870391845703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,16,4,balanced,0.2539430427551269
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,16,4,balanced,0.3120780754089355
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,8,8,power_law_1.2,0.11118975639343262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,16,4,balanced,0.42855808258056644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,16,4,balanced,0.026822400093078614
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,16,4,balanced,0.5481868743896484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,16,4,balanced,0.026448640823364254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,16,4,balanced,0.029272320270538332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,16,4,balanced,0.802929916381836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,16,4,balanced,0.03543807983398438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,16,4,balanced,0.04118271827697754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,16,4,balanced,0.048485121726989745
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,16,4,balanced,0.050196480751037595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,16,4,balanced,0.048657917976379396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,16,4,balanced,0.05796224117279053
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,16,4,balanced,0.05852287769317627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,16,4,balanced,0.06025856018066407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,16,4,balanced,0.06163839817047119
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,16,4,balanced,0.06348415851593017
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,16,4,balanced,0.06564864158630371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,16,4,balanced,0.06774784088134765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,16,4,balanced,0.07343103885650634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,16,4,balanced,0.08499072074890138
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,16,4,balanced,0.09477503776550293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,16,4,balanced,0.1139686393737793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,16,4,balanced,0.1314803218841553
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,16,4,balanced,0.05828991889953613
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,16,4,balanced,1.058319396972656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,16,4,balanced,0.17038463592529296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,16,4,balanced,0.2096985626220703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,16,4,balanced,0.30417919158935547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,16,4,balanced,0.38041473388671876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,16,4,balanced,0.5597747039794922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,16,4,balanced,0.7289177703857421
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,8,8,power_law_1.2,0.16100223541259767
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,8,8,power_law_1.2,0.20991615295410154
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,8,8,power_law_1.2,0.32319103240966796
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,8,8,power_law_1.2,0.4745344161987305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,16,4,balanced,0.02841599941253662
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,16,4,balanced,0.04705920219421387
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,16,4,balanced,0.03046655893325806
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,16,4,balanced,0.04660736083984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,16,4,balanced,0.046649599075317384
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,16,4,balanced,0.04658048152923584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,16,4,balanced,0.029239039421081543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,16,4,balanced,0.04676864147186279
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,16,4,balanced,0.04696832180023193
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,16,4,balanced,0.04712831974029541
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,16,4,balanced,0.04944767951965332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,16,4,balanced,0.049651198387145996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,16,4,balanced,0.05297023773193359
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,16,4,balanced,0.05566336154937744
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,16,4,balanced,0.057631998062133795
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,16,4,balanced,0.06262527942657471
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,16,4,balanced,0.02798847913742065
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,16,4,balanced,0.06812543869018554
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,16,4,balanced,0.03374592065811157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,16,4,balanced,0.0447488021850586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,16,4,power_law_1.01,0.02312704086303711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,16,4,power_law_1.01,0.02592895984649658
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,16,4,power_law_1.01,0.031275520324707026
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,16,4,power_law_1.01,0.037884159088134764
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,16,4,power_law_1.01,0.04231679916381836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,16,4,power_law_1.01,0.04497663974761963
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,16,4,power_law_1.01,0.05032959938049316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,16,4,power_law_1.01,0.048355841636657716
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,16,4,power_law_1.01,0.07686783790588378
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,16,4,power_law_1.01,0.07809152126312255
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,16,4,power_law_1.01,0.07942272186279296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,16,4,power_law_1.01,0.08439167976379394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,16,4,power_law_1.01,0.08615424156188964
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,16,4,power_law_1.01,0.08984576225280762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,16,4,power_law_1.01,0.0927359962463379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,16,4,power_law_1.01,0.09645055770874024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,16,4,power_law_1.01,0.10347007751464843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,16,4,power_law_1.01,0.11973888397216796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,16,4,power_law_1.01,0.13387136459350585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,16,4,power_law_1.01,0.16564607620239258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,16,4,power_law_1.01,0.1910425567626953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,16,4,power_law_1.01,0.24650239944458008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,16,4,power_law_1.01,0.30609024047851563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,16,4,balanced,0.044581117630004885
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,16,4,power_law_1.01,0.4254272079467773
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,16,4,balanced,0.04655231952667237
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,16,4,power_law_1.01,0.5718822479248047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,16,4,power_law_1.01,0.026050560474395752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,16,4,power_law_1.01,0.030145280361175537
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,16,4,power_law_1.01,0.8068889617919922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,16,4,power_law_1.01,0.03365888118743897
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,16,4,power_law_1.01,1.0623731231689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,16,4,power_law_1.01,0.037406721115112306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,16,4,power_law_1.01,0.039888639450073246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,16,4,power_law_1.01,0.04481919765472412
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,16,4,power_law_1.01,0.04273536205291748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,16,4,power_law_1.01,0.04487423896789551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,16,4,power_law_1.01,0.05276415824890137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,16,4,power_law_1.01,0.05388671874999999
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,16,4,power_law_1.01,0.05503744125366211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,16,4,power_law_1.01,0.05664256095886231
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,16,4,power_law_1.01,0.05855999946594238
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,16,4,power_law_1.01,0.060762882232666016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,16,4,power_law_1.01,0.06464000225067139
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,16,4,power_law_1.01,0.06634880065917968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,16,4,power_law_1.01,0.07229184150695801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,16,4,power_law_1.01,0.08042495727539063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,16,4,power_law_1.01,0.09417856216430665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,16,4,power_law_1.01,0.11481087684631348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,16,4,power_law_1.01,0.1369536018371582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,16,4,power_law_1.01,0.18278656005859376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,16,4,power_law_1.01,0.2247590446472168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,16,4,power_law_1.01,0.3191654396057129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,16,4,power_law_1.01,0.4093081665039063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,16,4,power_law_1.01,0.029667840003967282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,16,4,power_law_1.01,0.5619622421264648
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,16,4,power_law_1.01,0.03131392002105713
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,16,4,power_law_1.01,0.03300096035003662
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,16,4,power_law_1.01,0.7469197082519531
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,16,4,power_law_1.01,0.03460479974746704
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,16,4,power_law_1.01,0.03584768056869507
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,16,4,power_law_1.01,0.0374617600440979
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,16,4,power_law_1.01,0.04171135902404785
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,16,4,power_law_1.01,0.04181248188018799
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,16,4,power_law_1.01,0.04131328105926514
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,16,4,power_law_1.01,0.04317952156066894
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,16,4,power_law_1.01,0.04272384166717529
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,16,4,power_law_1.01,0.04276735782623291
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,16,4,power_law_1.01,0.04464767932891846
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,16,4,power_law_1.01,0.045483517646789554
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,16,4,power_law_1.01,0.04695680141448975
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,16,4,power_law_1.01,0.04910079956054687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,16,4,power_law_1.01,0.05255807876586914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,16,4,balanced,0.09807744026184081
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,16,4,balanced,0.15297408103942872
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,16,4,balanced,0.11166080474853515
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,16,4,balanced,0.19084543228149414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,16,4,power_law_1.01,0.08353792190551758
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,16,4,power_law_1.01,0.06743296146392821
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,16,4,power_law_1.01,0.06146431922912597
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,16,4,power_law_1.01,0.09809920310974121
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,16,4,power_law_1.01,0.11907711982727051
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,16,4,power_law_1.2,0.022693119049072265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,16,4,power_law_1.2,0.025122559070587157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,16,4,power_law_1.2,0.03219327926635742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,16,4,power_law_1.2,0.03860352039337158
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,16,4,power_law_1.01,0.14662528038024902
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,16,4,power_law_1.2,0.04167808055877685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,16,4,power_law_1.2,0.04414591789245605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,16,4,power_law_1.2,0.04874239921569824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,16,4,power_law_1.2,0.047978239059448244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,16,4,power_law_1.2,0.07764607906341553
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,16,4,power_law_1.2,0.07776383876800537
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,16,4,power_law_1.2,0.08066944122314454
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,16,4,power_law_1.01,0.1854092788696289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,16,4,power_law_1.2,0.08326784133911133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,16,4,power_law_1.2,0.08543359756469726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,16,4,power_law_1.2,0.0891750431060791
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,16,4,power_law_1.2,0.09661952018737793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,16,4,power_law_1.2,0.10464127540588379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,16,4,power_law_1.2,0.11954688072204589
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,16,4,power_law_1.2,0.13617919921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,16,4,power_law_1.2,0.16583295822143557
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,16,4,power_law_1.2,0.19634176254272462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,16,4,power_law_1.2,0.2543424034118652
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,16,4,power_law_1.2,0.3212428665161133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,16,4,power_law_1.2,0.09437952041625977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,16,4,power_law_1.2,0.42493953704833987
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,16,4,power_law_1.2,0.5720896148681641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,16,4,power_law_1.2,0.026357760429382326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,16,4,power_law_1.2,0.03011967897415161
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,16,4,power_law_1.2,0.8065497589111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,16,4,power_law_1.2,0.031256320476531985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,16,4,power_law_1.2,1.0815321350097657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,16,4,power_law_1.2,0.03705984115600586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,16,4,power_law_1.2,0.04014207839965821
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,16,4,power_law_1.2,0.04238080024719239
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,16,4,power_law_1.2,0.04356736183166504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,16,4,power_law_1.2,0.04446208000183106
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,16,4,power_law_1.2,0.054364161491394045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,16,4,power_law_1.2,0.05581952095031738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,16,4,power_law_1.2,0.05391104221343994
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,16,4,power_law_1.2,0.056979198455810544
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,16,4,power_law_1.2,0.05977983951568604
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,16,4,power_law_1.2,0.06472064018249511
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,16,4,power_law_1.2,0.06572415828704833
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,16,4,power_law_1.2,0.07353856086730956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,16,4,power_law_1.2,0.0846463966369629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,16,4,power_law_1.2,0.09617792129516602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,16,4,power_law_1.2,0.11825152397155762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,16,4,power_law_1.2,0.13996031761169433
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,16,4,power_law_1.2,0.18718719482421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,16,4,power_law_1.2,0.23116416931152345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,16,4,power_law_1.2,0.3236185455322266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,16,4,power_law_1.2,0.06249087810516357
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,16,4,power_law_1.2,0.4221977615356446
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,16,4,power_law_1.2,0.03009151935577393
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,16,4,power_law_1.2,0.030922238826751706
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,16,4,power_law_1.2,0.5951103973388672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,16,4,power_law_1.2,0.031231999397277832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,16,4,power_law_1.2,0.03340159893035889
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,16,4,power_law_1.2,0.035846400260925296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,16,4,power_law_1.2,0.784031982421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,16,4,power_law_1.2,0.037740800380706784
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,16,4,power_law_1.2,0.03952383995056152
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,16,4,power_law_1.2,0.04058879852294922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,16,4,power_law_1.2,0.0421504020690918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,16,4,power_law_1.2,0.04357120037078858
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,16,4,power_law_1.2,0.0445030403137207
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,16,4,power_law_1.2,0.044582400321960446
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,16,4,power_law_1.2,0.04656511783599854
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,16,4,power_law_1.2,0.04770559787750244
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,16,4,power_law_1.2,0.049180159568786616
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,16,4,power_law_1.2,0.059102721214294426
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,16,4,power_law_1.2,0.06836607933044433
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,16,4,power_law_1.2,0.08179455757141113
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,16,4,power_law_1.2,0.16288639068603517
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,16,4,power_law_1.2,0.040757761001586915
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,16,4,power_law_1.01,0.2500876808166504
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,16,4,power_law_1.01,0.3623769760131836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,16,4,power_law_1.2,0.11546624183654786
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,16,4,power_law_1.2,0.19309823989868163
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,16,4,power_law_1.2,0.09932160377502441
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,32,2,balanced,0.021914880275726318
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,32,2,balanced,0.022695679664611817
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,32,2,balanced,0.028582398891448972
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,16,4,power_law_1.2,0.3194790458679199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,32,2,balanced,0.038227200508117676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,32,2,balanced,0.062024960517883296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,32,2,balanced,0.06208640098571777
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,32,2,balanced,0.06303487777709961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,32,2,balanced,0.06395520210266113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,32,2,balanced,0.06443647861480713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,32,2,balanced,0.06575232028961182
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,32,2,balanced,0.1052518367767334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,32,2,balanced,0.10668543815612792
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,32,2,balanced,0.025588479042053226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,32,2,balanced,0.1103769588470459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,32,2,balanced,0.11292927742004395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,32,2,balanced,0.1163315200805664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,32,2,balanced,0.13764863967895508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,32,2,balanced,0.15171711921691894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,32,2,balanced,0.17940095901489256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,32,2,balanced,0.20830080032348636
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,32,2,balanced,0.2692441558837891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,32,2,balanced,0.3249715042114258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,32,2,balanced,0.44296447753906254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,32,2,balanced,0.5823411178588868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,32,2,balanced,0.8325119781494141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,32,2,balanced,0.02724479913711548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,32,2,balanced,0.12354816436767577
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,32,2,balanced,0.027276799678802487
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,32,2,balanced,0.030094079971313477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,32,2,balanced,1.0766438293457032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,32,2,balanced,0.03871999979019165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,32,2,balanced,0.05439487934112549
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,32,2,balanced,0.06857984066009522
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,32,2,balanced,0.06864128112792969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,32,2,balanced,0.06896768093109132
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,32,2,balanced,0.0698035192489624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,32,2,balanced,0.07145088195800782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,32,2,balanced,0.08818047523498536
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,32,2,balanced,0.08851455688476563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,32,2,balanced,0.09126655578613281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,32,2,balanced,0.09195775985717773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,32,2,balanced,0.0951795196533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,32,2,balanced,0.06865664005279541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,32,2,balanced,0.09938176155090332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,32,2,balanced,0.1086732769012451
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,32,2,balanced,0.11783424377441407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,32,2,balanced,0.13633791923522948
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,32,2,balanced,0.1541043186187744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,32,2,balanced,0.19131263732910156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,32,2,balanced,0.2279052734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,32,2,balanced,0.35208065032958985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,32,2,balanced,0.42362110137939457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,32,2,balanced,0.6157798385620117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,32,2,power_law_1.01,0.022481920719146727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,32,2,balanced,0.8150502777099609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,32,2,power_law_1.01,0.025176320075988766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,32,2,power_law_1.01,0.03782655954360962
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,32,2,power_law_1.01,0.046014719009399414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,32,2,power_law_1.01,0.050737919807434084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,32,2,power_law_1.01,0.05444479942321777
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,32,2,power_law_1.01,0.05504127979278565
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,16,4,power_law_1.2,0.3382361602783203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,32,2,power_law_1.01,0.059208960533142084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,32,2,power_law_1.01,0.06001023769378662
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,32,2,power_law_1.01,0.10240768432617187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,16,4,power_law_1.2,0.5894118499755859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,32,2,power_law_1.01,0.027235839366912838
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,32,2,power_law_1.01,0.1081497573852539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,32,2,power_law_1.01,0.1048089599609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,32,2,power_law_1.01,0.11519616127014161
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,32,2,power_law_1.01,0.11087871551513673
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,32,2,power_law_1.01,0.12266240119934083
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,32,2,power_law_1.01,0.1379648017883301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,32,2,power_law_1.01,0.15428863525390626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,32,2,power_law_1.01,0.18445568084716796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,32,2,power_law_1.01,0.21330944061279297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,32,2,power_law_1.01,0.27386367797851563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,32,2,power_law_1.01,0.33550209045410156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,32,2,power_law_1.01,0.05813888072967529
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,32,2,power_law_1.01,0.4731814575195313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,32,2,power_law_1.01,0.026785280704498295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,32,2,power_law_1.01,0.607850227355957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,32,2,power_law_1.01,0.030201599597930905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,32,2,power_law_1.01,0.03386240005493164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,32,2,power_law_1.01,0.04071040153503418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,32,2,power_law_1.01,0.8586649322509766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,32,2,power_law_1.01,0.05140351772308349
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,32,2,power_law_1.01,0.05547008037567138
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,32,2,power_law_1.01,0.06257791996002197
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,32,2,power_law_1.01,1.121454086303711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,32,2,power_law_1.01,0.06453120231628418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,32,2,power_law_1.01,0.06531712055206298
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,32,2,power_law_1.01,0.0844211196899414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,32,2,power_law_1.01,0.08210687637329102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,32,2,power_law_1.01,0.08309760093688964
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,32,2,power_law_1.01,0.05954559803009033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,32,2,power_law_1.01,0.09600640296936035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,32,2,power_law_1.01,0.10728063583374023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,32,2,power_law_1.01,0.1213798427581787
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,32,2,power_law_1.01,0.14902527809143068
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,32,2,power_law_1.01,0.18013439178466797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,32,2,power_law_1.01,0.0843507194519043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,32,2,power_law_1.01,0.22650495529174805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,32,2,power_law_1.01,0.2745408058166504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,32,2,power_law_1.01,0.06282112121582031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,32,2,power_law_1.01,0.0874841594696045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,32,2,power_law_1.2,0.022755839824676515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,32,2,power_law_1.01,0.3751295852661133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,32,2,power_law_1.01,0.47373313903808595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,32,2,power_law_1.2,0.025169920921325684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,32,2,power_law_1.2,0.027435519695281983
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,32,2,power_law_1.2,0.037825279235839844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,32,2,power_law_1.01,0.6700057220458985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,32,2,power_law_1.2,0.04482944011688232
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,32,2,power_law_1.2,0.047975678443908695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,32,2,power_law_1.01,0.8720294189453124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,32,2,power_law_1.2,0.05309567928314209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,32,2,power_law_1.2,0.058974719047546385
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,32,2,power_law_1.2,0.059362559318542486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,32,2,power_law_1.2,0.10245759963989258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,32,2,power_law_1.2,0.1045644760131836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,32,2,power_law_1.2,0.1083788776397705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,32,2,power_law_1.2,0.1122316837310791
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,32,2,power_law_1.2,0.1154700756072998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,32,2,power_law_1.2,0.12321279525756837
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,32,2,power_law_1.2,0.1390067195892334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,32,2,power_law_1.2,0.15519359588623047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,32,2,power_law_1.2,0.18600704193115233
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,32,2,power_law_1.2,0.05246592044830323
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,32,2,power_law_1.2,0.21553920745849608
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,32,2,power_law_1.2,0.2738137626647949
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,32,2,power_law_1.2,0.33902721405029296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,32,2,power_law_1.2,0.47712257385253903
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,32,2,power_law_1.2,0.6065318298339843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,32,2,power_law_1.2,0.025949440002441405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,32,2,power_law_1.2,0.030076160430908206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,32,2,power_law_1.2,0.8660787200927734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,32,2,power_law_1.2,0.03211008071899414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,32,2,power_law_1.2,0.057483520507812504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,32,2,power_law_1.2,0.03872128009796143
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,32,2,power_law_1.2,0.052170238494873046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,32,2,power_law_1.2,1.1363455963134765
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,16,4,power_law_1.2,0.7941798400878907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,32,2,power_law_1.2,0.05960576057434082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,32,2,power_law_1.2,0.0611891222000122
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,32,2,power_law_1.2,0.06403327941894531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,32,2,power_law_1.2,0.06329599857330323
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,32,2,power_law_1.2,0.07993087768554688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,32,2,power_law_1.2,0.0808844757080078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,32,2,power_law_1.2,0.0836415958404541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,32,2,power_law_1.2,0.08818816184997559
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,32,2,power_law_1.2,0.09673600196838379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,32,2,power_law_1.2,0.11002880096435548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,32,2,power_law_1.2,0.05668992042541504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,32,2,power_law_1.2,0.12636032104492187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,32,2,power_law_1.2,0.155097599029541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,32,2,power_law_1.2,0.18111999511718752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,32,2,power_law_1.2,0.06538112163543701
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,32,2,power_law_1.2,0.23389184951782224
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,32,2,power_law_1.2,0.2816703987121582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,32,2,power_law_1.2,0.0865766429901123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,32,2,power_law_1.2,0.3877836990356445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,64,balanced,0.028012800216674804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,32,2,power_law_1.2,0.48718078613281246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,64,balanced,0.024798719882965087
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,64,balanced,0.028055040836334227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,64,balanced,0.042722558975219725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,32,2,power_law_1.2,0.6810700988769531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,64,balanced,0.0447705602645874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,64,balanced,0.045207037925720214
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,64,balanced,0.046443519592285154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,64,balanced,0.0472435188293457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,32,2,power_law_1.2,0.877434844970703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,64,balanced,0.04733695983886719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,64,balanced,0.048935680389404296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,64,balanced,0.051022081375122076
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,64,balanced,0.05247744083404541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,64,balanced,0.05372416019439698
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,64,balanced,0.05565055847167969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,64,balanced,0.05993855953216552
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,64,balanced,0.0672883176803589
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,64,balanced,0.08174976348876953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,64,balanced,0.046743040084838865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,64,balanced,0.09750911712646484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,64,balanced,0.11541376113891602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,64,balanced,0.12953344345092774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,64,balanced,0.05026815891265869
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,64,balanced,0.1782374382019043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,64,balanced,0.21388416290283202
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,64,balanced,0.02765952110290527
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,64,balanced,0.046443519592285154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,64,balanced,0.09056896209716797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,1,64,balanced,0.01773695945739746
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,1,64,balanced,0.017789440155029295
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,1,64,balanced,0.017822719812393188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,1,64,balanced,0.01814911961555481
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,1,64,balanced,0.018136320114135743
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,1,64,balanced,0.01774719953536987
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,1,64,balanced,0.01820672035217285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,1,64,balanced,0.018137600421905518
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,1,64,balanced,0.018141440153121947
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,1,64,balanced,0.01815808057785034
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,1,64,balanced,0.018187520503997804
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,1,64,balanced,0.01813375949859619
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,1,64,balanced,0.017740800380706787
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,1,64,balanced,0.018658560514450074
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,1,64,balanced,0.01847040057182312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,1,64,balanced,0.017889280319213868
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,1,64,balanced,0.018654719591140748
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,1,64,balanced,0.018570239543914797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,1,64,balanced,0.017756160497665405
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,1,64,balanced,0.017802239656448365
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,1,64,balanced,0.01705471992492676
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,1,64,balanced,0.019809279441833496
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,1,64,balanced,0.019878400564193727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,64,power_law_1.01,0.02847872018814087
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,64,power_law_1.01,0.027648000717163085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,64,power_law_1.01,0.04452991962432861
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,1,64,balanced,0.02603775978088379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,64,power_law_1.01,0.04525184154510498
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,1,64,balanced,0.02488192081451416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,64,power_law_1.01,0.04666751861572266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,64,power_law_1.01,0.04920191764831543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,1,64,balanced,0.03128448009490967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,64,power_law_1.01,0.055220479965209956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,64,power_law_1.01,0.05338111877441406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,64,power_law_1.01,0.055997438430786126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,64,power_law_1.01,0.055125761032104495
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,64,power_law_1.01,0.06308351993560792
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,64,power_law_1.01,0.07708415985107422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,64,power_law_1.01,0.07810815811157226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,64,power_law_1.01,0.08235520362854004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,64,power_law_1.01,0.09223936080932618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,64,power_law_1.01,0.10561792373657226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,64,power_law_1.01,0.12095999717712402
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,64,power_law_1.01,0.13825152397155763
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,64,power_law_1.01,0.20633983612060547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,64,power_law_1.01,0.25293056488037113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,64,power_law_1.01,0.048174080848693845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,64,power_law_1.01,0.3446745681762695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,64,power_law_1.01,0.05468287944793702
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,64,power_law_1.01,0.43133438110351563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,1,64,power_law_1.01,0.017812479734420777
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,1,64,power_law_1.01,0.017816319465637206
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,1,64,power_law_1.01,0.01775488018989563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,1,64,power_law_1.01,0.017798399925231932
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,1,64,power_law_1.01,0.017838079929351807
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,1,64,balanced,0.03616767883300781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,1,64,power_law_1.01,0.017816319465637206
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,1,64,power_law_1.01,0.017825280427932737
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,1,64,power_law_1.01,0.01814911961555481
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,64,power_law_1.01,0.9486361694335936
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,1,64,power_law_1.01,0.018135039806365966
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,1,64,power_law_1.01,0.01822335958480835
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,1,64,power_law_1.01,0.018236160278320312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,1,64,power_law_1.01,0.018205440044403075
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,1,64,power_law_1.01,0.018152960538864137
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,1,64,power_law_1.01,0.01817855954170227
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,1,64,power_law_1.01,0.018638080358505248
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,1,64,power_law_1.01,0.018388479948043823
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,1,64,power_law_1.01,0.019394559860229494
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,64,power_law_1.01,0.6608819580078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,64,power_law_1.01,0.02632575988769531
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,1,64,power_law_1.01,0.03088639974594116
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,1,64,power_law_1.01,0.03662336111068726
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,1,64,power_law_1.01,0.026309120655059814
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,1,64,power_law_1.01,0.044948477745056156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,1,64,power_law_1.01,0.019823360443115234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,1,64,power_law_1.01,0.02191871881484985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,64,power_law_1.2,0.027216639518737797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,64,power_law_1.2,0.028170239925384522
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,64,power_law_1.2,0.043968000411987306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,64,power_law_1.2,0.0460595178604126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,64,power_law_1.2,0.05069439888000489
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,64,power_law_1.2,0.0496127986907959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,64,power_law_1.2,0.05356800079345703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,64,power_law_1.2,0.05451007843017578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,64,power_law_1.2,0.0560038423538208
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,64,power_law_1.2,0.05858943939208985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,64,power_law_1.2,0.05773056030273438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,64,power_law_1.2,0.06948607921600342
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,64,power_law_1.2,0.07929728031158448
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,64,power_law_1.2,0.07219840049743652
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,64,power_law_1.2,0.0464358377456665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,64,power_law_1.2,0.08692480087280273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,64,power_law_1.2,0.10262528419494629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,64,power_law_1.2,0.11310208320617676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,64,power_law_1.2,0.13600000381469726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,64,power_law_1.2,0.163110408782959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,64,power_law_1.2,0.26487167358398434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,64,power_law_1.2,0.3638911819458008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,64,power_law_1.2,0.4784230422973633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,64,power_law_1.2,0.7196659088134766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,1,64,power_law_1.2,0.01784832000732422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,1,64,power_law_1.2,0.017789440155029295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,64,power_law_1.2,1.106789093017578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,1,64,power_law_1.01,0.05528448104858399
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,1,64,power_law_1.2,0.01775488018989563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,1,64,power_law_1.2,0.017799680233001706
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,1,64,power_law_1.2,0.01790336012840271
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,1,64,power_law_1.2,0.018129919767379758
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,1,64,power_law_1.2,0.018177280426025393
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,1,64,power_law_1.2,0.01815935969352722
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,1,64,power_law_1.2,0.01798143982887268
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,1,64,power_law_1.2,0.01819007992744446
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,1,64,power_law_1.2,0.01824512004852295
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,1,64,power_law_1.2,0.01825152039527893
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,1,64,power_law_1.2,0.018424320220947265
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,1,64,power_law_1.2,0.01900032043457031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,64,power_law_1.2,0.028083200454711917
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,1,64,power_law_1.2,0.017826559543609618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,64,power_law_1.2,1.6182566833496093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,1,64,power_law_1.01,0.11166591644287109
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,1,64,power_law_1.01,0.1519974422454834
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,1,64,power_law_1.01,0.0731712007522583
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,1,64,power_law_1.2,0.022709760665893555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,1,64,power_law_1.2,0.023137280941009523
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,1,64,power_law_1.2,0.02023168087005615
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,1,64,power_law_1.2,0.019398399591445924
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,1,64,power_law_1.2,0.028056321144104
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,32,balanced,0.021488640308380127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,32,balanced,0.021543679237365724
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,32,balanced,0.021518080234527587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,32,balanced,0.023555839061737062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,32,balanced,0.037544960975646975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,32,balanced,0.03909888029098511
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,32,balanced,0.03950720071792603
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,32,balanced,0.040501761436462405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,32,balanced,0.03951872110366821
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,32,balanced,0.04136447906494141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,32,balanced,0.04301951885223389
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,32,balanced,0.0435865592956543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,32,balanced,0.04522367954254151
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,32,balanced,0.04609663963317871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,32,balanced,0.04804096221923828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,32,balanced,0.0496230411529541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,32,balanced,0.04974847793579101
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,32,balanced,0.05889920234680176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,32,balanced,0.06138879776000976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,32,balanced,0.07195775985717774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,32,balanced,0.07318272113800048
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,32,balanced,0.08724479675292969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,32,balanced,0.10040960311889649
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,32,balanced,0.12652031898498534
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,32,balanced,0.14892031669616698
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,32,balanced,0.20765312194824218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,32,balanced,0.2655731201171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,1,64,power_law_1.2,0.03423104047775268
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,1,64,power_law_1.2,0.04754559993743897
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,1,64,power_law_1.2,0.06067840099334717
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,1,64,power_law_1.2,0.09090687751770019
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,1,64,power_law_1.2,0.1751923179626465
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,1,64,power_law_1.2,0.12895872116088866
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,1,64,power_law_1.2,0.25456384658813475
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,2,32,balanced,0.017756160497665405
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,2,32,balanced,0.018210560083389282
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,2,32,balanced,0.018195199966430663
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,2,32,balanced,0.018269439935684202
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,2,32,balanced,0.018576639890670776
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,2,32,balanced,0.018209279775619508
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,2,32,balanced,0.018347519636154174
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,2,32,balanced,0.018301440477371218
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,2,32,balanced,0.018608640432357788
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,2,32,balanced,0.018602240085601806
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,2,32,balanced,0.018684159517288208
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,2,32,balanced,0.018685439825057985
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,2,32,balanced,0.019425280094146728
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,2,32,balanced,0.01943935990333557
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,2,32,balanced,0.017838079929351807
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,2,32,balanced,0.017355519533157348
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,2,32,balanced,0.018182400465011597
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,2,32,balanced,0.01820799946784973
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,2,32,balanced,0.01823359966278076
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,2,32,balanced,0.018553600311279297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,32,power_law_1.01,0.024451839923858642
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,32,power_law_1.01,0.025126399993896487
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,32,power_law_1.01,0.025181438922882084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,32,power_law_1.01,0.03747967958450317
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,2,32,balanced,0.018265600204467776
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,32,power_law_1.01,0.039055359363555905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,32,power_law_1.01,0.03945856094360352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,32,power_law_1.01,0.039961600303649904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,32,power_law_1.01,0.04845695972442627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,32,power_law_1.01,0.04501632213592529
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,32,power_law_1.01,0.04766592025756836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,32,power_law_1.01,0.0482803201675415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,32,power_law_1.01,0.05754752159118652
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,32,power_law_1.01,0.05959167957305909
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,32,power_law_1.01,0.05855103969573975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,32,power_law_1.01,0.06345983982086181
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,32,power_law_1.01,0.06439551830291748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,32,power_law_1.01,0.06960256099700927
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,32,power_law_1.01,0.07662335872650147
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,32,power_law_1.01,0.09010815620422363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,32,power_law_1.01,0.10262784004211425
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,32,power_law_1.01,0.11978752136230468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,32,power_law_1.01,0.1743052864074707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,32,power_law_1.01,0.21003007888793945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,32,power_law_1.01,0.2855027198791504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,32,power_law_1.01,0.3533427047729492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,32,power_law_1.01,0.6225600051879883
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,2,32,power_law_1.01,0.018182400465011597
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,32,power_law_1.01,0.7267520141601562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,2,32,power_law_1.01,0.018250240087509157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,2,32,power_law_1.01,0.018236160278320312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,2,32,power_law_1.01,0.01816447973251343
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,2,32,power_law_1.01,0.0182259202003479
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,2,32,power_law_1.01,0.018216960430145264
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,2,32,power_law_1.01,0.01821568012237549
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,2,32,power_law_1.01,0.01821823954582214
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,2,32,power_law_1.01,0.018236160278320312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,2,32,power_law_1.01,0.01860991954803467
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,2,32,power_law_1.01,0.01856768012046814
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,2,32,power_law_1.01,0.0186188805103302
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,2,32,power_law_1.01,0.01895040035247803
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,2,32,power_law_1.01,0.018638080358505248
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,2,32,balanced,0.022293760776519775
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,2,32,power_law_1.01,0.018947839736938477
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,2,32,power_law_1.01,0.01983232021331787
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,2,32,power_law_1.01,0.018570239543914797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,2,32,balanced,0.022691841125488284
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,2,32,balanced,0.0279257607460022
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,2,32,balanced,0.028925440311431884
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,2,32,balanced,0.03877248048782349
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,2,32,balanced,0.044961280822753906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,2,32,power_law_1.01,0.021454079151153566
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,2,32,power_law_1.01,0.022159359455108642
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,2,32,power_law_1.01,0.027105278968811035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,32,power_law_1.2,0.024496641159057617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,32,power_law_1.2,0.024855039119720458
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,2,32,power_law_1.01,0.029232640266418457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,32,power_law_1.2,0.024963839054107668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,32,power_law_1.2,0.03866240024566651
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,32,power_law_1.2,0.03882496118545532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,32,power_law_1.2,0.04478079795837402
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,32,power_law_1.2,0.04175487995147705
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,2,32,power_law_1.01,0.037478399276733396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,32,power_law_1.2,0.04327936172485351
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,32,power_law_1.2,0.047740159034729
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,32,power_law_1.2,0.04946047782897949
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,32,power_law_1.2,0.06068992137908935
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,32,power_law_1.2,0.06025728225708008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,32,power_law_1.2,0.06811391830444335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,32,power_law_1.2,0.06925439834594727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,32,power_law_1.2,0.07204480171203613
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,32,power_law_1.2,0.07433343887329101
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,32,power_law_1.2,0.08371711730957031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,2,32,power_law_1.01,0.04371840000152588
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,32,power_law_1.2,0.09354496002197266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,32,power_law_1.2,0.12989824295043945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,32,power_law_1.2,0.13033599853515626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,32,power_law_1.2,0.2089676856994629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,32,power_law_1.2,0.2559091186523438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,32,power_law_1.2,0.45797889709472656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,32,power_law_1.2,0.5866918563842773
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,2,32,power_law_1.2,0.018246400356292724
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,2,32,power_law_1.2,0.01819391965866089
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,2,32,power_law_1.2,0.01815551996231079
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,2,32,power_law_1.2,0.0182041597366333
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,2,32,power_law_1.2,0.01795456051826477
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,2,32,power_law_1.2,0.01821823954582214
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,2,32,power_law_1.2,0.018150399923324585
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,2,32,power_law_1.2,0.01844864010810852
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,2,32,power_law_1.2,0.01838207960128784
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,2,32,power_law_1.2,0.01855487942695618
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,2,32,power_law_1.2,0.018584320545196535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,32,power_law_1.2,0.7877490997314454
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,2,32,power_law_1.2,0.018632960319519044
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,2,32,power_law_1.01,0.06484096050262451
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,2,32,power_law_1.2,0.018746880292892457
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,2,32,power_law_1.01,0.0716377592086792
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,2,32,power_law_1.2,0.020195839405059816
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,2,32,power_law_1.2,0.018595839738845828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,2,32,power_law_1.2,0.024779520034790038
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,32,power_law_1.2,1.0625933074951173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,32,power_law_1.2,0.04585343837738037
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,2,32,power_law_1.01,0.09479552268981933
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,2,32,power_law_1.01,0.12139391899108887
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,2,32,power_law_1.2,0.022338559627532957
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,2,32,power_law_1.2,0.019466240406036377
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,2,32,power_law_1.2,0.021035521030426024
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,2,32,power_law_1.2,0.032240641117095944
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,2,32,power_law_1.2,0.03231872081756591
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,4,16,balanced,0.021858561038970947
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,4,16,balanced,0.021437439918518066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,4,16,balanced,0.022350080013275146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,4,16,balanced,0.025131518840789797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,4,16,balanced,0.03956736087799072
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,4,16,balanced,0.041113600730896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,4,16,balanced,0.041521921157836914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,4,16,balanced,0.04162047863006592
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,4,16,balanced,0.04651391983032226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,4,16,balanced,0.04602367877960205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,4,16,balanced,0.04895359992980957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,4,16,balanced,0.048564481735229495
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,4,16,balanced,0.05273983955383301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,4,16,balanced,0.055129599571228025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,4,16,balanced,0.058826241493225094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,4,16,balanced,0.05842175960540772
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,4,16,balanced,0.061903362274169926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,4,16,balanced,0.0651916790008545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,4,16,balanced,0.020290560722351074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,4,16,balanced,0.08257663726806641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,4,16,balanced,0.09112959861755371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,4,16,balanced,0.11332608222961427
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,2,32,power_law_1.2,0.046903038024902345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,4,16,balanced,0.1296243190765381
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,4,16,balanced,0.17199871063232422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,4,16,balanced,0.2093414306640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,4,16,balanced,0.30099327087402344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,4,16,balanced,0.3931417465209961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,4,16,power_law_1.01,0.024354560375213625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,4,16,power_law_1.01,0.024384000301361085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,4,16,power_law_1.01,0.026758399009704593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,4,16,power_law_1.01,0.0387059211730957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,4,16,power_law_1.01,0.02346112012863159
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,4,16,power_law_1.01,0.04217984199523926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,4,16,power_law_1.01,0.043381757736206054
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,4,16,power_law_1.01,0.0456601619720459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,4,16,power_law_1.01,0.04684160232543945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,4,16,power_law_1.01,0.05083903789520263
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,4,16,power_law_1.01,0.04189695835113526
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,4,16,power_law_1.01,0.038147840499877926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,4,16,power_law_1.01,0.05542272090911865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,4,16,power_law_1.01,0.06740608215332031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,4,16,power_law_1.01,0.06694528102874756
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,4,16,power_law_1.01,0.06462207794189453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,4,16,power_law_1.01,0.06589312076568603
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,4,16,power_law_1.01,0.07535103797912598
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,4,16,power_law_1.01,0.08956543922424316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,4,16,power_law_1.01,0.10459775924682617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,4,16,power_law_1.01,0.11757951736450196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,4,16,power_law_1.01,0.15818240165710448
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,4,16,power_law_1.01,0.22013311386108397
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,4,16,power_law_1.01,0.2745446395874024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,4,16,power_law_1.01,0.3494579315185547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,2,32,power_law_1.2,0.059677438735961916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,4,16,power_law_1.2,0.023110399246215822
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,4,16,power_law_1.01,0.5504665756225586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,4,16,power_law_1.2,0.026851840019226074
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,2,32,power_law_1.2,0.07559552192687988
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,4,16,power_law_1.2,0.02469248056411743
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,4,16,power_law_1.2,0.025477120876312254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,4,16,power_law_1.2,0.03789695978164673
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,4,16,power_law_1.2,0.038813440799713134
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,4,16,power_law_1.2,0.04017536163330078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,4,16,power_law_1.2,0.04320256233215332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,4,16,power_law_1.2,0.04730879783630371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,4,16,power_law_1.2,0.051502079963684076
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,4,16,power_law_1.2,0.050681600570678706
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,4,16,power_law_1.2,0.050677762031555176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,4,16,power_law_1.2,0.05896063804626465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,4,16,power_law_1.2,0.061872639656066895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,4,16,power_law_1.2,0.06031744003295898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,4,16,power_law_1.2,0.06368256092071534
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,4,16,power_law_1.2,0.07623040199279785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,4,16,power_law_1.2,0.06925439834594727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,4,16,power_law_1.2,0.09143039703369141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,4,16,power_law_1.2,0.1055065631866455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,4,16,power_law_1.2,0.1396787166595459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,4,16,power_law_1.01,0.7637721252441406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,4,16,power_law_1.2,0.20951679229736325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,4,16,power_law_1.2,0.26113792419433596
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,4,16,power_law_1.2,0.37937278747558595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,8,8,balanced,0.0215116810798645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,8,8,balanced,0.022050559520721436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,8,8,balanced,0.02224128007888794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,8,8,balanced,0.027240960597991942
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,8,8,balanced,0.033733119964599614
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,4,16,power_law_1.2,0.7959539031982421
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,8,8,balanced,0.05382400035858155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,8,8,balanced,0.05662208080291749
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,8,8,balanced,0.059596800804138185
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,8,8,balanced,0.0580083179473877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,8,8,balanced,0.0602291202545166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,8,8,balanced,0.06449151992797851
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,8,8,balanced,0.06506112098693848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,8,8,balanced,0.07458943843841552
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,8,8,balanced,0.07218048095703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,8,8,balanced,0.073918719291687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,8,8,balanced,0.07229440212249756
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,8,8,balanced,0.08063232421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,8,8,balanced,0.08790143966674804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,8,8,balanced,0.09994239807128906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,8,8,balanced,0.13463040351867678
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,4,16,power_law_1.2,0.4556057739257812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,8,8,balanced,0.14988287925720215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,8,8,balanced,0.1980121612548828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,8,8,balanced,0.22523775100708007
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,8,8,balanced,0.3180057525634766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,8,8,balanced,0.3944947052001953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,8,8,power_law_1.01,0.02719615936279297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,8,8,balanced,0.578265609741211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,8,8,power_law_1.01,0.03091200113296509
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,8,8,power_law_1.01,0.029218559265136716
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,8,8,balanced,0.7537190246582031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,8,8,power_law_1.01,0.033471999168395994
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,8,8,power_law_1.01,0.05098368167877197
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,8,8,power_law_1.01,0.05729407787322997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,8,8,power_law_1.01,0.05886720180511475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,8,8,power_law_1.01,0.05961088180541992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,8,8,power_law_1.01,0.07030144214630127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,8,8,power_law_1.01,0.06786431789398192
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,8,8,power_law_1.01,0.07246208190917969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,8,8,power_law_1.01,0.07603456020355225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,8,8,power_law_1.01,0.07740543842315674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,8,8,power_law_1.01,0.07867136001586914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,8,8,power_law_1.01,0.08621567726135254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,8,8,power_law_1.01,0.034630401134490965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,8,8,power_law_1.01,0.10214271545410156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,8,8,power_law_1.01,0.05445759773254395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,8,8,power_law_1.01,0.1160204792022705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,8,8,power_law_1.01,0.13230976104736328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,8,8,power_law_1.01,0.16205440521240236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,8,8,power_law_1.01,0.2059916877746582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,8,8,power_law_1.01,0.2648038482666016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,8,8,power_law_1.01,0.36813438415527344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,8,8,power_law_1.2,0.02560256004333496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,8,8,power_law_1.01,0.45938560485839847
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,8,8,power_law_1.2,0.02805248022079468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,8,8,power_law_1.01,0.6464678192138672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,8,8,power_law_1.2,0.02926464080810547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,8,8,power_law_1.2,0.032249600887298585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,8,8,power_law_1.2,0.03362047910690308
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,8,8,power_law_1.2,0.05147776126861572
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,8,8,power_law_1.01,0.9410240173339843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,8,8,power_law_1.2,0.0539353609085083
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,8,8,power_law_1.2,0.05857920169830322
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,8,8,power_law_1.2,0.059828481674194335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,8,8,power_law_1.2,0.06066815853118897
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,8,8,power_law_1.2,0.07072896003723145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,8,8,power_law_1.2,0.0687001609802246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,8,8,power_law_1.2,0.07344255924224855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,8,8,power_law_1.2,0.07502719879150391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,8,8,power_law_1.2,0.0787059211730957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,8,8,power_law_1.2,0.0792409610748291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,8,8,power_law_1.2,0.09024895668029785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,8,8,power_law_1.2,0.10646400451660157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,8,8,power_law_1.2,0.11920767784118654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,8,8,power_law_1.2,0.14837759971618653
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,8,8,power_law_1.2,0.16661247253417968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,8,8,power_law_1.2,0.22860031127929686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,8,8,power_law_1.2,0.31202816009521483
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,8,8,power_law_1.2,0.4306700897216797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,8,8,power_law_1.2,0.5279628753662109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,4,16,power_law_1.2,0.994708480834961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,16,4,balanced,0.025207040309906004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,16,4,balanced,0.023895039558410644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,16,4,balanced,0.025958399772644043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,16,4,balanced,0.03128448009490967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,8,8,power_law_1.2,0.769428482055664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,16,4,balanced,0.03953023910522461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,16,4,balanced,0.04815487861633301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,16,4,balanced,0.06248064041137695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,16,4,balanced,0.0632204818725586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,16,4,balanced,0.06412159919738769
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,16,4,balanced,0.06575615882873535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,16,4,balanced,0.07526016235351562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,16,4,balanced,0.07486976146697998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,16,4,balanced,0.0704908800125122
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,16,4,balanced,0.07597439765930177
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,16,4,balanced,0.08034687995910644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,16,4,balanced,0.08807295799255371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,16,4,balanced,0.09913472175598145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,16,4,balanced,0.11160832405090333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,16,4,balanced,0.1525376033782959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,16,4,balanced,0.17176448822021484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,2,32,power_law_1.2,0.09709440231323242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,2,32,power_law_1.2,0.141461763381958
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,16,4,balanced,0.2680191993713379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,16,4,balanced,0.4084198379516602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,16,4,balanced,0.06994304180145264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,16,4,balanced,0.5066611099243163
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,16,4,power_law_1.01,0.024677119255065917
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,16,4,power_law_1.01,0.029240319728851317
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,16,4,balanced,0.7419481658935547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,16,4,power_law_1.01,0.029232640266418457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,16,4,power_law_1.01,0.034593279361724856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,16,4,power_law_1.01,0.0625548791885376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,16,4,balanced,0.9807820892333984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,16,4,power_law_1.01,0.06342783927917481
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,16,4,balanced,0.22756607055664063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,16,4,power_law_1.01,0.06873216152191161
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,16,4,power_law_1.01,0.07681663990020751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,16,4,power_law_1.01,0.0727449607849121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,16,4,power_law_1.01,0.07381120204925537
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,16,4,power_law_1.01,0.07580671787261964
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,16,4,power_law_1.01,0.07812479972839356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,16,4,power_law_1.01,0.08310015678405762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,16,4,power_law_1.01,0.1070911979675293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,16,4,power_law_1.01,0.11699328422546387
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,2,32,power_law_1.2,0.1972313690185547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,16,4,power_law_1.01,0.14485887527465818
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,16,4,power_law_1.01,0.17200511932373047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,16,4,power_law_1.01,0.24084352493286132
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,16,4,power_law_1.01,0.3254886245727539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,16,4,power_law_1.01,0.4283891296386718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,16,4,power_law_1.01,0.09034239768981933
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,16,4,power_law_1.2,0.024698879718780518
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,16,4,power_law_1.01,0.5323046493530273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,16,4,power_law_1.2,0.026801919937133788
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,16,4,power_law_1.01,0.8229401397705078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,16,4,power_law_1.2,0.03447168111801148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,16,4,power_law_1.01,0.039121921062469485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,16,4,power_law_1.01,0.06464384078979493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,16,4,power_law_1.2,0.04711552143096924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,16,4,power_law_1.01,1.0379724884033203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,16,4,power_law_1.2,0.05958144187927246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,16,4,power_law_1.2,0.06356991767883301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,16,4,power_law_1.2,0.07231359958648682
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,16,4,power_law_1.2,0.07043200016021729
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,16,4,power_law_1.2,0.07605504035949708
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,16,4,power_law_1.2,0.07266687870025634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,16,4,power_law_1.2,0.07639679908752442
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,16,4,power_law_1.2,0.08169088363647461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,16,4,power_law_1.01,0.05148543834686279
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,16,4,power_law_1.2,0.08394240379333497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,16,4,power_law_1.2,0.0916646385192871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,16,4,power_law_1.2,0.10737919807434082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,16,4,power_law_1.2,0.14930432319641113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,16,4,power_law_1.2,0.18279680252075195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,16,4,power_law_1.2,0.26756992340087893
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,16,4,power_law_1.2,0.037761280536651606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,16,4,power_law_1.2,0.3152396774291992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,16,4,power_law_1.2,0.5848960113525391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,8,8,power_law_1.2,1.1336959838867187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,32,2,balanced,0.022266879081726074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,32,2,balanced,0.02248703956604004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,16,4,power_law_1.2,0.46079486846923834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,16,4,power_law_1.2,0.8810572814941405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,32,2,balanced,0.02697727918624878
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,32,2,balanced,0.03466752052307129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,32,2,balanced,0.059572482109069826
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,32,2,balanced,0.04731391906738282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,32,2,balanced,0.057747201919555666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,16,4,power_law_1.2,0.058385920524597165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,32,2,balanced,0.05999360084533691
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,16,4,power_law_1.2,1.1537023925781251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,32,2,balanced,0.07220479965209961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,32,2,balanced,0.07425536155700684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,32,2,balanced,0.07435520172119141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,32,2,balanced,0.0788479995727539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,32,2,balanced,0.08834048271179198
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,32,2,balanced,0.07925888061523437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,32,2,balanced,0.08921343803405761
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,16,4,power_law_1.2,0.029265921115875244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,32,2,balanced,0.09585023880004882
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,32,2,balanced,0.10922112464904785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,32,2,balanced,0.12087167739868163
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,32,2,balanced,0.1639910316467285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,32,2,balanced,0.18407424926757812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,32,2,balanced,0.2435980796813965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,32,2,balanced,0.2834560012817383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,16,4,power_law_1.2,0.12199423789978028
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,32,2,balanced,0.4325235366821289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,32,2,balanced,0.5308953475952148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,32,2,power_law_1.01,0.027621119022369384
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,32,2,power_law_1.01,0.02935424089431763
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,32,2,balanced,0.07227392196655273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,32,2,power_law_1.01,0.03458559989929199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,32,2,balanced,0.780343017578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,32,2,power_law_1.01,0.0558080005645752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,32,2,power_law_1.01,0.05077760219573975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,32,2,power_law_1.01,0.04264063835144043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,32,2,power_law_1.01,0.055357441902160645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,32,2,power_law_1.01,0.07172351837158203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,32,2,power_law_1.01,0.0734822416305542
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,32,2,balanced,1.0297894287109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,32,2,power_law_1.01,0.02313983917236328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,32,2,power_law_1.01,0.07628159999847413
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,32,2,power_law_1.01,0.08668160438537598
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,32,2,power_law_1.01,0.08281855583190918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,32,2,power_law_1.01,0.0915494441986084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,32,2,power_law_1.01,0.0990783977508545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,32,2,power_law_1.01,0.1182975959777832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,32,2,power_law_1.01,0.13393792152404785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,32,2,power_law_1.01,0.16160255432128906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,32,2,power_law_1.01,0.19347711563110354
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,32,2,power_law_1.01,0.25802495956420896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,32,2,power_law_1.01,0.34100990295410155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,32,2,power_law_1.2,0.02281984090805054
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,32,2,power_law_1.01,0.4545280075073242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,32,2,power_law_1.2,0.026836481094360355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,32,2,power_law_1.01,0.06740096092224121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,32,2,power_law_1.01,0.07508480072021484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,32,2,power_law_1.2,0.029589760303497314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,32,2,power_law_1.2,0.03470463991165161
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,32,2,power_law_1.2,0.03927680015563965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,32,2,power_law_1.2,0.05630208015441894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,32,2,power_law_1.2,0.05497471809387207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,32,2,power_law_1.01,0.8343206024169921
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,32,2,power_law_1.2,0.0516812801361084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,32,2,power_law_1.2,0.06811264038085937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,32,2,power_law_1.2,0.07493120193481445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,32,2,power_law_1.2,0.07377408027648927
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,32,2,power_law_1.2,0.07498752117156983
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,32,2,power_law_1.2,0.083056640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,32,2,power_law_1.2,0.0901465606689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,32,2,power_law_1.2,0.08774144172668456
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,32,2,power_law_1.01,1.0853286743164063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,32,2,power_law_1.2,0.09437439918518067
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,32,2,power_law_1.2,0.1025881576538086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,32,2,power_law_1.2,0.11977984428405761
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,32,2,power_law_1.2,0.13616767883300782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,32,2,power_law_1.2,0.16453376770019532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,32,2,power_law_1.2,0.19487232208251953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,32,2,power_law_1.2,0.2817510414123535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,64,balanced,0.03333120107650757
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,32,2,power_law_1.01,0.6003417587280273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,32,2,power_law_1.2,0.34863105773925784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,64,balanced,0.07146880149841309
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,32,2,power_law_1.2,0.4626611328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,32,2,power_law_1.2,0.6167308807373046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,64,balanced,0.15033472061157227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,64,balanced,0.15567104339599608
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,32,2,power_law_1.2,0.8562226867675781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,64,balanced,0.16360191345214842
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,64,balanced,0.17787519454956055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,64,balanced,0.16558975219726563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,64,balanced,0.17247360229492187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,64,balanced,0.17491327285766603
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,64,balanced,0.19276544570922852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,64,balanced,0.2042598342895508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,64,balanced,0.2165670394897461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,64,balanced,0.0643993616104126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,64,balanced,0.23396991729736327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,64,balanced,0.26235008239746094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,64,balanced,0.1190118408203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,64,balanced,0.14644864082336426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,64,balanced,0.41795711517333983
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,64,balanced,0.16219776153564452
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,64,balanced,0.4814604949951172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,64,balanced,0.6304025650024414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,64,balanced,0.03487744092941284
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,64,balanced,0.03542655944824219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,64,balanced,1.0985177612304686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,64,balanced,0.2930944061279297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,64,balanced,0.06247680187225342
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,64,balanced,0.08806400299072266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,64,balanced,1.4641830444335937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,64,balanced,0.0899123191833496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,64,balanced,0.09649151802062989
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,64,balanced,0.09528703689575195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,64,balanced,0.7488998413085938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,64,balanced,0.09702015876770019
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,64,balanced,0.11050368309020997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,64,balanced,0.09830911636352539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,64,balanced,0.10068736076354981
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,64,balanced,0.04576767921447754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,64,balanced,2.232838439941406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,64,balanced,0.11265407562255861
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,64,balanced,0.1167039966583252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,64,balanced,0.12330240249633788
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,64,balanced,0.1365939235687256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,64,balanced,0.1456601619720459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,64,balanced,0.16028287887573242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,64,balanced,0.20045951843261717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,64,balanced,0.24040576934814456
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,64,balanced,2.918742980957031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,64,balanced,0.33564670562744137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,64,balanced,0.4094348907470703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,64,balanced,0.5785011291503906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,64,balanced,0.10183423995971679
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,64,balanced,0.7215219116210937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,64,balanced,1.0624243164062501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,32,2,power_law_1.2,1.1049664306640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,64,balanced,1.373236541748047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,1,64,balanced,0.02063744068145752
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,1,64,balanced,0.022417919635772703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,1,64,balanced,0.04837887763977051
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,1,64,balanced,0.046868481636047364
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,1,64,balanced,0.04689792156219482
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,1,64,balanced,0.048211197853088375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,1,64,balanced,0.04795648097991943
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,1,64,balanced,0.04793856143951416
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,1,64,balanced,0.048951039314270015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,1,64,balanced,0.04716032028198242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,1,64,balanced,0.04787968158721924
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,1,64,balanced,0.04894336223602295
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,1,64,balanced,0.02143359899520874
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,1,64,balanced,0.04917119979858399
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,1,64,balanced,0.04918655872344971
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,1,64,balanced,0.04903679847717285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,1,64,balanced,0.04974976062774658
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,1,64,balanced,0.051244797706604006
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,1,64,balanced,0.047248640060424806
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,1,64,balanced,0.03927551984786987
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,1,64,balanced,0.0475110387802124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,64,power_law_1.01,0.11769472122192384
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,64,power_law_1.01,0.12053503990173339
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,64,power_law_1.01,0.13216511726379393
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,64,power_law_1.01,0.13798656463623046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,64,power_law_1.01,0.14542847633361816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,64,power_law_1.01,0.15233280181884765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,64,power_law_1.01,0.1646143913269043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,64,power_law_1.01,0.17003135681152343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,64,power_law_1.01,0.16917503356933594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,64,power_law_1.01,0.17990144729614257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,64,power_law_1.01,0.10894975662231446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,64,power_law_1.01,0.1879897689819336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,64,power_law_1.01,0.20444927215576172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,64,power_law_1.01,0.23490560531616209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,64,power_law_1.01,0.25395456314086917
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,64,power_law_1.01,0.29062528610229493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,64,power_law_1.01,0.3192793655395508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,64,power_law_1.01,0.4132799911499023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,64,power_law_1.01,0.48677631378173825
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,64,power_law_1.01,0.639758071899414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,64,power_law_1.01,0.8152780914306641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,64,power_law_1.01,1.090453796386719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,64,power_law_1.01,1.5510182189941406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,64,power_law_1.01,2.256341705322266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,64,power_law_1.01,3.0746752929687498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,64,power_law_1.01,4.798849182128906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,64,power_law_1.01,0.07765888214111329
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,64,power_law_1.01,0.07707776069641113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,64,power_law_1.01,0.07824255943298339
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,64,power_law_1.01,0.08020352363586426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,64,power_law_1.01,0.07451263904571534
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,64,power_law_1.01,0.08695167541503906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,64,power_law_1.01,0.08859519958496094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,64,power_law_1.01,6.597037963867187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,64,power_law_1.01,0.09932800292968749
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,64,power_law_1.01,0.10125439643859864
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,64,power_law_1.01,0.09841792106628418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,64,power_law_1.01,0.10044032096862794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,64,power_law_1.01,0.11072256088256835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,64,power_law_1.01,0.11789823532104493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,64,power_law_1.01,0.13675007820129395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,64,power_law_1.01,0.14988287925720215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,64,power_law_1.01,0.1522214412689209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,64,power_law_1.01,0.1578777599334717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,64,power_law_1.01,0.20595584869384767
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,64,power_law_1.01,0.23643520355224606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,64,power_law_1.01,0.3114944076538086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,64,power_law_1.01,0.43472896575927733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,64,power_law_1.01,0.5740083312988281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,64,power_law_1.01,0.6663577270507812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,64,power_law_1.01,1.153347854614258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,1,64,balanced,0.0673689603805542
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,64,power_law_1.01,1.3902784729003907
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,1,64,balanced,0.08524800300598144
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,1,64,balanced,0.06514431953430175
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,1,64,power_law_1.01,0.04730495929718018
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,1,64,balanced,0.09017215728759766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,1,64,power_law_1.01,0.047957758903503414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,1,64,power_law_1.01,0.047375359535217286
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,1,64,balanced,0.1530521583557129
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,1,64,power_law_1.01,0.046964478492736814
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,1,64,balanced,0.12851327896118164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,1,64,power_law_1.01,0.04298495769500733
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,1,64,power_law_1.01,0.046048002243041994
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,1,64,power_law_1.01,0.04563968181610108
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,1,64,power_law_1.01,0.04560512065887451
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,1,64,power_law_1.01,0.04624639987945557
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,1,64,power_law_1.01,0.04534143924713135
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,1,64,power_law_1.01,0.04657536029815674
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,1,64,power_law_1.01,0.04639103889465332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,1,64,power_law_1.01,0.04647552013397217
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,1,64,power_law_1.01,0.047831039428710934
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,1,64,power_law_1.01,0.04797696113586426
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,1,64,power_law_1.01,0.048712959289550786
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,64,power_law_1.01,2.2195237731933597
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,64,power_law_1.01,3.0158258056640626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,1,64,power_law_1.01,0.08391167640686034
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,1,64,power_law_1.01,0.050621438026428225
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,1,64,power_law_1.01,0.056076798439025875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,1,64,power_law_1.01,0.06204671859741211
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,1,64,power_law_1.01,0.12656000137329101
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,64,power_law_1.2,0.1101478385925293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,1,64,power_law_1.01,0.07240831851959229
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,64,power_law_1.2,0.11729151725769042
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,1,64,power_law_1.01,0.11110527992248535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,64,power_law_1.2,0.12059136390686036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,64,power_law_1.2,0.1274726390838623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,64,power_law_1.2,0.13622912406921386
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,64,power_law_1.2,0.15315711975097657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,64,power_law_1.2,0.145666561126709
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,64,power_law_1.2,0.16665983200073242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,64,power_law_1.2,0.1684864044189453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,64,power_law_1.2,0.17131391525268555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,64,power_law_1.2,0.18747648239135742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,1,64,power_law_1.01,0.17499008178710937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,64,power_law_1.2,0.19430015563964845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,64,power_law_1.2,0.21759359359741212
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,64,power_law_1.2,0.2550489616394043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,64,power_law_1.2,0.2814028739929199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,64,power_law_1.2,0.2799577522277832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,64,power_law_1.2,0.35118465423583983
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,64,power_law_1.2,0.41676414489746094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,64,power_law_1.2,0.5369369506835937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,64,power_law_1.2,0.7688422393798828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,64,power_law_1.2,0.9796160125732423
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,64,power_law_1.2,1.495663299560547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,64,power_law_1.2,2.0379136657714847
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,64,power_law_1.2,2.8720474243164062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,64,power_law_1.2,0.07594624042510986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,64,power_law_1.2,0.07679999828338623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,64,power_law_1.2,0.07813119888305664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,64,power_law_1.2,0.07639808177947997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,64,power_law_1.2,0.06769408226013184
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,64,power_law_1.2,0.08089983940124512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,64,power_law_1.2,0.08842623710632325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,64,power_law_1.2,0.09958399772644043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,64,power_law_1.2,0.09716352462768554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,64,power_law_1.2,4.574513854980468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,64,power_law_1.2,0.09980928421020507
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,64,power_law_1.2,0.10619647979736328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,64,power_law_1.2,0.12082431793212892
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,64,power_law_1.2,0.11275008201599121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,64,power_law_1.2,0.13778047561645507
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,64,power_law_1.2,0.14382719993591309
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,64,power_law_1.2,0.15847807884216308
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,64,power_law_1.2,0.1725606346130371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,64,power_law_1.2,0.2096588706970215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,64,power_law_1.2,0.28050559997558594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,64,power_law_1.2,0.37525630950927735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,64,power_law_1.2,0.4310873413085938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,64,power_law_1.2,0.64674560546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,64,power_law_1.2,6.502857055664063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,64,power_law_1.2,0.8387315368652344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,64,power_law_1.2,1.2531404876708985
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,1,64,power_law_1.2,0.04746623992919922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,1,64,power_law_1.2,0.047211518287658696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,64,power_law_1.2,1.7773989868164062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,1,64,power_law_1.2,0.04747136116027832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,1,64,power_law_1.2,0.04651008129119873
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,1,64,power_law_1.2,0.03936768054962158
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,1,64,power_law_1.2,0.0452019214630127
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,1,64,power_law_1.2,0.04521088123321533
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,1,64,power_law_1.2,0.04577536106109619
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,1,64,power_law_1.2,0.04563712120056153
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,64,power_law_1.2,2.85035400390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,1,64,power_law_1.2,0.04623871803283691
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,1,64,power_law_1.2,0.04695551872253418
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,1,64,power_law_1.2,0.04763775825500488
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,1,64,power_law_1.2,0.047502079010009766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,1,64,power_law_1.2,0.04824575901031494
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,1,64,power_law_1.2,0.04888576030731201
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,1,64,power_law_1.2,0.0526361608505249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,64,power_law_1.2,3.7302886962890627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,64,power_law_1.2,8.925277709960938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,1,64,power_law_1.01,0.22345855712890622
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,1,64,power_law_1.01,0.3727743911743164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,1,64,power_law_1.01,0.46912769317626957
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,1,64,power_law_1.2,0.06141568183898925
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,1,64,power_law_1.2,0.04809855937957764
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,1,64,power_law_1.2,0.08235136032104493
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,1,64,power_law_1.2,0.06663551807403564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,32,balanced,0.0293887996673584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,32,balanced,0.04949632167816162
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,32,balanced,0.06122879981994629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,32,balanced,0.09915648460388184
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,32,balanced,0.1318899154663086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,32,balanced,0.14313728332519532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,32,balanced,0.17383808135986328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,32,balanced,0.15313920021057129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,32,balanced,0.16434944152832032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,32,balanced,0.15938048362731932
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,32,balanced,0.18167808532714844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,1,64,power_law_1.2,0.1094540786743164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,32,balanced,0.18892032623291016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,32,balanced,0.19256832122802733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,32,balanced,0.21691007614135743
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,32,balanced,0.22378496170043943
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,32,balanced,0.21811712265014646
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,32,balanced,0.24238464355468753
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,32,balanced,0.2963366317749023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,32,balanced,0.34215808868408204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,32,balanced,0.48578815460205077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,32,balanced,0.576226577758789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,32,balanced,0.7940057373046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,32,balanced,0.9631513977050782
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,1,64,power_law_1.2,0.13911680221557615
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,32,balanced,1.4241165161132812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,1,64,power_law_1.2,0.16642688751220702
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,32,balanced,0.0321343994140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,32,balanced,1.879278106689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,1,64,power_law_1.2,0.2505241584777832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,32,balanced,0.04210303783416748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,32,balanced,0.06376959800720214
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,32,balanced,0.08744192123413086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,32,balanced,0.09727104187011719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,32,balanced,0.11269887924194336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,32,balanced,0.09917056083679199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,32,balanced,0.10161408424377441
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,32,balanced,0.10233983993530274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,32,balanced,0.10888192176818848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,32,balanced,0.03323391914367676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,32,balanced,0.11263360023498534
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,32,balanced,0.1177344036102295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,32,balanced,0.13827327728271482
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,32,balanced,0.14729855537414552
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,32,balanced,0.13261311531066894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,32,balanced,2.836708984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,32,balanced,0.1449843215942383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,32,balanced,0.17456256866455078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,32,balanced,0.2104089546203613
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,32,balanced,0.26212095260620116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,32,balanced,0.32225536346435546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,32,balanced,0.4580518341064453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,32,balanced,0.5721472167968751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,32,balanced,3.755125732421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,32,balanced,0.8186675262451171
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,32,balanced,1.0348518371582032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,32,balanced,1.5276031494140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,32,balanced,2.0051058959960937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,1,64,power_law_1.2,0.34371711730957033
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,1,64,power_law_1.2,0.5378009414672851
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,1,64,power_law_1.2,0.6948274993896485
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,2,32,balanced,0.02250112056732178
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,2,32,balanced,0.020778241157531737
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,2,32,balanced,0.04979968070983887
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,2,32,balanced,0.05030911922454834
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,2,32,balanced,0.050424318313598636
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,2,32,balanced,0.050113282203674316
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,2,32,balanced,0.04985343933105469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,2,32,balanced,0.05100927829742432
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,2,32,balanced,0.02476032018661499
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,2,32,balanced,0.050241279602050784
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,2,32,balanced,0.050956802368164064
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,2,32,balanced,0.05099904060363769
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,2,32,balanced,0.050301442146301265
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,2,32,balanced,0.05344768047332764
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,2,32,balanced,0.0533516788482666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,2,32,balanced,0.055902719497680664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,2,32,balanced,0.03871360063552857
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,2,32,balanced,0.05002111911773681
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,2,32,balanced,0.050170879364013675
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,2,32,balanced,0.0505344009399414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,32,power_law_1.01,0.11430784225463868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,32,power_law_1.01,0.11339008331298828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,2,32,balanced,0.04966271877288818
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,32,power_law_1.01,0.11367168426513671
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,32,power_law_1.01,0.11959424018859863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,32,power_law_1.01,0.126112003326416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,32,power_law_1.01,0.1408243179321289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,32,power_law_1.01,0.1489408016204834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,32,power_law_1.01,0.1573631954193115
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,32,power_law_1.01,0.15510399818420412
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,32,power_law_1.01,0.1548736000061035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,32,power_law_1.01,0.1690073585510254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,32,power_law_1.01,0.17559423446655273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,32,power_law_1.01,0.18904191970825196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,32,power_law_1.01,0.21824384689331056
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,32,power_law_1.01,0.22369663238525392
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,32,power_law_1.01,0.2241881561279297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,32,power_law_1.01,0.2591551971435547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,32,power_law_1.01,0.3670707321166992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,32,power_law_1.01,0.43863681793212894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,32,power_law_1.01,0.5644723129272461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,32,power_law_1.01,0.7249945831298829
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,32,power_law_1.01,1.029876480102539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,32,power_law_1.01,1.3296141052246093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,32,power_law_1.01,2.0089382934570312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,32,power_law_1.01,2.2581875610351565
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,2,32,balanced,0.07296512126922608
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,2,32,balanced,0.0993126392364502
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,2,32,balanced,0.07590784072875977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,32,power_law_1.01,0.07885056018829346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,32,power_law_1.01,0.08110976219177246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,32,power_law_1.01,0.0791654396057129
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,2,32,balanced,0.10677120208740234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,32,power_law_1.01,0.08279040336608887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,32,power_law_1.01,0.08211968421936036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,32,power_law_1.01,0.08923520088195801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,32,power_law_1.01,0.10023807525634765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,32,power_law_1.01,0.10203264236450196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,32,power_law_1.01,0.0993894386291504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,32,power_law_1.01,0.09484288215637207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,32,power_law_1.01,0.10994943618774414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,32,power_law_1.01,0.11380736351013183
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,32,power_law_1.01,0.11927552223205566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,32,power_law_1.01,0.14590335845947267
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,32,power_law_1.01,0.1383206367492676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,32,power_law_1.01,0.1461516761779785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,32,power_law_1.01,0.19465599060058594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,32,power_law_1.01,0.23682176589965823
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,32,power_law_1.01,3.9390988159179683
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,32,power_law_1.01,0.3033011245727539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,32,power_law_1.01,0.3608038330078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,2,32,balanced,0.15420672416687012
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,32,power_law_1.01,0.14066944122314454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,32,power_law_1.01,0.5242086410522461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,2,32,power_law_1.01,0.050242562294006345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,32,power_law_1.01,0.7199628448486328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,32,power_law_1.01,1.0134015655517579
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,2,32,power_law_1.01,0.04909311771392823
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,2,32,power_law_1.01,0.04985727787017823
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,2,32,power_law_1.01,0.049684481620788576
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,2,32,power_law_1.01,0.047699198722839356
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,2,32,power_law_1.01,0.044650239944458006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,32,power_law_1.01,5.268684692382813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,2,32,power_law_1.01,0.045987839698791506
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,2,32,power_law_1.01,0.04572927951812744
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,2,32,power_law_1.01,0.04698495864868164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,2,32,power_law_1.01,0.0488588809967041
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,2,32,power_law_1.01,0.04960512161254883
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,2,32,power_law_1.01,0.04642176151275635
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,32,power_law_1.01,1.9546226501464843
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,2,32,power_law_1.01,0.05041279792785645
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,2,32,power_law_1.01,0.051476478576660156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,2,32,power_law_1.01,0.05126656055450439
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,2,32,power_law_1.01,0.06318336009979249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,32,power_law_1.01,1.3242137145996093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,2,32,power_law_1.01,0.049610238075256344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,32,power_law_1.01,2.8504257202148438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,2,32,balanced,0.18435712814331057
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,2,32,power_law_1.01,0.05845248222351074
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,2,32,power_law_1.01,0.08385791778564453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,2,32,power_law_1.01,0.05372032165527344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,2,32,power_law_1.01,0.07353983879089356
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,2,32,power_law_1.01,0.11180543899536133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,32,power_law_1.2,0.11650176048278808
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,32,power_law_1.2,0.113821439743042
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,2,32,power_law_1.01,0.14789504051208496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,32,power_law_1.2,0.11433216094970704
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,2,32,power_law_1.01,0.1898419189453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,32,power_law_1.2,0.12001536369323731
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,32,power_law_1.2,0.1228492832183838
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,32,power_law_1.2,0.13774335861206055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,32,power_law_1.2,0.14596735954284668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,32,power_law_1.2,0.1456460762023926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,32,power_law_1.2,0.15697664260864258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,32,power_law_1.2,0.1590220832824707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,32,power_law_1.2,0.176181755065918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,32,power_law_1.2,0.1833510398864746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,32,power_law_1.2,0.19977855682373047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,32,power_law_1.2,0.2277132797241211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,32,power_law_1.2,0.22519552230834963
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,32,power_law_1.2,0.2348633575439453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,32,power_law_1.2,0.3012236785888672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,32,power_law_1.2,0.3760076904296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,32,power_law_1.2,0.4792115020751953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,32,power_law_1.2,0.6169843292236328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,2,32,power_law_1.01,0.2480166435241699
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,32,power_law_1.2,0.7982733154296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,32,power_law_1.2,1.1510272216796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,32,power_law_1.2,0.0796134376525879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,32,power_law_1.2,1.5675315856933594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,32,power_law_1.2,0.07899648189544678
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,32,power_law_1.2,2.2345011901855467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,32,power_law_1.2,0.08017024040222168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,32,power_law_1.2,0.08311552047729492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,32,power_law_1.2,0.08271615982055665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,32,power_law_1.2,0.0961241626739502
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,32,power_law_1.2,0.0867686367034912
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,32,power_law_1.2,3.0764389038085938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,32,power_law_1.2,0.09818367958068848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,32,power_law_1.2,0.09694335937500001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,32,power_law_1.2,0.09990528106689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,32,power_law_1.2,0.10849535942077637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,32,power_law_1.2,0.11479167938232422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,32,power_law_1.2,0.1400780773162842
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,32,power_law_1.2,0.1489574432373047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,32,power_law_1.2,0.1421836757659912
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,32,power_law_1.2,0.1626790428161621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,32,power_law_1.2,0.20540672302246094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,32,power_law_1.2,4.583084716796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,32,power_law_1.2,0.2400192070007324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,32,power_law_1.2,0.3127231979370117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,32,power_law_1.2,0.1125823974609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,32,power_law_1.2,0.4122060775756836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,32,power_law_1.2,0.6263948822021485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,32,power_law_1.2,0.666487045288086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,2,32,power_law_1.2,0.050275840759277345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,32,power_law_1.2,1.1014527893066406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,2,32,power_law_1.2,0.05033472061157227
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,2,32,power_law_1.2,0.049548802375793455
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,2,32,power_law_1.2,0.04851583957672119
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,32,power_law_1.2,1.54158203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,2,32,power_law_1.2,0.04818816184997558
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,2,32,power_law_1.2,0.0429747200012207
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,2,32,power_law_1.2,0.04484479904174805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,2,32,power_law_1.2,0.04644735813140869
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,2,32,power_law_1.2,0.04799232006072998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,32,power_law_1.2,2.1606207275390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,2,32,power_law_1.2,0.049095678329467776
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,2,32,power_law_1.2,0.04831103801727295
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,2,32,power_law_1.2,0.04857344150543213
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,2,32,power_law_1.2,0.04973824024200439
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,2,32,power_law_1.2,0.04905600070953369
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,2,32,power_law_1.2,0.05245183944702149
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,2,32,power_law_1.2,0.062067198753356936
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,2,32,power_law_1.2,0.05435647964477539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,32,power_law_1.2,6.324925537109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,32,power_law_1.2,2.992967529296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,2,32,power_law_1.01,0.3399372863769531
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,2,32,power_law_1.01,0.533759994506836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,2,32,power_law_1.2,0.3986764907836914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,2,32,power_law_1.2,0.06916351795196533
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,2,32,power_law_1.2,0.09172224044799805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,2,32,power_law_1.2,0.053392639160156255
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,4,16,balanced,0.04398591995239258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,2,32,power_law_1.2,0.10881279945373534
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,4,16,balanced,0.05989376068115234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,4,16,balanced,0.07534848213195801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,4,16,balanced,0.1125055980682373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,4,16,balanced,0.14026495933532715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,4,16,balanced,0.14492416381835938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,4,16,balanced,0.153953275680542
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,4,16,balanced,0.16119295120239258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,4,16,balanced,0.16866304397583007
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,4,16,balanced,0.17380479812622068
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,4,16,balanced,0.19231744766235354
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,4,16,balanced,0.19724288940429688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,4,16,balanced,0.22520576477050783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,4,16,balanced,0.20806400299072267
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,4,16,balanced,0.22665727615356443
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,4,16,balanced,0.25244800567626957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,4,16,balanced,0.02880000114440918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,2,32,power_law_1.2,0.14761728286743164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,4,16,balanced,0.36865409851074216
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,2,32,power_law_1.2,0.16848512649536135
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,4,16,balanced,0.5427993774414063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,4,16,balanced,0.6562290954589843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,4,16,balanced,0.9144755554199218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,4,16,balanced,0.31016447067260744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,4,16,balanced,1.1334054565429688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,4,16,balanced,0.030516479015350338
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,4,16,balanced,0.03496959924697876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,4,16,balanced,0.03827455997467041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,4,16,balanced,1.6877850341796876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,4,16,balanced,0.05645823955535889
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,4,16,balanced,0.08354944229125977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,4,16,balanced,0.10429823875427247
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,4,16,balanced,0.10569727897644043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,4,16,balanced,0.10804991722106934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,4,16,balanced,2.183171844482422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,4,16,balanced,0.11298432350158691
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,4,16,balanced,0.11616127967834473
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,4,16,balanced,0.12477952003479005
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,4,16,balanced,0.13543295860290527
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,4,16,balanced,0.1462399959564209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,4,16,balanced,0.15837951660156252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,4,16,balanced,0.14027008056640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,4,16,balanced,0.15614336013793945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,4,16,balanced,0.1676812744140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,4,16,balanced,0.21159551620483397
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,4,16,balanced,0.24777343749999997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,4,16,balanced,0.3169049644470215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,4,16,balanced,3.2519284057617184
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,4,16,balanced,0.3880358505249023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,4,16,balanced,0.5623014450073243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,4,16,balanced,0.6986137390136719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,4,16,balanced,4.331119384765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,4,16,balanced,1.0176934051513673
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,4,16,balanced,1.300674591064453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,4,16,balanced,1.938030090332031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,4,16,balanced,2.5306597900390626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,2,32,power_law_1.2,0.27418495178222657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,2,32,power_law_1.2,0.5616307067871094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,4,16,balanced,0.023527679443359376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,4,16,balanced,0.05714176177978516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,4,16,balanced,0.05722496032714843
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,4,16,balanced,0.05702655792236329
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,4,16,balanced,0.05747968196868897
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,4,16,balanced,0.057292799949646
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,4,16,balanced,0.02380160093307495
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,4,16,balanced,0.057867522239685054
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,4,16,balanced,0.05828608036041259
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,4,16,balanced,0.0577177619934082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,4,16,balanced,0.026385920047760008
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,4,16,balanced,0.05779967784881592
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,4,16,balanced,0.05934847831726074
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,4,16,balanced,0.06059008121490479
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,4,16,balanced,0.0631987190246582
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,4,16,balanced,0.06477183818817138
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,4,16,balanced,0.06830719947814942
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,2,32,power_law_1.2,0.8559385681152344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,4,16,balanced,0.036677119731903074
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,4,16,balanced,0.05606912136077881
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,4,16,balanced,0.05106175899505615
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,4,16,power_law_1.01,0.0863475227355957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,4,16,power_law_1.01,0.11290752410888671
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,4,16,power_law_1.01,0.11382399559020998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,4,16,power_law_1.01,0.11330816268920899
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,4,16,power_law_1.01,0.1259225559234619
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,4,16,balanced,0.056799998283386235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,4,16,power_law_1.01,0.1347929573059082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,4,16,power_law_1.01,0.15118464469909668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,4,16,power_law_1.01,0.1423078441619873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,4,16,power_law_1.01,0.14686464309692382
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,4,16,power_law_1.01,0.15688447952270507
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,4,16,power_law_1.01,0.16297216415405275
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,4,16,power_law_1.01,0.17365503311157227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,4,16,power_law_1.01,0.18429824829101563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,4,16,power_law_1.01,0.20878591537475585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,4,16,power_law_1.01,0.2046976089477539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,4,16,power_law_1.01,0.21059328079223633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,4,16,power_law_1.01,0.23521919250488282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,4,16,power_law_1.01,0.31738752365112305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,4,16,power_law_1.01,0.39931903839111327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,4,16,power_law_1.01,0.5234265518188477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,4,16,power_law_1.01,0.6509286499023437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,4,16,power_law_1.01,0.8971123504638673
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,4,16,power_law_1.01,1.1955379486083983
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,4,16,power_law_1.01,1.6747430419921876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,4,16,power_law_1.01,2.223582763671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,4,16,power_law_1.01,3.3426162719726564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,4,16,power_law_1.01,0.06380288124084472
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,4,16,power_law_1.01,0.08726400375366211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,4,16,power_law_1.01,0.08513792037963867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,4,16,power_law_1.01,0.09006336212158203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,4,16,power_law_1.01,4.145354309082031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,4,16,power_law_1.01,0.09376383781433105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,4,16,power_law_1.01,0.11424384117126465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,4,16,power_law_1.01,0.09626367568969726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,4,16,power_law_1.01,0.10445440292358399
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,4,16,power_law_1.01,0.10567551612854005
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,4,16,power_law_1.01,0.1096895980834961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,4,16,power_law_1.01,0.11820287704467773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,4,16,power_law_1.01,0.12280447959899903
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,4,16,power_law_1.01,0.1344486427307129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,4,16,power_law_1.01,0.151975679397583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,4,16,power_law_1.01,0.1478758430480957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,4,16,power_law_1.01,0.14221183776855467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,4,16,power_law_1.01,0.16245759963989256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,4,16,power_law_1.01,0.20363136291503908
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,4,16,power_law_1.01,0.23632383346557617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,4,16,power_law_1.01,0.3114137649536133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,4,16,power_law_1.01,0.39719806671142577
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,4,16,power_law_1.01,0.5344627380371094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,4,16,power_law_1.01,0.6855615997314454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,4,16,power_law_1.01,1.0257331085205077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,4,16,power_law_1.01,1.2794649505615234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,4,16,balanced,0.10073856353759765
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,4,16,power_law_1.01,0.045480961799621585
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,4,16,balanced,0.0947929573059082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,4,16,balanced,0.1369279956817627
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,4,16,power_law_1.01,0.05647488117218018
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,4,16,power_law_1.01,0.056556801795959476
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,4,16,power_law_1.01,0.055429120063781735
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,4,16,power_law_1.01,0.05368832111358642
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,4,16,balanced,0.14957695960998535
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,4,16,power_law_1.01,0.0551142406463623
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,4,16,power_law_1.01,0.052249598503112796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,4,16,power_law_1.01,1.9367027282714844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,4,16,power_law_1.01,0.04973951816558838
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,4,16,power_law_1.01,0.05254528045654296
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,4,16,power_law_1.01,0.053035521507263185
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,4,16,power_law_1.01,0.055910401344299324
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,4,16,power_law_1.01,0.05594111919403076
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,4,16,balanced,0.2198348808288574
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,4,16,power_law_1.01,0.0574451208114624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,4,16,balanced,0.2700172805786133
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,4,16,power_law_1.01,0.06082816123962402
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,4,16,power_law_1.01,2.5334259033203126
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,4,16,power_law_1.01,0.05911295890808106
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,4,16,power_law_1.01,0.05465472221374512
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,4,16,power_law_1.01,0.0784499216079712
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,4,16,power_law_1.01,0.09549823760986328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,4,16,power_law_1.01,0.0706060791015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,4,16,power_law_1.01,0.0651904010772705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,4,16,power_law_1.2,0.08686976432800293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,4,16,power_law_1.2,0.09889408111572265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,4,16,power_law_1.2,0.11153535842895508
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,4,16,power_law_1.01,0.15238143920898437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,4,16,power_law_1.2,0.11367808341979982
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,4,16,power_law_1.01,0.18626943588256836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,4,16,power_law_1.2,0.12797311782836912
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,4,16,power_law_1.2,0.1314572811126709
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,4,16,power_law_1.2,0.14433279991149903
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,4,16,power_law_1.2,0.1424729633331299
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,4,16,power_law_1.2,0.14812671661376953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,4,16,power_law_1.2,0.1561638355255127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,4,16,power_law_1.2,0.16689407348632812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,4,16,power_law_1.2,0.17795711517333984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,4,16,power_law_1.2,0.18962303161621094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,4,16,power_law_1.2,0.2172275161743164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,4,16,power_law_1.2,0.20567295074462894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,4,16,power_law_1.2,0.2176639938354492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,4,16,power_law_1.2,0.24548992156982422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,4,16,power_law_1.2,0.34314624786376957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,4,16,power_law_1.2,0.4059622573852539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,4,16,power_law_1.2,0.527454719543457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,4,16,power_law_1.2,0.6854169464111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,4,16,power_law_1.2,0.9654528045654297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,4,16,power_law_1.2,1.2697177886962892
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,4,16,power_law_1.2,1.925861053466797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,4,16,power_law_1.2,2.4777842712402345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,4,16,power_law_1.2,0.06239744186401367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,4,16,power_law_1.2,0.08552831649780274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,4,16,power_law_1.2,0.08406784057617187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,4,16,power_law_1.2,3.5615603637695314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,4,16,power_law_1.2,0.09059583663940429
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,4,16,power_law_1.2,0.09431936264038085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,4,16,power_law_1.2,0.11046400070190429
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,4,16,power_law_1.2,0.10684415817260742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,4,16,power_law_1.2,0.10675840377807617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,4,16,power_law_1.2,0.11000960350036622
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,4,16,power_law_1.2,0.11764479637145997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,4,16,power_law_1.2,0.12233471870422363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,4,16,power_law_1.2,0.13869567871093752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,4,16,power_law_1.2,0.15626239776611328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,4,16,power_law_1.2,0.0986457633972168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,4,16,power_law_1.2,0.13615872383117675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,4,16,power_law_1.2,0.14614399909973144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,4,16,power_law_1.2,0.16359807968139647
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,4,16,power_law_1.2,0.20148735046386718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,4,16,power_law_1.2,0.23550336837768554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,4,16,power_law_1.2,0.32217857360839847
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,4,16,power_law_1.2,0.41054336547851566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,4,16,power_law_1.2,0.5734182357788086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,4,16,power_law_1.2,0.749775390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,4,16,power_law_1.2,4.880245666503906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,4,16,power_law_1.2,1.0577305603027345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,4,16,power_law_1.2,1.3599232482910157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,4,16,power_law_1.2,0.04544896125793457
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,4,16,power_law_1.2,0.05288064002990722
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,4,16,power_law_1.2,0.053547520637512204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,4,16,power_law_1.2,1.9904754638671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,4,16,power_law_1.2,0.05671167850494385
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,4,16,power_law_1.2,0.05521279811859131
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,4,16,power_law_1.2,0.05484799861907959
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,4,16,power_law_1.2,0.049848318099975586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,4,16,power_law_1.2,0.051041278839111325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,4,16,power_law_1.2,2.7403250122070313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,4,16,power_law_1.2,0.05235199928283692
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,4,16,power_law_1.2,0.056241922378540046
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,4,16,power_law_1.2,0.055656957626342776
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,4,16,power_law_1.2,0.05214335918426514
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,4,16,power_law_1.2,0.05638271808624268
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,4,16,power_law_1.2,0.061249279975891115
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,4,16,power_law_1.2,0.05837952136993409
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,4,16,power_law_1.01,0.26238592147827144
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,4,16,power_law_1.01,0.31761535644531247
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,4,16,power_law_1.2,0.06320127964019776
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,4,16,power_law_1.2,0.06856448173522948
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,4,16,power_law_1.01,0.5044761657714844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,4,16,power_law_1.01,0.6466470336914062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,4,16,power_law_1.2,0.0882265567779541
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,4,16,power_law_1.2,0.24147327423095702
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,4,16,power_law_1.2,0.1397158432006836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,4,16,power_law_1.2,0.1253004837036133
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,4,16,power_law_1.2,0.19675775527954104
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,4,16,power_law_1.2,0.07519487857818603
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,8,8,balanced,0.030154240131378175
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,8,8,balanced,0.026087679862976075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,8,8,balanced,0.06886784076690675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,8,8,balanced,0.05056896209716797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,8,8,balanced,0.09706624031066895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,8,8,balanced,0.1280396842956543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,4,16,power_law_1.2,0.3931097412109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,8,8,balanced,0.13981696128845217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,8,8,balanced,0.14226688385009764
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,8,8,balanced,0.16360448837280273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,8,8,balanced,0.16601472854614258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,8,8,balanced,0.17161600112915038
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,8,8,balanced,0.18567295074462892
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,8,8,balanced,0.19528959274291993
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,8,8,balanced,0.21351423263549804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,8,8,balanced,0.19626880645751954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,8,8,balanced,0.21015935897827148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,8,8,balanced,0.2372991943359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,8,8,balanced,0.29345279693603515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,8,8,balanced,0.34509056091308593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,8,8,balanced,0.5118873596191407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,8,8,balanced,0.615393295288086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,8,8,balanced,0.8732275390625001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,8,8,balanced,1.0720550537109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,8,8,balanced,1.588494110107422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,8,8,balanced,0.030088961124420166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,8,8,balanced,0.03583103895187378
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,8,8,balanced,2.0613388061523437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,8,8,balanced,0.05803647994995117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,8,8,balanced,0.07761663913726806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,8,8,balanced,0.102357759475708
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,8,8,balanced,0.11077759742736817
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,8,8,balanced,0.11231103897094727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,8,8,balanced,3.056188049316406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,8,8,balanced,0.12944512367248534
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,8,8,balanced,0.13218048095703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,8,8,balanced,0.13705087661743165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,8,8,balanced,0.1497612762451172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,8,8,balanced,0.15451519966125488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,8,8,balanced,0.17105152130126952
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,8,8,balanced,0.14980607986450195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,8,8,balanced,0.16044544219970702
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,8,8,balanced,0.17527040481567382
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,8,8,balanced,3.989462890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,8,8,balanced,0.20947711944580077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,8,8,balanced,0.24069887161254883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,8,8,balanced,0.3096396827697754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,8,8,balanced,0.37248897552490234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,8,8,balanced,0.5510707092285156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,8,8,balanced,0.6780172729492188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,8,8,balanced,0.043740158081054685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,8,8,balanced,0.9835545349121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,8,8,power_law_1.01,0.06322175979614257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,8,8,balanced,1.2343641662597657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,8,8,power_law_1.01,0.08784768104553223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,8,8,power_law_1.01,0.07999231815338134
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,8,8,power_law_1.01,0.09096320152282715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,8,8,power_law_1.01,0.11278976440429686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,8,8,balanced,1.8593394470214843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,8,8,power_law_1.01,0.13980928421020508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,8,8,power_law_1.01,0.1374003219604492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,8,8,power_law_1.01,0.15534079551696778
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,8,8,balanced,2.4113754272460937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,8,8,power_law_1.01,0.1617305564880371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,8,8,power_law_1.01,0.18334720611572267
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,8,8,power_law_1.01,0.20732543945312498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,8,8,power_law_1.01,0.1247539234161377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,8,8,power_law_1.01,0.14887807846069337
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,8,8,power_law_1.01,0.19305599212646485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,8,8,power_law_1.01,0.23590911865234374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,8,8,power_law_1.01,0.18100223541259766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,8,8,power_law_1.01,0.3109990310668945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,8,8,power_law_1.01,0.38729598999023435
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,8,8,power_law_1.01,0.5114112091064453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,8,8,power_law_1.01,0.6249164962768555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,8,8,power_law_1.01,0.20754816055297853
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,8,8,power_law_1.01,0.8903206634521486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,8,8,power_law_1.01,1.1396620941162108
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,8,8,power_law_1.01,1.7111308288574218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,8,8,power_law_1.01,0.04348031997680664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,8,8,power_law_1.01,2.1596185302734376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,8,8,power_law_1.01,0.06872576236724853
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,8,8,power_law_1.01,0.0677849578857422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,8,8,power_law_1.01,0.08084735870361329
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,8,8,power_law_1.01,0.09209216117858887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,8,8,power_law_1.01,0.09858816146850587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,8,8,power_law_1.01,0.10734335899353027
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,8,8,power_law_1.01,3.2415679931640624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,4,16,power_law_1.2,0.5147379302978516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,8,8,power_law_1.01,0.108920316696167
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,8,8,power_law_1.01,0.12388223648071288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,8,8,power_law_1.01,0.12540672302246095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,8,8,power_law_1.01,0.14388992309570312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,8,8,power_law_1.01,0.1307430362701416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,8,8,power_law_1.01,0.15204863548278807
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,8,8,power_law_1.01,0.16827903747558595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,8,8,power_law_1.01,0.14576767921447753
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,8,8,power_law_1.01,0.15457023620605467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,8,8,power_law_1.01,0.17294208526611327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,8,8,power_law_1.01,0.21128704071044924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,8,8,power_law_1.01,0.23806335449218752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,8,8,power_law_1.01,0.3177984046936035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,8,8,power_law_1.01,4.191121826171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,8,8,power_law_1.01,0.4067724609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,8,8,power_law_1.01,0.544666862487793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,4,16,power_law_1.2,0.7928691101074219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,8,8,power_law_1.01,0.6945394897460938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,8,8,power_law_1.01,1.0272447967529297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,8,8,power_law_1.2,0.06259967803955077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,8,8,power_law_1.2,0.07334911823272705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,8,8,power_law_1.01,1.2827430725097657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,8,8,power_law_1.2,0.08382080078125001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,8,8,power_law_1.2,0.09910655975341796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,8,8,power_law_1.2,0.12369664192199707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,8,8,power_law_1.2,0.1372211170196533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,8,8,power_law_1.01,1.93546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,8,8,power_law_1.2,0.15608960151672363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,8,8,power_law_1.2,0.16718080520629883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,8,8,power_law_1.01,2.5310092163085938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,8,8,power_law_1.2,0.17791488647460937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,8,8,power_law_1.2,0.10888959884643554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,8,8,power_law_1.2,0.1875699234008789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,8,8,power_law_1.2,0.19456256866455077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,8,8,power_law_1.2,0.21046783447265627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,8,8,power_law_1.2,0.2417318344116211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,8,8,power_law_1.2,0.3103615951538086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,8,8,power_law_1.2,0.12561792373657227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,8,8,power_law_1.2,0.4021427154541016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,8,8,power_law_1.2,0.5200153732299805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,8,8,power_law_1.2,0.6365414428710937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,8,8,power_law_1.2,0.9196812438964844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,8,8,power_law_1.2,0.14975616455078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,8,8,power_law_1.2,1.1711974334716797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,4,16,power_law_1.2,1.1435430145263672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,8,8,power_law_1.2,0.21074304580688477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,8,8,power_law_1.2,0.043196158409118654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,8,8,power_law_1.2,0.05753727912902832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,8,8,power_law_1.2,1.7737126159667969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,8,8,power_law_1.2,0.06778624057769775
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,8,8,power_law_1.2,0.07295616149902344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,8,8,power_law_1.2,0.10576767921447754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,8,8,power_law_1.2,0.09512319564819335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,8,8,power_law_1.2,0.10916223526000976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,8,8,power_law_1.2,0.09726592063903808
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,8,8,power_law_1.2,0.1259455966949463
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,8,8,power_law_1.2,2.2607513427734376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,8,8,power_law_1.2,0.12683520317077637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,8,8,power_law_1.2,0.14355072021484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,8,8,power_law_1.2,0.1512934398651123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,8,8,power_law_1.2,0.1698355293273926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,8,8,power_law_1.2,0.14744576454162597
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,8,8,power_law_1.2,0.15677184104919434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,8,8,power_law_1.2,0.17680383682250977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,8,8,power_law_1.2,0.214465274810791
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,8,8,power_law_1.2,0.245534725189209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,8,8,power_law_1.2,0.3215283203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,8,8,power_law_1.2,0.4129638290405273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,8,8,power_law_1.2,3.3167486572265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,8,8,power_law_1.2,0.13487104415893553
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,8,8,power_law_1.2,0.5616563034057618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,8,8,power_law_1.2,0.7246284484863281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,16,4,balanced,0.03585920095443725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,16,4,balanced,0.07270016193389892
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,16,4,balanced,0.11388671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,8,8,power_law_1.2,1.0273664093017578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,16,4,balanced,0.17940351486206055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,16,4,balanced,0.2406502342224121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,8,8,power_law_1.2,1.3271629333496093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,16,4,balanced,0.2556403160095215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,16,4,balanced,0.2630131149291992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,16,4,balanced,0.27604480743408205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,16,4,balanced,0.2796979141235352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,16,4,balanced,0.03952383995056152
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,16,4,balanced,0.28879743576049804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,8,8,power_law_1.2,1.971216583251953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,16,4,balanced,0.32795265197753903
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,16,4,balanced,0.33305088043212894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,16,4,balanced,0.35312896728515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,16,4,balanced,0.3336383819580078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,16,4,balanced,0.3470732879638672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,16,4,balanced,0.3719244766235351
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,16,4,balanced,0.423240966796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,16,4,balanced,0.4741606521606445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,8,8,power_law_1.2,2.5954815673828127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,16,4,balanced,0.7573465728759766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,16,4,balanced,0.8518297576904297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,16,4,balanced,1.2242022705078124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,16,4,balanced,0.0774015998840332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,16,4,balanced,1.4169818115234374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,16,4,balanced,0.15387904167175293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,16,4,balanced,0.21769472122192385
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,16,4,balanced,0.3658240127563476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,16,4,balanced,0.6609510040283203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,16,4,balanced,2.1377484130859377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,8,8,power_law_1.2,4.384102478027343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,16,4,balanced,2.7046105957031252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,16,4,balanced,0.859810562133789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,16,4,balanced,0.8759487915039064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,16,4,balanced,0.865560302734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,16,4,balanced,0.8699225616455077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,16,4,balanced,0.8809804534912109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,16,4,balanced,1.2011942291259765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,16,4,balanced,3.964320068359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,16,4,balanced,1.1477593231201173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,16,4,balanced,1.1729459381103515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,16,4,balanced,1.3406681823730469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,16,4,balanced,5.0607629394531255
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,16,4,balanced,1.3853312683105468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,16,4,balanced,1.4087986755371094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,16,4,balanced,1.216843490600586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,16,4,balanced,1.2274559783935548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,16,4,balanced,1.4730035400390624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,16,4,balanced,1.5316493225097656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,16,4,balanced,1.673675537109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,16,4,power_law_1.01,0.06281472206115722
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,16,4,power_law_1.01,0.09988991737365722
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,16,4,power_law_1.01,0.10433152198791504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,16,4,balanced,2.8550067138671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,16,4,balanced,3.08927490234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,16,4,power_law_1.01,0.1296076774597168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,16,4,power_law_1.01,0.16698623657226563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,16,4,power_law_1.01,0.21523967742919922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,16,4,power_law_1.01,0.22525312423706056
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,16,4,power_law_1.01,0.23271551132202148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,16,4,power_law_1.01,0.25494783401489257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,16,4,balanced,3.866830139160156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,16,4,power_law_1.01,0.25864959716796876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,16,4,power_law_1.01,0.3020428848266602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,16,4,power_law_1.01,0.31808895111083985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,16,4,power_law_1.01,0.3356723022460938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,16,4,balanced,4.461285095214843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,16,4,power_law_1.01,0.3281612777709961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,16,4,power_law_1.01,0.3859379196166992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,16,4,power_law_1.01,0.4820057678222656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,16,4,power_law_1.01,0.5759526443481445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,16,4,power_law_1.01,0.7351987457275391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,16,4,power_law_1.01,0.28502912521362306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,16,4,power_law_1.01,0.8799040222167969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,16,4,power_law_1.01,1.222586898803711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,16,4,power_law_1.01,1.5372633361816406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,16,4,balanced,8.084869384765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,16,4,power_law_1.01,0.34929279327392576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,16,4,power_law_1.01,2.2012838745117187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,16,4,power_law_1.01,0.20321407318115234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,16,4,power_law_1.01,0.32978561401367185
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,16,4,power_law_1.01,0.35498497009277347
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,16,4,power_law_1.01,2.8840908813476562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,16,4,power_law_1.01,0.5386969757080078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,16,4,power_law_1.01,0.5884812927246094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,16,4,power_law_1.01,0.7804045104980469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,16,4,power_law_1.01,0.8201177978515626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,16,4,balanced,9.869276123046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,16,4,power_law_1.01,0.9773542022705077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,16,4,power_law_1.01,0.8091993713378907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,16,4,power_law_1.01,0.8659865570068359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,16,4,power_law_1.01,4.264112548828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,16,4,power_law_1.01,1.0795903778076172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,16,4,power_law_1.01,1.2239884948730468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,16,4,power_law_1.01,1.0812723541259766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,16,4,power_law_1.01,1.320939483642578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,16,4,power_law_1.01,1.4177792358398438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,16,4,power_law_1.01,5.438363037109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,16,4,power_law_1.01,1.3215129089355468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,16,4,power_law_1.01,1.5953216552734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,16,4,power_law_1.01,1.6384191894531248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,16,4,power_law_1.01,2.3125132751464843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,16,4,power_law_1.01,2.9965939331054687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,16,4,power_law_1.2,0.06162303924560547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,16,4,power_law_1.01,3.5559295654296874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,16,4,power_law_1.2,0.07935999870300293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,16,4,power_law_1.2,0.10425600051879882
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,16,4,power_law_1.01,3.9695730590820313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,16,4,power_law_1.2,0.1332083225250244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,16,4,power_law_1.2,0.15845760345458984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,16,4,power_law_1.2,0.21444736480712892
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,16,4,power_law_1.2,0.222872314453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,16,4,power_law_1.2,0.21985919952392577
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,16,4,power_law_1.2,0.25896064758300785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,16,4,power_law_1.2,0.26820095062255855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,16,4,power_law_1.2,0.29233024597167967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,16,4,power_law_1.2,0.3124595260620117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,16,4,power_law_1.2,0.32203647613525394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,16,4,power_law_1.2,0.35289089202880863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,16,4,power_law_1.2,0.3420620727539062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,16,4,power_law_1.2,0.3523660659790039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,16,4,power_law_1.2,0.39351806640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,16,4,power_law_1.01,5.895789794921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,16,4,power_law_1.2,0.4829580688476563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,16,4,power_law_1.2,0.5915391921997071
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,16,4,power_law_1.2,0.7492185974121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,16,4,power_law_1.01,6.789328613281251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,16,4,power_law_1.2,0.909359359741211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,16,4,power_law_1.2,1.247688980102539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,16,4,power_law_1.2,1.5680076599121093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,16,4,power_law_1.2,0.21838720321655272
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,16,4,power_law_1.2,0.292545280456543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,16,4,power_law_1.2,2.2938841247558592
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,16,4,power_law_1.2,0.3668364715576172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,16,4,power_law_1.2,2.870018615722656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,16,4,power_law_1.2,0.5278464126586914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,16,4,power_law_1.2,0.5975936126708985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,16,4,power_law_1.2,0.7961369323730468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,16,4,power_law_1.2,0.895879669189453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,16,4,power_law_1.2,4.356599121093749
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,16,4,power_law_1.2,0.7903820800781249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,16,4,power_law_1.2,0.9399820709228516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,16,4,power_law_1.2,0.9303756713867187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,16,4,power_law_1.01,12.037620849609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,16,4,power_law_1.2,0.8373798370361328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,16,4,power_law_1.01,1.093232650756836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,16,4,power_law_1.2,1.2906944274902343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,16,4,power_law_1.2,1.268072967529297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,16,4,power_law_1.2,1.3474176025390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,16,4,power_law_1.2,1.1268096160888672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,16,4,power_law_1.2,1.3801510620117188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,16,4,power_law_1.2,1.48671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,16,4,power_law_1.2,1.5594329833984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,16,4,power_law_1.2,1.6342527770996091
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,16,4,power_law_1.2,1.9728909301757813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,16,4,power_law_1.2,2.0293798828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,16,4,power_law_1.2,5.682005615234376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,32,2,balanced,0.03504384040832519
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,32,2,balanced,0.060650238990783686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,16,4,power_law_1.2,3.1082763671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,32,2,balanced,0.08733183860778808
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,32,2,balanced,0.2469503974914551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,32,2,balanced,0.3434688186645508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,16,4,power_law_1.2,4.067723388671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,32,2,balanced,0.3532979202270508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,32,2,balanced,0.3736076736450195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,32,2,balanced,0.14640000343322754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,32,2,balanced,0.35986560821533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,32,2,balanced,0.37392127990722657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,32,2,balanced,0.38451969146728515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,32,2,balanced,0.39958782196044923
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,32,2,balanced,0.4083596801757812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,16,4,power_law_1.2,5.299197387695313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,32,2,balanced,0.42984062194824213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,32,2,balanced,0.4409100723266602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,32,2,balanced,0.45294975280761723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,32,2,balanced,0.47631359100341797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,32,2,balanced,0.5242521667480469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,32,2,balanced,0.5725683212280274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,16,4,power_law_1.2,6.078853149414063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,32,2,balanced,0.9188044738769531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,32,2,balanced,1.017712631225586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,32,2,balanced,1.4494808959960939
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,16,4,power_law_1.01,1.1217702484130858
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,32,2,balanced,1.6464282226562499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,32,2,balanced,0.13542783737182618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,16,4,power_law_1.2,7.222384643554688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,32,2,balanced,3.1354763793945315
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,32,2,balanced,0.23182592391967774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,32,2,balanced,0.43881855010986326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,32,2,balanced,0.8082061004638671
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,32,2,balanced,1.146151657104492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,32,2,balanced,4.5984344482421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,32,2,balanced,1.1531110382080079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,32,2,balanced,2.475741424560547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,32,2,balanced,1.1552089691162108
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,32,2,balanced,5.807761840820312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,16,4,power_law_1.2,13.038118896484374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,32,2,balanced,1.1745228576660156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,32,2,balanced,0.09410304069519043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,32,2,balanced,1.1601945495605468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,32,2,balanced,1.1928307342529298
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,32,2,balanced,1.2797682952880858
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,32,2,balanced,1.3798323059082032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,32,2,balanced,2.0045120239257814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,32,2,balanced,1.4538227844238283
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,32,2,balanced,1.6647424316406252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,32,2,balanced,1.6896102905273438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,32,2,balanced,2.403600616455078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,32,2,balanced,2.380912628173828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,32,2,balanced,2.4923724365234374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,32,2,balanced,1.357224884033203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,32,2,power_law_1.01,0.045203199386596685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,32,2,power_law_1.01,0.10633472442626953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,32,2,balanced,3.189720458984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,32,2,power_law_1.01,0.14648960113525392
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,32,2,power_law_1.01,0.28292863845825195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,32,2,power_law_1.01,0.19709695816040038
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,32,2,power_law_1.01,0.30133375167846677
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,32,2,balanced,4.470759582519531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,32,2,power_law_1.01,0.3224204635620117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,32,2,power_law_1.01,0.34089599609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,32,2,power_law_1.01,0.3819007873535156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,32,2,power_law_1.01,0.41052928924560544
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,32,2,power_law_1.01,0.31161855697631835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,32,2,power_law_1.01,0.44779392242431637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,32,2,balanced,5.78594970703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,32,2,power_law_1.01,0.5255206298828126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,32,2,power_law_1.01,0.4519667053222657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,32,2,power_law_1.01,0.4882124710083008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,32,2,power_law_1.01,0.08065024375915528
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,32,2,power_law_1.01,0.6076761627197266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,32,2,balanced,6.2385321044921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,32,2,power_law_1.01,0.6837567901611328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,32,2,power_law_1.01,0.43225856781005856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,32,2,power_law_1.01,0.8710604858398436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,32,2,power_law_1.01,1.0345471954345702
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,32,2,power_law_1.01,1.4075482177734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,32,2,power_law_1.01,1.7721011352539062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,32,2,power_law_1.01,0.15184639930725097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,32,2,power_law_1.01,2.512138214111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,32,2,power_law_1.01,0.2515571212768555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,32,2,power_law_1.01,0.31859968185424803
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,32,2,power_law_1.01,3.270424194335938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,32,2,power_law_1.01,0.5877824020385742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,32,2,power_law_1.01,0.6805542755126953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,32,2,power_law_1.01,0.9620531463623048
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,32,2,power_law_1.01,1.1540966033935547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,32,2,power_law_1.01,0.9880358123779297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,32,2,power_law_1.01,4.737676696777344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,32,2,power_law_1.01,1.1652403259277344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,32,2,power_law_1.01,1.0339456176757813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,32,2,power_law_1.01,1.0869798278808593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,32,2,power_law_1.01,1.2328844451904297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,32,2,power_law_1.01,1.098831329345703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,32,2,balanced,12.921267089843749
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,32,2,power_law_1.01,1.2040332794189452
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,32,2,power_law_1.01,6.061757202148438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,32,2,power_law_1.01,1.320641326904297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,32,2,power_law_1.01,1.4004698181152344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,32,2,power_law_1.01,1.7085401916503904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,32,2,balanced,15.2100732421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,32,2,power_law_1.01,2.695973205566406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,32,2,power_law_1.01,3.069690856933594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,32,2,power_law_1.01,2.8221554565429687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,32,2,power_law_1.01,3.2615795898437496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,32,2,power_law_1.2,0.04524543762207031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,32,2,power_law_1.2,0.07862016201019287
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,32,2,power_law_1.2,0.10743167877197266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,32,2,power_law_1.01,3.7048229980468745
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,32,2,power_law_1.2,0.1435251235961914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,32,2,power_law_1.2,0.18449407577514648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,32,2,power_law_1.2,0.2819596862792969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,32,2,power_law_1.2,0.2980838394165039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,32,2,power_law_1.2,0.30621952056884766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,32,2,power_law_1.2,0.33481727600097655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,32,2,power_law_1.2,0.35042049407958986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,32,2,power_law_1.2,0.38996864318847657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,32,2,power_law_1.2,0.42986240386962893
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,32,2,power_law_1.2,0.4609651184082031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,32,2,power_law_1.2,0.5364403152465821
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,32,2,power_law_1.2,0.445976333618164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,32,2,power_law_1.2,0.4573952102661133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,32,2,power_law_1.01,5.7695361328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,32,2,power_law_1.2,0.5047065734863281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,32,2,power_law_1.2,0.6208972930908203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,32,2,power_law_1.2,0.6941516876220704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,32,2,power_law_1.2,0.8922099304199218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,32,2,power_law_1.2,1.0524441528320314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,32,2,power_law_1.2,1.4314892578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,32,2,power_law_1.2,1.7892556762695313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,32,2,power_law_1.2,2.5752703857421873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,32,2,power_law_1.01,1.874977264404297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,32,2,power_law_1.01,8.911181030273436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,32,2,power_law_1.2,0.14076928138732908
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,32,2,power_law_1.2,3.323827209472656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,32,2,power_law_1.2,0.43108097076416013
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,32,2,power_law_1.2,0.5795788955688477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,32,2,power_law_1.2,0.567185935974121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,32,2,power_law_1.2,0.9988224029541015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,32,2,power_law_1.2,0.9528486633300781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,32,2,power_law_1.2,0.9999769592285157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,32,2,power_law_1.2,4.859640197753906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,32,2,power_law_1.2,1.0943321228027343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,32,2,power_law_1.2,1.0728806304931642
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,32,2,power_law_1.2,1.1122700500488283
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,32,2,power_law_1.01,13.11445068359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,32,2,power_law_1.2,1.0416844940185546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,32,2,power_law_1.2,1.1887462615966797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,32,2,power_law_1.2,0.21831424713134764
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,32,2,power_law_1.2,1.214851837158203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,32,2,power_law_1.2,1.4592819213867188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,32,2,power_law_1.2,1.6585997009277342
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,32,2,power_law_1.2,2.2732249450683595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,32,2,power_law_1.2,2.4603123474121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,32,2,power_law_1.2,2.0588441467285157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,32,2,power_law_1.2,2.3529638671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,32,2,power_law_1.2,2.5286886596679685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,32,2,power_law_1.2,6.1547393798828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,16,8,balanced,0.025936639308929442
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,16,8,balanced,0.02602240085601807
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,16,8,balanced,0.025950720310211183
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,32,2,power_law_1.01,17.906221923828124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,16,8,balanced,0.0464140796661377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,16,8,balanced,0.0479040002822876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,16,8,balanced,0.04950911998748779
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,16,8,balanced,0.04978176116943359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,32,2,power_law_1.2,3.5156683349609374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,16,8,balanced,0.05016320228576661
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,16,8,balanced,0.05115647792816162
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,16,8,balanced,0.05185408115386962
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,16,8,balanced,0.05180287837982178
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,16,8,balanced,0.05296383857727051
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,16,8,balanced,0.053071360588073734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,16,8,balanced,0.05477888107299804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,16,8,balanced,0.055672321319580075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,16,8,balanced,0.05603583812713623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,16,8,balanced,0.05688831806182861
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,16,8,balanced,0.060015358924865723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,16,8,balanced,0.062452478408813475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,16,8,balanced,0.0701964807510376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,16,8,balanced,0.06653952121734619
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,16,8,balanced,0.08833536148071289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,16,8,balanced,0.10274175643920898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,16,8,balanced,0.12452992439270019
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,16,8,balanced,0.15935104370117187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,16,8,balanced,0.22417024612426756
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,16,8,balanced,0.027645440101623536
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,16,8,balanced,0.29289344787597654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,16,8,balanced,0.028037118911743163
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,16,8,balanced,0.02797312021255493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,16,8,balanced,0.035764479637146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,16,8,balanced,0.03658495903015137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,16,8,balanced,0.036177918910980225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,32,2,power_law_1.2,5.800980224609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,16,8,balanced,0.036327679157257084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,16,8,balanced,0.0366425609588623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,16,8,balanced,0.037085440158844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,16,8,balanced,0.037063679695129394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,16,8,balanced,0.03739007949829102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,32,2,power_law_1.2,5.754423217773438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,16,8,balanced,0.037857279777526856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,16,8,balanced,0.03827455997467041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,16,8,balanced,0.03591808080673218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,16,8,balanced,0.038309121131896974
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,16,8,balanced,0.04028031826019287
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,16,8,balanced,0.04192512035369873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,16,8,balanced,0.04322815895080566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,16,8,balanced,0.04689663887023926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,16,8,balanced,0.044843521118164066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,16,8,balanced,0.050586881637573244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,16,8,balanced,0.05893760204315186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,16,8,balanced,0.07400320053100587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,16,8,balanced,0.036601600646972654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,16,8,balanced,0.08716032028198242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,16,8,balanced,0.12481792449951172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,16,8,balanced,0.1570304012298584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,32,2,power_law_1.2,7.330595703125001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,32,2,power_law_1.2,12.451435546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,32,2,power_law_1.2,14.803765869140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,16,8,balanced,0.0161843204498291
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,16,8,balanced,0.015772160291671753
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,16,8,balanced,0.016179200410842896
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,16,8,balanced,0.01615231990814209
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,16,8,balanced,0.016524800062179566
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,16,8,balanced,0.016503039598464966
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,16,8,balanced,0.016520960330963137
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,16,8,balanced,0.016567039489746093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,16,8,balanced,0.01611647963523865
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,16,8,balanced,0.016592639684677123
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,16,8,balanced,0.01625216007232666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,16,8,balanced,0.01613055944442749
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,16,8,balanced,0.015777280330657957
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,16,8,balanced,0.016609280109405516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,16,8,balanced,0.01621119976043701
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,16,8,balanced,0.016537599563598633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,16,8,balanced,0.016113920211791994
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,16,8,balanced,0.01731071949005127
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,16,8,balanced,0.01739392042160034
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,16,8,balanced,0.01858816027641296
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,16,8,balanced,0.019397120475769043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,16,8,power_law_1.01,0.04341248035430908
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,16,8,power_law_1.01,0.04690176010131836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,16,8,power_law_1.01,0.047989759445190426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,16,8,power_law_1.01,0.049387521743774414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,16,8,power_law_1.01,0.0509887981414795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,16,8,power_law_1.01,0.025623040199279783
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,16,8,balanced,0.021543679237365724
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,16,8,power_law_1.01,0.05224832057952881
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,16,8,power_law_1.01,0.05386112213134766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,16,8,power_law_1.01,0.054764800071716316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,16,8,power_law_1.01,0.05634431838989258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,16,8,power_law_1.01,0.05715583801269532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,16,8,power_law_1.01,0.05888895988464356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,16,8,power_law_1.01,0.06043519973754883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,16,8,power_law_1.01,0.0668287992477417
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,16,8,power_law_1.01,0.06913023948669433
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,16,8,power_law_1.01,0.06888448238372803
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,16,8,power_law_1.01,0.055142397880554195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,16,8,power_law_1.01,0.08124544143676757
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,16,8,balanced,0.022749440670013426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,16,8,power_law_1.01,0.10134271621704101
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,16,8,power_law_1.01,0.1189516830444336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,16,8,power_law_1.01,0.1483903980255127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,16,8,power_law_1.01,0.23624832153320313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,16,8,balanced,0.031106560230255126
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,16,8,balanced,0.034973440170288084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,16,8,power_law_1.01,0.20865152359008787
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,16,8,power_law_1.01,0.027674880027771
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,16,8,power_law_1.01,0.03577343940734863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,16,8,power_law_1.01,0.03622528076171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,16,8,power_law_1.01,0.03624959945678711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,16,8,power_law_1.01,0.036282880306243895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,16,8,power_law_1.01,0.03702143907546997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,16,8,power_law_1.01,0.037474560737609866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,16,8,power_law_1.01,0.037425920963287354
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,16,8,power_law_1.01,0.037876479625701905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,16,8,power_law_1.01,0.038264319896698
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,16,8,power_law_1.01,0.039481599330902103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,16,8,power_law_1.01,0.04069119930267334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,16,8,power_law_1.01,0.041498880386352535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,16,8,power_law_1.01,0.04275968074798584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,16,8,power_law_1.01,0.04518655776977539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,16,8,power_law_1.01,0.046069760322570805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,16,8,power_law_1.01,0.04656511783599854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,16,8,power_law_1.01,0.04929791927337647
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,16,8,power_law_1.01,0.05902592182159424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,16,8,power_law_1.01,0.07018496036529541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,16,8,power_law_1.01,0.3990617752075195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,16,8,power_law_1.01,0.08504960060119629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,16,8,power_law_1.01,0.524218864440918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,16,8,power_law_1.01,0.12665599822998047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,16,8,power_law_1.01,0.1157811164855957
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,16,8,balanced,0.047548160552978516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,16,8,balanced,0.05964288234710693
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,16,8,power_law_1.01,0.016106239557266235
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,16,8,power_law_1.01,0.016191999912261962
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,16,8,power_law_1.01,0.016153600215911865
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,16,8,power_law_1.01,0.01627135992050171
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,16,8,power_law_1.01,0.01625983953475952
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,16,8,power_law_1.01,0.016093440055847168
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,16,8,power_law_1.01,0.016503039598464966
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,16,8,power_law_1.01,0.016176639795303344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,16,8,power_law_1.01,0.0164902400970459
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,16,8,power_law_1.01,0.016560640335083008
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,16,8,power_law_1.01,0.016584960222244264
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,16,8,power_law_1.01,0.01660159945487976
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,16,8,power_law_1.01,0.017315839529037473
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,16,8,power_law_1.01,0.017379839420318604
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,16,8,power_law_1.01,0.8080051422119141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,16,8,power_law_1.01,0.28014848709106444
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,16,8,power_law_1.01,0.020663039684295652
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,16,8,power_law_1.01,0.2175654411315918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,16,8,power_law_1.01,0.3950835037231445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,16,8,power_law_1.01,0.5801958465576172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,16,8,power_law_1.01,0.017986559867858888
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,16,8,power_law_1.01,0.018237440586090087
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,16,8,power_law_1.01,0.01941248059272766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,16,8,power_law_1.01,0.022520320415496824
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,16,8,power_law_1.01,0.02806528091430664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,16,8,power_law_1.01,0.0342579197883606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,16,8,power_law_1.2,0.025544960498809815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,16,8,power_law_1.2,0.04315648078918457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,16,8,power_law_1.2,0.04724224090576172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,16,8,power_law_1.2,0.04714240074157715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,16,8,power_law_1.2,0.049710078239440916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,16,8,power_law_1.2,0.05094016075134278
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,16,8,power_law_1.2,0.0526259183883667
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,16,8,power_law_1.2,0.05385471820831299
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,16,8,power_law_1.2,0.05415040016174316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,16,8,power_law_1.01,1.0453862762451172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,16,8,power_law_1.2,0.05493760108947754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,16,8,power_law_1.2,0.0573145580291748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,16,8,power_law_1.2,0.058728961944580084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,16,8,power_law_1.2,0.06116352081298828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,16,8,power_law_1.2,0.06665215969085694
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,16,8,power_law_1.2,0.056158719062805174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,16,8,power_law_1.2,0.06870399951934815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,16,8,power_law_1.2,0.06812543869018554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,16,8,power_law_1.2,0.08417152404785157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,16,8,power_law_1.2,0.09845888137817382
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,16,8,power_law_1.2,0.1184934425354004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,16,8,power_law_1.2,0.15324416160583496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,16,8,power_law_1.2,0.19992576599121095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,16,8,power_law_1.2,0.27339136123657226
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,16,8,power_law_1.01,0.04608767986297607
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,16,8,power_law_1.2,0.4789952087402344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,16,8,power_law_1.01,0.05940480232238769
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,16,8,power_law_1.2,0.027996160984039307
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,16,8,power_law_1.2,0.03548543930053711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,16,8,power_law_1.2,0.03583487987518311
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,16,8,power_law_1.2,0.03623424053192138
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,16,8,power_law_1.2,0.03629312038421631
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,16,8,power_law_1.2,0.03681663990020752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,16,8,power_law_1.2,0.03699455976486206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,16,8,power_law_1.2,0.037501440048217774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,16,8,power_law_1.2,0.03820800065994263
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,16,8,power_law_1.2,0.03852159976959228
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,16,8,power_law_1.2,0.03988095998764038
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,16,8,power_law_1.2,0.040687360763549806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,16,8,power_law_1.2,0.04117248058319092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,16,8,power_law_1.2,0.04312064170837403
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,16,8,power_law_1.2,0.044764161109924316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,16,8,power_law_1.2,0.41251968383789067
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,16,8,power_law_1.2,0.04648191928863525
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,16,8,power_law_1.01,0.0810483169555664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,16,8,power_law_1.2,0.04527743816375732
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,16,8,power_law_1.2,0.04901375770568848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,16,8,power_law_1.2,0.058410239219665525
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,16,8,power_law_1.2,0.07149184226989747
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,16,8,power_law_1.2,0.08692095756530763
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,16,8,power_law_1.2,0.1084390354156494
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,16,8,power_law_1.01,0.10275712013244628
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,16,8,power_law_1.01,0.13126912117004394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,16,8,power_law_1.2,0.14934271812438965
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,16,8,power_law_1.2,0.015983359813690184
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,16,8,power_law_1.2,0.015878399610519407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,16,8,power_law_1.2,0.22779264450073242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,16,8,power_law_1.2,0.01616768002510071
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,16,8,power_law_1.2,0.016113920211791994
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,16,8,power_law_1.2,0.016171519756317136
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,16,8,power_law_1.2,0.8300070190429688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,16,8,power_law_1.2,0.016163840293884277
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,16,8,power_law_1.2,0.01618559956550598
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,16,8,power_law_1.2,0.016523519754409792
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,16,8,power_law_1.2,0.016592639684677123
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,16,8,power_law_1.2,0.016524800062179566
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,16,8,power_law_1.2,0.01658112049102783
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,16,8,power_law_1.2,0.017341439723968503
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,16,8,power_law_1.2,0.01733504056930542
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,16,8,power_law_1.2,1.0094989013671873
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,16,8,power_law_1.2,0.018576639890670776
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,16,8,power_law_1.2,0.017367039918899534
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,16,8,power_law_1.2,0.018183679580688478
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,16,8,power_law_1.2,0.019366400241851805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,16,8,power_law_1.2,0.021114881038665774
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,16,8,power_law_1.2,0.02273024082183838
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,16,8,power_law_1.2,0.3023654365539551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,16,8,power_law_1.2,0.43724033355712894
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,16,8,power_law_1.01,0.17866880416870118
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,16,8,power_law_1.2,0.047316479682922366
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,16,8,power_law_1.2,0.5795225524902344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,32,4,balanced,0.024099841117858886
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,32,4,balanced,0.023932158946990967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,32,4,balanced,0.026438400745391843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,32,4,balanced,0.04401152133941651
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,32,4,balanced,0.04448128223419189
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,32,4,balanced,0.04684415817260742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,32,4,balanced,0.047238402366638184
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,32,4,balanced,0.04788864135742187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,32,4,balanced,0.04888319969177246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,32,4,balanced,0.04887423992156982
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,32,4,balanced,0.04973951816558838
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,32,4,balanced,0.05080192089080811
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,32,4,balanced,0.0517632007598877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,32,4,balanced,0.05340415954589843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,32,4,balanced,0.05519360065460205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,32,4,balanced,0.056337919235229496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,32,4,balanced,0.05849472045898437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,32,4,balanced,0.06460544109344482
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,32,4,balanced,0.06129151821136475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,32,4,balanced,0.07681407928466796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,32,4,balanced,0.09071488380432129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,32,4,balanced,0.10948479652404786
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,32,4,balanced,0.12626303672790526
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,32,4,balanced,0.15698047637939455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,32,4,balanced,0.20333440780639647
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,32,4,balanced,0.3099750328063965
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,16,8,power_law_1.2,0.0804428768157959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,32,4,balanced,0.026378240585327152
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,32,4,balanced,0.38713985443115234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,32,4,balanced,0.02636928081512451
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,32,4,balanced,0.03538944005966187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,32,4,balanced,0.03537152051925659
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,32,4,balanced,0.035479040145874025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,32,4,balanced,0.03585920095443725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,32,4,balanced,0.035828480720520015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,32,4,balanced,0.03581696033477783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,32,4,balanced,0.0362444806098938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,32,4,balanced,0.03627264022827149
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,32,4,balanced,0.037063679695129394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,32,4,balanced,0.03704960107803344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,32,4,balanced,0.038278400897979736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,32,4,balanced,0.03867392063140869
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,32,4,balanced,0.039496960639953616
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,32,4,balanced,0.04212863922119141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,32,4,balanced,0.04598527908325195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,32,4,balanced,0.04346496105194092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,32,4,balanced,0.04894591808319092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,32,4,balanced,0.05713535785675049
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,32,4,balanced,0.06845183849334717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,32,4,balanced,0.08135295867919921
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,32,4,balanced,0.10654080390930176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,32,4,balanced,0.13260543823242188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,32,4,balanced,0.20073215484619142
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,32,4,balanced,0.25508224487304687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,32,4,balanced,0.029249279499053954
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,16,8,power_law_1.2,0.029434878826141358
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,16,8,power_law_1.2,0.03476736068725586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,16,8,power_law_1.2,0.05707392215728759
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,16,8,power_law_1.2,0.13042304039001465
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,16,8,power_law_1.2,0.10296192169189453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,32,4,balanced,0.016129280328750613
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,32,4,balanced,0.0170035195350647
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,32,4,balanced,0.01675647974014282
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,32,4,balanced,0.017405439615249634
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,32,4,balanced,0.017004799842834473
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,32,4,balanced,0.017382400035858152
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,32,4,balanced,0.01705471992492676
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,32,4,balanced,0.01655295968055725
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,32,4,balanced,0.0174233603477478
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,32,4,balanced,0.017356799840927122
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,32,4,balanced,0.017375999689102174
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,32,4,balanced,0.0173363196849823
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,32,4,balanced,0.016536320447921755
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,32,4,balanced,0.016940799951553346
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,32,4,balanced,0.016543999910354615
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,32,4,balanced,0.01693824052810669
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,32,4,balanced,0.016633599996566772
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,32,4,balanced,0.0186188805103302
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,32,4,balanced,0.019397120475769043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,32,4,balanced,0.022259199619293214
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,32,4,balanced,0.021029119491577146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,32,4,power_law_1.01,0.02354687929153442
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,32,4,power_law_1.01,0.025633280277252195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,32,4,power_law_1.01,0.04363391876220703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,32,4,power_law_1.01,0.04519807815551758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,32,4,power_law_1.01,0.046407680511474605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,32,4,power_law_1.01,0.04725503921508789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,32,4,power_law_1.01,0.04913919925689697
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,32,4,power_law_1.01,0.050195198059082034
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,32,4,power_law_1.01,0.05134463787078858
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,32,4,power_law_1.01,0.05200128078460693
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,32,4,power_law_1.01,0.052887039184570314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,32,4,power_law_1.01,0.05454592227935791
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,32,4,power_law_1.01,0.055671038627624514
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,32,4,power_law_1.01,0.05844607830047608
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,32,4,power_law_1.01,0.06164608001708984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,32,4,power_law_1.01,0.06249343872070312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,32,4,power_law_1.01,0.0735487985610962
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,32,4,power_law_1.01,0.08027520179748535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,32,4,power_law_1.01,0.09331711769104004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,32,4,power_law_1.01,0.1075648021697998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,32,4,power_law_1.01,0.1255244827270508
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,32,4,balanced,0.025180160999298096
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,32,4,balanced,0.028488960266113285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,32,4,balanced,0.04006271839141846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,32,4,power_law_1.01,0.16350080490112304
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,32,4,balanced,0.04720384120941162
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,32,4,balanced,0.06683008193969726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,32,4,power_law_1.01,0.0272652792930603
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,16,8,power_law_1.2,0.1856755256652832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,32,4,power_law_1.01,0.028956160545349118
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,32,4,power_law_1.01,0.03527551889419556
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,32,4,power_law_1.01,0.0358515191078186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,32,4,power_law_1.01,0.03533184051513672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,32,4,power_law_1.01,0.03609855890274048
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,32,4,power_law_1.01,0.036551680564880375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,32,4,power_law_1.01,0.036753919124603275
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,32,4,power_law_1.01,0.03717504024505615
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,32,4,power_law_1.01,0.03751935958862305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,32,4,power_law_1.01,0.03836416006088257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,32,4,power_law_1.01,0.21456256866455078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,32,4,power_law_1.01,0.039560959339141846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,32,4,power_law_1.01,0.04032127857208252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,32,4,power_law_1.01,0.0442137622833252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,32,4,power_law_1.01,0.041521921157836914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,32,4,power_law_1.01,0.044824318885803224
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,32,4,power_law_1.01,0.04902143955230713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,32,4,power_law_1.01,0.04605184078216552
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,32,4,power_law_1.01,0.05654528141021729
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,32,4,power_law_1.01,0.06864384174346924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,32,4,power_law_1.01,0.07972608089447022
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,32,4,balanced,0.08123904228210449
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,32,4,power_law_1.01,0.1025049591064453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,32,4,power_law_1.01,0.1347097587585449
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,32,4,power_law_1.01,0.3160332870483399
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,32,4,power_law_1.01,0.01627519965171814
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,32,4,power_law_1.01,0.016524800062179566
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,32,4,power_law_1.01,0.016698880195617678
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,32,4,power_law_1.01,0.016631040573120117
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,32,4,power_law_1.01,0.01692800045013428
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,32,4,power_law_1.01,0.5822758483886719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,32,4,power_law_1.01,0.016755199432373045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,32,4,power_law_1.01,0.19748607635498047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,32,4,power_law_1.01,0.0169267201423645
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,32,4,power_law_1.01,0.01719167947769165
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,32,4,power_law_1.01,0.01704192042350769
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,32,4,power_law_1.01,0.0173363196849823
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,32,4,power_law_1.01,0.01720960021018982
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,32,4,power_law_1.01,0.01815808057785034
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,32,4,power_law_1.01,0.4374003219604492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,32,4,power_law_1.01,0.2769753646850586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,32,4,power_law_1.01,0.02061311960220337
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,32,4,power_law_1.01,0.3710451126098633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,32,4,power_law_1.01,0.018636800050735473
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,32,4,power_law_1.01,0.018169599771499633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,32,4,power_law_1.01,0.019407360553741454
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,32,4,power_law_1.01,0.019118080139160158
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,32,4,power_law_1.01,0.5238604736328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,32,4,power_law_1.01,0.05107840061187744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,32,4,power_law_1.01,0.7143679809570312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,32,4,power_law_1.01,0.02179840087890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,32,4,power_law_1.01,0.025181438922882084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,32,4,power_law_1.2,0.02437119960784912
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,32,4,power_law_1.2,0.02610048055648804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,32,4,power_law_1.2,0.04444799900054931
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,32,4,power_law_1.2,0.04482816219329834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,32,4,power_law_1.2,0.046476798057556154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,32,4,power_law_1.2,0.048078079223632816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,32,4,power_law_1.2,0.0491264009475708
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,32,4,power_law_1.2,0.050496001243591306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,32,4,power_law_1.2,0.05138688087463379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,32,4,power_law_1.2,0.05215104103088379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,32,4,power_law_1.2,0.05337984085083007
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,32,4,power_law_1.2,0.05543039798736572
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,32,4,power_law_1.2,0.05543039798736572
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,32,4,power_law_1.2,0.05835775852203369
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,32,4,power_law_1.2,0.061201920509338384
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,32,4,power_law_1.2,0.06418560028076173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,32,4,power_law_1.2,0.06655231952667237
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,32,4,power_law_1.2,0.07993216037750243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,32,4,power_law_1.2,0.09164416313171386
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,32,4,power_law_1.2,0.10750080108642579
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,32,4,power_law_1.01,0.029602560997009274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,32,4,power_law_1.2,0.12462464332580567
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,32,4,power_law_1.01,0.03677567958831787
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,32,4,power_law_1.01,0.05809408187866211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,32,4,power_law_1.2,0.17425024032592773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,32,4,power_law_1.2,0.20174591064453126
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,32,4,power_law_1.01,0.07646592140197754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,32,4,power_law_1.2,0.026362879276275637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,32,4,power_law_1.2,0.028552958965301512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,32,4,power_law_1.2,0.035024640560150144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,32,4,power_law_1.2,0.03559168100357056
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,32,4,power_law_1.01,0.08776960372924805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,32,4,power_law_1.2,0.03582463979721069
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,32,4,power_law_1.2,0.03584127902984619
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,32,4,power_law_1.2,0.6393254470825196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,32,4,power_law_1.2,0.036756479740142824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,32,4,power_law_1.2,0.036917760372161865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,32,4,power_law_1.2,0.03764991998672486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,32,4,power_law_1.2,0.03871999979019165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,32,4,power_law_1.2,0.03988095998764038
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,32,4,power_law_1.2,0.04000127792358398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,32,4,power_law_1.2,0.03619328022003174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,32,4,power_law_1.2,0.041994237899780275
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,32,4,power_law_1.2,0.04403200149536133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,32,4,power_law_1.2,0.04602367877960205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,32,4,power_law_1.2,0.044961280822753906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,32,4,power_law_1.2,0.04880512237548828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,32,4,power_law_1.2,0.3013478469848633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,32,4,power_law_1.2,0.058950400352478026
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,32,4,power_law_1.2,0.06998144149780273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,32,4,power_law_1.2,0.08154751777648925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,32,4,power_law_1.2,0.5979891204833985
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,32,4,power_law_1.01,0.13596287727355957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,32,4,power_law_1.2,0.10514559745788574
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,32,4,power_law_1.2,0.01610751986503601
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,32,4,power_law_1.2,0.01655168056488037
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,32,4,power_law_1.2,0.016593919992446898
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,32,4,power_law_1.2,0.01668992042541504
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,32,4,power_law_1.2,0.016524800062179566
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,32,4,power_law_1.2,0.016910079717636108
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,32,4,power_law_1.2,0.016940799951553346
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,32,4,power_law_1.2,0.017351679801940918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,32,4,power_law_1.2,0.01698815941810608
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,32,4,power_law_1.2,0.017349120378494263
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,32,4,power_law_1.2,0.01701120018959045
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,32,4,power_law_1.2,0.018179839849472045
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,32,4,power_law_1.2,0.01824895977973938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,32,4,power_law_1.2,0.01825919985771179
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,32,4,power_law_1.2,0.019383039474487305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,32,4,power_law_1.2,0.13305472373962401
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,32,4,power_law_1.2,0.01943295955657959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,32,4,power_law_1.2,0.201790714263916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,32,4,power_law_1.2,0.43499519348144533
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,32,4,power_law_1.2,0.027279360294342043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,32,4,power_law_1.2,0.38318592071533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,32,4,power_law_1.2,0.020610558986663818
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,32,4,power_law_1.2,0.021862399578094483
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,32,4,power_law_1.2,0.26009471893310543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,32,4,power_law_1.2,0.025148160457611084
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,32,4,power_law_1.2,0.03657984018325806
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,32,4,power_law_1.01,0.16384639739990234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,16,8,balanced,0.03335040092468262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,16,8,balanced,0.033278720378875734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,16,8,balanced,0.03338752031326294
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,16,8,balanced,0.05836800098419189
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,16,8,balanced,0.06120704174041748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,16,8,balanced,0.06426368236541749
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,16,8,balanced,0.06523647785186767
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,16,8,balanced,0.06568319797515869
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,16,8,balanced,0.06709760189056396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,16,8,balanced,0.06737919807434081
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,16,8,balanced,0.06851840019226074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,16,8,balanced,0.07032447814941406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,16,8,balanced,0.07116032123565674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,16,8,balanced,0.07256447792053222
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,16,8,balanced,0.07417856216430664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,16,8,balanced,0.0740339183807373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,16,8,balanced,0.07629824161529541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,16,8,balanced,0.08155008316040038
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,16,8,balanced,0.08476415634155274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,16,8,balanced,0.11279104232788087
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,16,8,balanced,0.11365887641906738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,16,8,balanced,0.1383027172088623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,16,8,balanced,0.15364864349365234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,16,8,balanced,0.18993791580200195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,16,8,balanced,0.24580736160278321
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,16,8,balanced,0.3565491104125976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,16,8,balanced,0.45256576538085935
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,16,8,balanced,0.032956159114837645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,16,8,balanced,0.03304320096969605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,16,8,balanced,0.0329638409614563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,16,8,balanced,0.04279679775238037
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,16,8,balanced,0.04239488124847412
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,32,4,power_law_1.2,0.04447103977203369
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,16,8,balanced,0.042442240715026856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,16,8,balanced,0.04276351928710938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,16,8,balanced,0.042516479492187495
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,16,8,balanced,0.043151359558105465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,16,8,balanced,0.04318336009979248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,16,8,balanced,0.04278272151947021
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,16,8,balanced,0.04314367771148682
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,16,8,balanced,0.04321023941040039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,16,8,balanced,0.04358016014099121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,16,8,balanced,0.04402304172515869
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,16,8,balanced,0.04399744033813476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,16,8,balanced,0.04609792232513428
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,16,8,balanced,0.04850304126739502
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,16,8,balanced,0.050608639717102055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,16,8,balanced,0.05750271797180175
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,16,8,balanced,0.05589888095855713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,16,8,balanced,0.06822144031524659
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,16,8,balanced,0.07745791912078857
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,16,8,balanced,0.09807999610900879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,32,4,power_law_1.2,0.5031475067138672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,16,8,balanced,0.17555328369140627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,16,8,balanced,0.2243891143798828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,16,8,balanced,0.11689472198486328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,32,4,power_law_1.2,0.055493121147155766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,32,4,power_law_1.2,0.07630847930908204
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,16,8,balanced,0.01739135980606079
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,16,8,balanced,0.01778815984725952
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,16,8,balanced,0.01744256019592285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,16,8,balanced,0.01742591977119446
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,16,8,balanced,0.017749760150909424
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,16,8,balanced,0.01777024030685425
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,16,8,balanced,0.0174835205078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,16,8,balanced,0.017816319465637206
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,16,8,balanced,0.017822719812393188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,16,8,balanced,0.01790336012840271
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,16,8,balanced,0.017775360345840454
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,16,8,balanced,0.018173439502716066
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,16,8,balanced,0.017791999578475954
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,16,8,balanced,0.018143999576568603
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,16,8,balanced,0.018568960428237916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,32,4,power_law_1.2,0.13301888465881348
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,32,4,power_law_1.2,0.09593471527099609
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,16,8,balanced,0.01771903991699219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,16,8,balanced,0.017779200077056884
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,32,4,power_law_1.2,0.16236160278320313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,16,8,balanced,0.01942271947860718
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,16,8,balanced,0.020661759376525878
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,16,8,balanced,0.01940608024597168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,16,8,power_law_1.01,0.0329638409614563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,16,8,power_law_1.01,0.0587391996383667
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,16,8,power_law_1.01,0.0610368013381958
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,16,8,power_law_1.01,0.06085760116577148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,16,8,power_law_1.01,0.0641919994354248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,16,8,power_law_1.01,0.06738944053649902
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,16,8,power_law_1.01,0.06978687763214111
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,16,8,balanced,0.022673919200897216
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,16,8,power_law_1.01,0.071244797706604
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,16,8,power_law_1.01,0.07273344039916993
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,16,8,power_law_1.01,0.07680384159088136
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,16,8,power_law_1.01,0.07586559772491455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,16,8,power_law_1.01,0.07876480102539063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,16,8,power_law_1.01,0.08017151832580567
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,16,8,power_law_1.01,0.0849728012084961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,16,8,power_law_1.01,0.10407296180725098
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,16,8,power_law_1.01,0.10988415718078612
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,16,8,power_law_1.01,0.11635711669921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,16,8,power_law_1.01,0.13388287544250488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,16,8,power_law_1.01,0.15017600059509278
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,16,8,power_law_1.01,0.18333440780639648
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,16,8,balanced,0.024321279525756835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,16,8,power_law_1.01,0.24624128341674806
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,16,8,balanced,0.026832640171051025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,16,8,power_law_1.01,0.3124928092956543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,16,8,power_law_1.01,0.4137382507324219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,16,8,balanced,0.037557759284973145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,16,8,power_law_1.01,0.03217152118682861
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,16,8,power_law_1.01,0.041751041412353515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,16,8,power_law_1.01,0.042327041625976565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,16,8,power_law_1.01,0.04232319831848145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,16,8,power_law_1.01,0.042388482093811033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,16,8,power_law_1.01,0.042789120674133305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,16,8,power_law_1.01,0.04280704021453857
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,16,8,power_law_1.01,0.04315264225006103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,16,8,power_law_1.01,0.04356736183166504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,16,8,power_law_1.01,0.04373119831085205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,16,8,power_law_1.01,0.04485375881195068
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,16,8,power_law_1.01,0.04684160232543945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,16,8,power_law_1.01,0.04766335964202881
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,16,8,power_law_1.01,0.04807295799255371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,16,8,power_law_1.01,0.051016960144042965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,16,8,power_law_1.01,0.054617600440979006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,16,8,power_law_1.01,0.05651072025299072
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,16,8,power_law_1.01,0.6137919998168945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,16,8,power_law_1.01,0.06847487926483155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,16,8,power_law_1.01,0.07528063774108887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,16,8,power_law_1.01,0.0879321575164795
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,16,8,balanced,0.045505280494689944
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,16,8,power_law_1.01,0.7961984252929687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,16,8,power_law_1.01,0.10766464233398439
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,16,8,power_law_1.01,0.17037055969238282
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,16,8,balanced,0.07952383995056153
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,16,8,balanced,0.06253056049346924
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,16,8,power_law_1.01,0.01740031957626343
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,16,8,power_law_1.01,0.01739904046058655
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,16,8,power_law_1.01,0.017763839960098265
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,16,8,power_law_1.01,0.017355519533157348
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,16,8,power_law_1.01,0.017790720462799073
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,16,8,power_law_1.01,0.01781888008117676
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,16,8,power_law_1.01,0.01777791976928711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,16,8,power_law_1.01,0.01787008047103882
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,16,8,power_law_1.01,0.01777791976928711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,16,8,power_law_1.01,0.01821951985359192
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,16,8,power_law_1.01,0.018214399814605715
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,16,8,power_law_1.01,0.019032319784164427
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,16,8,power_law_1.01,0.018920960426330565
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,16,8,power_law_1.01,0.019353599548339845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,16,8,power_law_1.01,0.3929702377319336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,16,8,power_law_1.01,0.22369920730590823
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,16,8,power_law_1.01,0.2970060729980469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,16,8,power_law_1.01,0.5441100692749024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,16,8,power_law_1.01,1.2459827423095704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,16,8,power_law_1.01,1.6219020080566406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,16,8,power_law_1.01,0.7437580871582031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,16,8,power_law_1.01,0.022342400550842283
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,16,8,power_law_1.01,0.020385279655456542
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,16,8,power_law_1.01,0.020217599868774413
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,16,8,power_law_1.01,0.0240447998046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,16,8,power_law_1.01,0.026455039978027346
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,16,8,power_law_1.01,0.03575680017471313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,16,8,power_law_1.2,0.03291392087936401
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,16,8,power_law_1.2,0.058791680335998533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,16,8,power_law_1.2,0.060784640312194826
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,16,8,power_law_1.2,0.05994880199432373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,16,8,power_law_1.2,0.06491648197174073
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,16,8,power_law_1.2,0.06773759841918944
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,16,8,power_law_1.2,0.07024511814117432
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,16,8,power_law_1.2,0.07142399787902833
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,16,8,power_law_1.2,0.0734438419342041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,16,8,power_law_1.2,0.07369472026824951
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,16,8,power_law_1.2,0.0754252815246582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,16,8,power_law_1.2,0.08003328323364259
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,16,8,power_law_1.2,0.08090496063232422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,16,8,power_law_1.01,0.045180158615112306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,16,8,power_law_1.2,0.09233023643493651
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,16,8,power_law_1.2,0.10509056091308593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,16,8,power_law_1.2,0.11428352355957032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,16,8,power_law_1.2,0.11819647789001464
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,16,8,power_law_1.2,0.13251711845397948
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,16,8,power_law_1.2,0.1508889579772949
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,16,8,power_law_1.01,0.06155648231506348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,16,8,power_law_1.2,0.18338176727294922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,16,8,power_law_1.2,0.24596992492675782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,16,8,power_law_1.2,0.35186561584472653
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,16,8,power_law_1.01,0.10729472160339355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,16,8,power_law_1.2,0.4508211135864258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,16,8,power_law_1.01,0.07876095771789551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,16,8,power_law_1.2,0.03249151945114136
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,16,8,power_law_1.2,0.04213503837585449
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,16,8,power_law_1.2,0.04272895812988281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,16,8,power_law_1.2,0.04233856201171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,16,8,power_law_1.2,0.04272768020629883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,16,8,power_law_1.2,0.04320511817932129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,16,8,power_law_1.2,0.04312575817108154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,16,8,power_law_1.2,0.043550720214843755
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,16,8,power_law_1.2,0.044002561569213866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,16,8,power_law_1.2,0.04436736106872559
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,16,8,power_law_1.2,0.045664000511169436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,16,8,power_law_1.2,0.04723968029022217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,16,8,power_law_1.2,0.04851712226867676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,16,8,power_law_1.2,0.04991360187530518
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,16,8,power_law_1.2,0.05189760208129883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,16,8,power_law_1.2,0.05500671863555908
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,16,8,power_law_1.01,0.1283993625640869
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,16,8,power_law_1.2,0.05550079822540284
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,16,8,power_law_1.2,0.06688127994537355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,16,8,power_law_1.2,0.07575039863586426
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,16,8,power_law_1.01,0.17286144256591798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,16,8,power_law_1.2,0.08890751838684083
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,16,8,power_law_1.2,0.106179838180542
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,16,8,power_law_1.2,0.6188659286499023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,16,8,power_law_1.2,0.14697983741760254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,16,8,power_law_1.2,0.2091187286376953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,16,8,power_law_1.01,0.24312320709228516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,16,8,power_law_1.2,0.8722943878173828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,16,8,power_law_1.2,0.017725440263748168
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,16,8,power_law_1.2,0.018145279884338377
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,16,8,power_law_1.2,0.017786879539489747
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,16,8,power_law_1.2,0.01817471981048584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,16,8,power_law_1.2,0.017799680233001706
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,16,8,power_law_1.2,0.018213119506835938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,16,8,power_law_1.2,0.01828096032142639
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,16,8,power_law_1.2,0.01821951985359192
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,16,8,power_law_1.2,0.0181388795375824
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,16,8,power_law_1.2,0.018531839847564697
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,16,8,power_law_1.2,1.2932928466796876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,16,8,power_law_1.2,0.019420160055160524
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,16,8,power_law_1.2,0.019031039476394652
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,16,8,power_law_1.2,0.32171646118164066
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,16,8,power_law_1.2,0.020309760570526122
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,16,8,power_law_1.2,0.6036940765380859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,16,8,power_law_1.2,0.42671871185302734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,16,8,power_law_1.2,0.6949209594726563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,16,8,power_law_1.2,0.034330880641937254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,16,8,power_law_1.2,1.644011535644531
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,16,8,power_law_1.2,0.01803264021873474
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,16,8,power_law_1.2,0.01903488039970398
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,16,8,power_law_1.2,0.020631039142608644
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,16,8,power_law_1.2,0.02202879905700684
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,16,8,power_law_1.2,0.02379647970199585
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,16,8,power_law_1.2,0.026616320610046384
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,32,4,balanced,0.030887680053710936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,32,4,balanced,0.03050112009048462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,32,4,balanced,0.03336832046508789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,32,4,balanced,0.057894401550292976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,32,4,balanced,0.06058623790740967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,32,4,balanced,0.06278016090393065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,32,4,balanced,0.06357888221740723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,32,4,balanced,0.06415999889373779
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,32,4,balanced,0.06526976108551026
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,32,4,balanced,0.06597631931304931
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,32,4,balanced,0.06684288024902343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,32,4,balanced,0.06859263896942139
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,32,4,balanced,0.06975232124328613
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,32,4,balanced,0.07193855762481689
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,32,4,balanced,0.07517824172973633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,32,4,balanced,0.07598464012145996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,32,4,balanced,0.07813632011413574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,32,4,balanced,0.0964799976348877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,32,4,balanced,0.10567296028137207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,32,4,balanced,0.12452095985412596
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,32,4,balanced,0.13880831718444825
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,32,4,balanced,0.16148479461669923
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,32,4,balanced,0.1840716743469238
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,32,4,balanced,0.23449344635009767
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,32,4,balanced,0.31471872329711914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,32,4,balanced,0.45477630615234377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,32,4,balanced,0.587663345336914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,32,4,balanced,0.03091200113296509
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,16,8,power_law_1.2,0.04288896083831787
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,32,4,balanced,0.030993919372558593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,32,4,balanced,0.03292671918869018
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,32,4,balanced,0.04035327911376953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,32,4,balanced,0.0411084794998169
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,32,4,balanced,0.04124800205230713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,32,4,balanced,0.041113600730896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,32,4,balanced,0.04119167804718017
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,32,4,balanced,0.041706237792968746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,32,4,balanced,0.04193280220031738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,32,4,balanced,0.042004480361938476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,32,4,balanced,0.04240511894226075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,32,4,balanced,0.04322559833526611
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,32,4,balanced,0.04403327941894532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,32,4,balanced,0.0456063985824585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,32,4,balanced,0.04939263820648193
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,32,4,balanced,0.04153728008270264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,32,4,balanced,0.054944000244140624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,32,4,balanced,0.05362304210662842
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,16,8,power_law_1.2,0.06102911949157715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,32,4,balanced,0.0644659185409546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,32,4,balanced,0.07268608093261719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,32,4,balanced,0.09123456001281738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,32,4,balanced,0.10712448120117188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,32,4,balanced,0.14435456275939942
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,32,4,balanced,0.18521343231201173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,32,4,balanced,0.2766579246520996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,32,4,balanced,0.3512985610961914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,16,8,power_law_1.2,0.18790271759033203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,16,8,power_law_1.2,0.07539328098297118
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,16,8,power_law_1.2,0.10721920013427735
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,16,8,power_law_1.2,0.11749247550964355
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,32,4,balanced,0.017757439613342287
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,32,4,balanced,0.018572800159454346
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,32,4,balanced,0.018152960538864137
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,32,4,balanced,0.018179839849472045
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,32,4,balanced,0.01821568012237549
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,32,4,balanced,0.018570239543914797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,32,4,balanced,0.018944000005722043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,32,4,balanced,0.01856384038925171
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,32,4,balanced,0.017834240198135377
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,32,4,balanced,0.018586879968643187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,32,4,balanced,0.01906048059463501
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,32,4,balanced,0.018981120586395263
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,32,4,balanced,0.019384319782257083
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,16,8,power_law_1.2,0.22162687301635745
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,32,4,balanced,0.01824128031730652
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,32,4,balanced,0.018257919549942016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,32,4,balanced,0.018559999465942383
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,32,4,balanced,0.018298879861831666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,32,4,balanced,0.021089279651641847
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,32,4,balanced,0.02340735912322998
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,32,4,balanced,0.022272000312805178
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,32,4,power_law_1.01,0.0308735990524292
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,32,4,power_law_1.01,0.03334784030914307
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,32,4,power_law_1.01,0.05870463848114014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,32,4,power_law_1.01,0.06036479949951172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,32,4,power_law_1.01,0.06281087875366212
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,32,4,power_law_1.01,0.06445312023162841
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,32,4,power_law_1.01,0.066561279296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,32,4,power_law_1.01,0.06770944118499755
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,32,4,power_law_1.01,0.0698905611038208
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,32,4,balanced,0.026060800552368164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,32,4,power_law_1.01,0.0701043176651001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,32,4,power_law_1.01,0.07156479835510254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,32,4,power_law_1.01,0.07395199775695802
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,32,4,power_law_1.01,0.07538303852081299
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,32,4,power_law_1.01,0.07791744232177734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,32,4,power_law_1.01,0.0906278419494629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,32,4,power_law_1.01,0.10743680000305175
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,32,4,power_law_1.01,0.11073792457580567
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,32,4,power_law_1.01,0.12437376022338867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,32,4,power_law_1.01,0.1412659168243408
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,32,4,balanced,0.02984447956085205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,32,4,power_law_1.01,0.16408319473266603
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,32,4,balanced,0.06268159866333008
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,32,4,balanced,0.052796158790588375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,32,4,balanced,0.03620223999023438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,32,4,power_law_1.01,0.261900806427002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,32,4,power_law_1.01,0.19844736099243163
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,32,4,power_law_1.01,0.030933759212493896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,32,4,power_law_1.01,0.0329036808013916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,32,4,power_law_1.01,0.04074240207672119
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,32,4,power_law_1.01,0.040797438621521
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,32,4,power_law_1.01,0.040871682167053225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,32,4,power_law_1.01,0.04150527954101563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,32,4,power_law_1.01,0.04159359931945801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,32,4,power_law_1.01,0.3331123352050781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,32,4,power_law_1.01,0.042502398490905764
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,32,4,power_law_1.01,0.042812161445617676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,32,4,power_law_1.01,0.04402431964874268
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,32,4,power_law_1.01,0.04517375946044922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,32,4,power_law_1.01,0.04608767986297607
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,32,4,power_law_1.01,0.04199039936065674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,32,4,power_law_1.01,0.047655677795410155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,32,4,power_law_1.01,0.05099008083343506
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,32,4,power_law_1.01,0.05305088043212891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,32,4,power_law_1.01,0.05506303787231446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,32,4,power_law_1.01,0.45557502746582035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,32,4,power_law_1.01,0.06589439868927002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,32,4,power_law_1.01,0.07397888183593751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,32,4,power_law_1.01,0.09100159645080566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,32,4,power_law_1.01,0.10751487731933593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,32,4,power_law_1.01,0.14809856414794922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,32,4,balanced,0.08458111763000489
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,32,4,power_law_1.01,0.19592575073242186
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,32,4,balanced,0.10605055809020995
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,32,4,power_law_1.01,0.017771519422531128
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,32,4,power_law_1.01,0.018010879755020144
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,32,4,power_law_1.01,0.01783936023712158
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,32,4,power_law_1.01,0.01821568012237549
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,32,4,power_law_1.01,0.01818112015724182
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,32,4,power_law_1.01,0.018247679471969605
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,32,4,power_law_1.01,0.018575359582901
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,32,4,power_law_1.01,0.01858816027641296
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,32,4,power_law_1.01,0.01900032043457031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,32,4,power_law_1.01,0.018585599660873413
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,32,4,power_law_1.01,0.018983680009841918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,32,4,power_law_1.01,0.6345318222045899
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,32,4,power_law_1.01,0.35889793395996095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,32,4,power_law_1.01,1.1371916961669921
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,32,4,power_law_1.01,0.2924224090576172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,32,4,power_law_1.01,0.942630386352539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,32,4,power_law_1.01,0.6692121887207032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,32,4,power_law_1.01,0.5065651321411133
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,32,4,power_law_1.01,0.01954303979873657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,32,4,power_law_1.01,0.019827200174331667
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,32,4,power_law_1.01,0.020254719257354736
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,32,4,power_law_1.01,0.021475839614868163
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,32,4,power_law_1.01,0.021556479930877687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,32,4,power_law_1.01,0.02350464105606079
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,32,4,power_law_1.01,0.027251200675964354
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,32,4,power_law_1.01,0.03012480020523071
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,32,4,power_law_1.01,0.037738239765167235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,32,4,power_law_1.2,0.030926079750061036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,32,4,power_law_1.2,0.0334169602394104
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,32,4,power_law_1.2,0.058721280097961424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,32,4,power_law_1.2,0.05974400043487549
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,32,4,power_law_1.2,0.06290431976318359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,32,4,power_law_1.2,0.06493824005126954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,32,4,power_law_1.2,0.0669158411026001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,32,4,power_law_1.2,0.068472318649292
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,32,4,power_law_1.2,0.06986752033233642
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,32,4,power_law_1.2,0.07028863906860353
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,32,4,power_law_1.2,0.07248000144958497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,32,4,power_law_1.2,0.0747648000717163
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,32,4,power_law_1.2,0.07558144092559814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,32,4,power_law_1.2,0.08514176368713379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,32,4,power_law_1.2,0.09766400337219239
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,32,4,power_law_1.01,0.045263361930847165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,32,4,power_law_1.2,0.1163212776184082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,32,4,power_law_1.2,0.10265600204467773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,32,4,power_law_1.2,0.12361472129821778
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,32,4,power_law_1.01,0.06613759994506836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,32,4,power_law_1.2,0.1401254367828369
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,32,4,power_law_1.01,0.07525631904602051
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,32,4,power_law_1.2,0.1655731201171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,32,4,power_law_1.2,0.19324928283691406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,32,4,power_law_1.2,0.25669504165649415
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,32,4,power_law_1.01,0.11274751663208007
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,32,4,power_law_1.2,0.3366796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,32,4,power_law_1.01,0.12012672424316406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,32,4,power_law_1.2,0.03168384075164795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,32,4,power_law_1.2,0.0337280011177063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,32,4,power_law_1.2,0.04118656158447266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,32,4,power_law_1.2,0.041520638465881346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,32,4,power_law_1.2,0.04118271827697754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,32,4,power_law_1.2,0.041820158958435064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,32,4,power_law_1.2,0.04194303989410401
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,32,4,power_law_1.2,0.042411518096923825
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,32,4,power_law_1.2,0.04315264225006103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,32,4,power_law_1.2,0.04322559833526611
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,32,4,power_law_1.2,0.04387968063354492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,32,4,power_law_1.2,0.045194239616394044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,32,4,power_law_1.2,0.04653696060180664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,32,4,power_law_1.2,0.048903679847717284
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,32,4,power_law_1.2,0.05097856044769287
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,32,4,power_law_1.2,0.05349120140075684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,32,4,power_law_1.2,0.05478784084320069
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,32,4,power_law_1.01,0.1882035255432129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,32,4,power_law_1.2,0.06569983959197997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,32,4,power_law_1.2,0.46064769744873046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,32,4,power_law_1.2,0.07391615867614745
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,32,4,power_law_1.2,0.08947199821472168
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,32,4,power_law_1.01,0.21793407440185547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,32,4,power_law_1.2,0.10716287612915039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,32,4,power_law_1.2,0.14807680130004883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,32,4,power_law_1.2,0.195994873046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,32,4,power_law_1.2,0.017802239656448365
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,32,4,power_law_1.2,0.017716480493545534
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,32,4,power_law_1.2,0.018163199424743655
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,32,4,power_law_1.2,0.01820032000541687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,32,4,power_law_1.2,0.018330880403518678
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,32,4,power_law_1.2,0.018238719701766968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,32,4,power_law_1.2,0.01856511950492859
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,32,4,power_law_1.2,0.018568960428237916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,32,4,power_law_1.2,0.018955520391464233
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,32,4,power_law_1.2,0.01856768012046814
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,32,4,power_law_1.2,0.018711040019989016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,32,4,power_law_1.2,0.9211750030517578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,32,4,power_law_1.2,0.34928768157958984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,32,4,power_law_1.2,0.019815679788589478
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,32,4,power_law_1.2,0.020244479179382324
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,32,4,power_law_1.2,0.021775360107421877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,32,4,power_law_1.2,0.6456896209716797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,32,4,power_law_1.2,0.29787263870239256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,32,4,power_law_1.2,0.5213375854492187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,32,4,power_law_1.2,0.6705318450927734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,32,4,power_law_1.2,0.048244481086730955
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,32,4,power_law_1.2,0.02004096031188965
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,32,4,power_law_1.2,0.021907200813293455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,32,4,power_law_1.2,1.2699967956542968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,32,4,power_law_1.2,0.023111679553985597
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,32,4,power_law_1.2,0.025606400966644287
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,32,4,power_law_1.2,0.07234687805175781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,128,balanced,0.06755328178405762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,128,balanced,0.06696959972381591
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,128,balanced,0.06710527896881104
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,128,balanced,0.06740479946136475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,128,balanced,0.06707071781158447
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,128,balanced,0.07825535774230957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,128,balanced,0.1383948802947998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,128,balanced,0.13814656257629393
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,32,4,power_law_1.2,0.030640640258789063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,128,balanced,0.1386291217803955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,128,balanced,0.1391155242919922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,128,balanced,0.14107904434204102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,128,balanced,0.14089599609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,128,balanced,0.1418252754211426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,128,balanced,0.144837121963501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,128,balanced,0.14001919746398925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,128,balanced,0.15499391555786132
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,128,balanced,0.15187968254089354
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,128,balanced,0.1564684772491455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,128,balanced,0.1682841682434082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,128,balanced,0.1608550453186035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,128,balanced,0.14499967575073242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,128,balanced,0.17532543182373045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,128,balanced,0.18665599822998047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,128,balanced,0.20879871368408204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,128,balanced,0.234279670715332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,128,balanced,0.3175321578979492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,128,balanced,0.3828736114501953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,128,balanced,0.03803136110305786
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,128,balanced,0.03815167903900146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,128,balanced,0.037916159629821776
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,128,balanced,0.03793920040130615
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,128,balanced,0.04379392147064209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,128,balanced,0.05055615901947021
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,128,balanced,0.05076735973358154
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,32,4,power_law_1.2,0.03660543918609619
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,128,balanced,0.050942721366882326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,128,balanced,0.050531840324401854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,128,balanced,0.05065983772277832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,128,balanced,0.05255936145782471
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,128,balanced,0.03827327966690063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,128,balanced,0.05188479900360108
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,128,balanced,0.053071360588073734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,128,balanced,0.054527997970581055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,128,balanced,0.05884416103363037
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,128,balanced,0.05555071830749512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,128,balanced,0.05655807971954345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,128,balanced,0.06374400138854981
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,128,balanced,0.0701580810546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,128,balanced,0.06671743869781495
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,128,balanced,0.07557248115539551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,128,balanced,0.08311167716979981
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,128,balanced,0.09757439613342285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,128,balanced,0.10891008377075195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,128,balanced,0.13875455856323243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,128,balanced,0.1730790328979492
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,32,4,power_law_1.2,0.06364927768707276
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,32,4,power_law_1.2,0.1789311981201172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,32,4,power_law_1.2,0.11420800209045409
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,1,128,balanced,0.020305919647216796
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,1,128,balanced,0.020573439598083495
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,1,128,balanced,0.02108544111251831
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,1,128,balanced,0.021103360652923585
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,1,128,balanced,0.021002240180969238
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,1,128,balanced,0.02109312057495117
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,1,128,balanced,0.021105918884277344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,1,128,balanced,0.021073920726776125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,1,128,balanced,0.021025280952453613
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,1,128,balanced,0.021486079692840575
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,1,128,balanced,0.021510400772094727
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,1,128,balanced,0.02144511938095093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,1,128,balanced,0.021610240936279296
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,1,128,balanced,0.021447679996490478
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,1,128,balanced,0.02165503978729248
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,1,128,balanced,0.021495039463043215
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,1,128,balanced,0.02223360061645508
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,1,128,balanced,0.022269439697265626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,1,128,balanced,0.022726399898529055
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,1,128,balanced,0.020648961067199705
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,32,4,power_law_1.2,0.1342527961730957
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,1,128,balanced,0.02060544013977051
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,128,power_law_1.01,0.0781606388092041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,128,power_law_1.01,0.07844351768493653
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,128,power_law_1.01,0.07862143993377686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,128,power_law_1.01,0.07831424236297607
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,128,power_law_1.01,0.13872511863708498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,128,power_law_1.01,0.1416703987121582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,128,power_law_1.01,0.1421363162994385
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,128,power_law_1.01,0.14368512153625487
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,128,power_law_1.01,0.1460275173187256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,128,power_law_1.01,0.14926976203918457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,128,power_law_1.01,0.151146240234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,128,power_law_1.01,0.15040127754211424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,128,power_law_1.01,0.15073792457580565
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,32,4,power_law_1.2,0.24900096893310547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,128,power_law_1.01,0.1599193572998047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,128,power_law_1.01,0.16302719116210937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,128,power_law_1.01,0.16847488403320313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,128,power_law_1.01,0.16999168395996095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,128,power_law_1.01,0.17869951248168944
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,128,power_law_1.01,0.19974655151367188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,128,power_law_1.01,0.24933759689331056
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,128,power_law_1.01,0.30727424621582033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,128,power_law_1.01,0.4288307189941406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,128,power_law_1.01,0.6462054443359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,128,power_law_1.01,0.8410111999511718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,128,power_law_1.01,0.9898252868652344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,128,power_law_1.01,1.5553779602050781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,128,power_law_1.01,0.04520063877105713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,128,power_law_1.01,0.044843521118164066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,128,power_law_1.01,0.04817535877227783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,128,power_law_1.01,2.317136688232422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,128,power_law_1.01,0.0452569580078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,128,power_law_1.01,0.0526963186264038
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,128,power_law_1.01,0.0549235200881958
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,128,power_law_1.01,0.053131518363952634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,128,power_law_1.01,0.05576704025268555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,128,power_law_1.01,0.055244798660278316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,128,power_law_1.01,0.05990272045135499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,128,power_law_1.01,0.059618558883666996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,128,power_law_1.01,0.05635712146759033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,128,power_law_1.01,0.060577278137207036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,128,power_law_1.01,0.06452095985412598
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,128,power_law_1.01,0.06875904083251953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,128,power_law_1.01,0.0671129608154297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,128,power_law_1.01,0.07460351943969726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,128,power_law_1.01,0.07803904056549073
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,128,power_law_1.01,0.08936320304870606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,128,power_law_1.01,0.1130303955078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,128,power_law_1.01,0.12167296409606934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,128,power_law_1.01,0.1673561668395996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,128,power_law_1.01,0.23252992630004882
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,128,power_law_1.01,0.3443199920654297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,1,128,balanced,0.02031359910964966
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,128,power_law_1.01,0.42959743499755854
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,1,128,power_law_1.01,0.02110464096069336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,1,128,power_law_1.01,0.021063680648803713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,128,power_law_1.01,0.7264972686767578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,1,128,power_law_1.01,0.02109055995941162
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,128,power_law_1.01,0.8913394927978515
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,1,128,power_law_1.01,0.021025280952453613
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,1,128,power_law_1.01,0.021246719360351562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,1,128,power_law_1.01,0.02109951972961426
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,1,128,power_law_1.01,0.021251840591430666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,1,128,power_law_1.01,0.021619200706481934
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,1,128,power_law_1.01,0.02183552026748657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,1,128,power_law_1.01,0.02182528018951416
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,1,128,power_law_1.01,0.021863679885864257
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,1,128,power_law_1.01,0.021857280731201172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,1,128,power_law_1.01,0.02218240022659302
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,1,128,power_law_1.01,0.021459200382232667
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,1,128,power_law_1.01,0.02176896095275879
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,1,128,power_law_1.01,0.021954560279846193
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,1,128,balanced,0.02063744068145752
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,1,128,balanced,0.029164800643920897
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,1,128,balanced,0.028959999084472655
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,1,128,balanced,0.04236032009124756
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,1,128,balanced,0.04766592025756836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,1,128,power_law_1.01,0.021505279541015627
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,1,128,power_law_1.01,0.025958399772644043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,1,128,power_law_1.01,0.035543038845062255
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,1,128,power_law_1.01,0.02772480010986328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,128,power_law_1.2,0.07829504013061524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,128,power_law_1.2,0.07849984169006348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,128,power_law_1.2,0.07841407775878906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,128,power_law_1.2,0.13895039558410643
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,128,power_law_1.2,0.13890815734863282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,128,power_law_1.2,0.14322943687438966
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,128,power_law_1.2,0.14393343925476074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,128,power_law_1.2,0.1454092788696289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,128,power_law_1.2,0.1471628761291504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,128,power_law_1.2,0.15185536384582518
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,128,power_law_1.2,0.14983936309814455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,128,power_law_1.2,0.15233152389526367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,128,power_law_1.2,0.1545753574371338
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,128,power_law_1.2,0.16420095443725585
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,1,128,power_law_1.01,0.04257535934448242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,128,power_law_1.2,0.16783231735229492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,128,power_law_1.2,0.17191808700561523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,128,power_law_1.2,0.17535232543945312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,128,power_law_1.2,0.20003328323364258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,1,128,power_law_1.01,0.05117440223693848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,128,power_law_1.2,0.23333503723144533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,128,power_law_1.2,0.32751232147216797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,128,power_law_1.2,0.3535539245605469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,128,power_law_1.2,0.6155571365356446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,128,power_law_1.2,0.8989785766601562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,1,128,power_law_1.01,0.07165952205657959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,128,power_law_1.2,1.2403968048095702
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,128,power_law_1.2,0.04370816230773926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,128,power_law_1.2,0.043598079681396486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,128,power_law_1.2,0.0437286376953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,128,power_law_1.2,1.8541413879394533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,128,power_law_1.2,0.04925695896148682
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,128,power_law_1.2,0.05263743877410889
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,128,power_law_1.2,0.051423997879028314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,128,power_law_1.2,0.05344384193420411
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,128,power_law_1.2,0.05278848171234131
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,128,power_law_1.2,0.05867008209228516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,128,power_law_1.2,0.05576064109802246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,128,power_law_1.2,0.061413121223449704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,128,power_law_1.2,0.06058623790740967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,128,power_law_1.2,2.9002725219726564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,128,power_law_1.2,0.06098688125610352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,128,power_law_1.2,0.06885632038116454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,128,power_law_1.2,0.0736294412612915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,128,power_law_1.2,0.07386879920959473
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,128,power_law_1.2,0.07328383922576905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,128,power_law_1.2,0.08626432418823242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,128,power_law_1.2,0.09609600067138671
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,128,power_law_1.2,0.11955583572387696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,128,power_law_1.2,0.17589632034301758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,128,power_law_1.2,3.636093444824219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,128,power_law_1.2,0.24693119049072268
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,128,power_law_1.2,0.33581695556640623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,128,power_law_1.2,0.5019494247436523
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,1,128,power_law_1.2,0.021072640419006347
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,128,power_law_1.2,0.7532569885253906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,1,128,power_law_1.2,0.02102783918380737
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,128,power_law_1.2,0.989617919921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,1,128,power_law_1.2,0.021456639766693115
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,1,128,power_law_1.2,0.021063680648803713
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,1,128,power_law_1.01,0.08700160026550294
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,1,128,power_law_1.2,0.021276159286499025
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,1,128,power_law_1.2,0.02106879949569702
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,1,128,power_law_1.2,0.02183295965194702
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,1,128,power_law_1.2,0.02173439979553223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,128,power_law_1.2,1.516768035888672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,1,128,power_law_1.2,0.021765120029449463
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,1,128,power_law_1.2,0.02181504011154175
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,1,128,power_law_1.2,0.021497600078582764
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,1,128,power_law_1.2,0.02187648057937622
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,1,128,power_law_1.2,0.02217600107192993
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,1,128,power_law_1.2,0.02192512035369873
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,1,128,power_law_1.2,0.027084798812866212
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,1,128,power_law_1.2,0.0210752010345459
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,1,128,power_law_1.01,0.11468159675598144
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,1,128,power_law_1.01,0.1879987144470215
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,1,128,power_law_1.01,0.17135488510131835
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,1,128,power_law_1.2,0.021047039031982424
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,1,128,power_law_1.2,0.031069440841674806
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,1,128,power_law_1.2,0.04203519821166992
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,1,128,power_law_1.2,0.02480128049850464
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,1,128,power_law_1.2,0.054257922172546394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,64,balanced,0.04030208110809326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,64,balanced,0.03666815996170044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,64,balanced,0.0395686411857605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,64,balanced,0.03993599891662598
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,64,balanced,0.06172800064086914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,64,balanced,0.07412735939025879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,64,balanced,0.12840319633483888
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,64,balanced,0.12857343673706054
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,64,balanced,0.1305075168609619
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,64,balanced,0.1305305576324463
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,64,balanced,0.13153663635253907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,64,balanced,0.13332223892211914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,64,balanced,0.13364352226257323
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,64,balanced,0.13543423652648925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,64,balanced,0.13702015876770018
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,64,balanced,0.14026623725891113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,64,balanced,0.1441267204284668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,64,balanced,0.1533836841583252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,64,balanced,0.1489740753173828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,64,balanced,0.16089727401733397
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,64,balanced,0.16672128677368164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,64,balanced,0.18782848358154297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,64,balanced,0.204138240814209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,64,balanced,0.24571647644042968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,64,balanced,0.2831590461730957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,64,balanced,0.3686118316650391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,64,balanced,0.035423998832702634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,64,balanced,0.48696063995361333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,64,balanced,0.035031039714813236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,64,balanced,0.03866751909255982
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,64,balanced,0.035470080375671384
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,64,balanced,0.037774078845977783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,64,balanced,0.04114304065704345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,64,balanced,0.04824960231781006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,64,balanced,0.05131904125213623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,64,balanced,0.04892032146453858
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,64,balanced,0.05183487892150879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,64,balanced,0.049364480972290035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,64,balanced,0.05629951953887939
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,64,balanced,0.05142784118652344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,64,balanced,0.061329917907714845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,64,balanced,0.0556006383895874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,64,balanced,0.058021121025085456
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,64,balanced,0.058620162010192864
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,64,balanced,0.06951039791107178
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,64,balanced,0.06404223918914795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,64,balanced,0.07121920108795166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,64,balanced,0.07746175765991212
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,64,balanced,0.08559871673583984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,64,balanced,0.0947481632232666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,64,balanced,0.12205183982849122
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,64,balanced,0.1441907215118408
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,64,balanced,0.19303295135498047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,64,balanced,0.2414336013793945
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,1,128,power_law_1.2,0.07135871887207032
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,1,128,power_law_1.2,0.08957951545715333
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,1,128,power_law_1.2,0.13315584182739257
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,1,128,power_law_1.2,0.23141504287719727
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,1,128,power_law_1.2,0.34576766967773437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,1,128,power_law_1.2,0.16247423171997072
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,2,64,balanced,0.019860479831695557
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,2,64,balanced,0.02225152015686035
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,2,64,balanced,0.022704639434814454
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,2,64,balanced,0.022709760665893555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,2,64,balanced,0.022741758823394777
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,2,64,balanced,0.023130879402160645
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,2,64,balanced,0.023055360317230225
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,2,64,balanced,0.023152639865875242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,2,64,balanced,0.02309119939804077
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,2,64,balanced,0.02312448024749756
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,2,64,balanced,0.019896320104598998
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,2,64,balanced,0.023141119480133057
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,2,64,balanced,0.023252480030059815
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,2,64,balanced,0.02363264083862305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,2,64,balanced,0.023466238975524904
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,2,64,balanced,0.02476799964904785
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,2,64,balanced,0.025201919078826907
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,2,64,balanced,0.019619840383529662
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,2,64,balanced,0.019814399480819703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,2,64,balanced,0.020363519191741945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,64,power_law_1.01,0.07329279899597167
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,64,power_law_1.01,0.07327871799468995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,64,power_law_1.01,0.07395584106445312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,64,power_law_1.01,0.07379327774047852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,64,power_law_1.01,0.1257151985168457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,64,power_law_1.01,0.12685055732727052
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,64,power_law_1.01,0.13115391731262208
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,64,power_law_1.01,0.1320908832550049
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,64,power_law_1.01,0.13240320205688477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,64,power_law_1.01,0.13444607734680175
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,64,power_law_1.01,0.13734911918640136
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,64,power_law_1.01,0.14027647972106932
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,64,power_law_1.01,0.14131327629089357
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,64,power_law_1.01,0.15507200241088867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,64,power_law_1.01,0.15754112243652343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,64,power_law_1.01,0.1540287971496582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,64,power_law_1.01,0.1642995262145996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,64,power_law_1.01,0.17259008407592774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,64,power_law_1.01,0.18815616607666014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,64,power_law_1.01,0.22757888793945313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,64,power_law_1.01,0.27582208633422856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,64,power_law_1.01,0.36273281097412113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,64,power_law_1.01,0.5035251235961914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,64,power_law_1.01,0.692547836303711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,2,64,balanced,0.02267008066177368
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,2,64,balanced,0.022301440238952634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,64,power_law_1.01,0.9084095764160157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,64,power_law_1.01,0.04199808120727539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,64,power_law_1.01,0.04158592224121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,64,power_law_1.01,1.3159986877441407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,2,64,balanced,0.022661120891571046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,64,power_law_1.01,0.042263040542602534
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,64,power_law_1.01,0.04358143806457519
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,64,power_law_1.01,0.04839424133300781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,64,power_law_1.01,0.047653121948242186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,64,power_law_1.01,1.6997080993652343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,64,power_law_1.01,0.04927743911743164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,64,power_law_1.01,0.054608640670776365
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,64,power_law_1.01,0.05231743812561035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,64,power_law_1.01,0.0573414421081543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,64,power_law_1.01,0.05517951965332031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,64,power_law_1.01,0.05797632217407227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,64,power_law_1.01,0.05716735839843749
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,64,power_law_1.01,0.064268798828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,64,power_law_1.01,0.06863232135772705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,64,power_law_1.01,0.07037055969238282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,64,power_law_1.01,0.07078271865844726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,64,power_law_1.01,0.0768998384475708
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,64,power_law_1.01,0.09031295776367188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,64,power_law_1.01,0.1061094379425049
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,64,power_law_1.01,0.12653183937072754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,64,power_law_1.01,0.19699199676513673
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,64,power_law_1.01,0.20569343566894532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,64,power_law_1.01,0.31041664123535156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,2,64,power_law_1.01,0.022753279209136962
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,64,power_law_1.01,0.40770687103271486
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,2,64,power_law_1.01,0.02257024049758911
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,64,power_law_1.01,0.6348672103881836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,2,64,power_law_1.01,0.022680320739746095
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,2,64,power_law_1.01,0.02273792028427124
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,2,64,power_law_1.01,0.021401600837707518
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,2,64,power_law_1.01,0.02185472011566162
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,2,64,power_law_1.01,0.022259199619293214
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,2,64,power_law_1.01,0.022641921043395997
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,2,64,power_law_1.01,0.02273792028427124
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,2,64,power_law_1.01,0.022672638893127442
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,2,64,power_law_1.01,0.022657279968261716
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,2,64,power_law_1.01,0.02283008098602295
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,2,64,power_law_1.01,0.022655999660491942
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,2,64,power_law_1.01,0.022380800247192384
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,2,64,power_law_1.01,0.02266239881515503
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,2,64,power_law_1.01,0.023121919631958008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,64,power_law_1.01,0.8255833435058595
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,2,64,power_law_1.01,0.022685439586639405
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,2,64,balanced,0.033145599365234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,2,64,balanced,0.03555711984634399
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,2,64,balanced,0.0498419189453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,2,64,balanced,0.057556481361389154
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,2,64,power_law_1.01,0.026374399662017822
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,2,64,power_law_1.01,0.03340800046920776
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,2,64,power_law_1.01,0.02915839910507202
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,64,power_law_1.2,0.07346432209014893
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,64,power_law_1.2,0.07392384052276611
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,2,64,power_law_1.01,0.039959039688110355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,64,power_law_1.2,0.07434751987457275
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,64,power_law_1.2,0.12572287559509276
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,64,power_law_1.2,0.1260185623168945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,64,power_law_1.2,0.12869888305664062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,64,power_law_1.2,0.13331456184387208
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,2,64,power_law_1.01,0.05253503799438477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,64,power_law_1.2,0.1319257640838623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,64,power_law_1.2,0.13532544136047364
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,64,power_law_1.2,0.14019071578979492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,64,power_law_1.2,0.1400089645385742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,64,power_law_1.2,0.14338687896728514
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,64,power_law_1.2,0.14820608139038086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,64,power_law_1.2,0.15566080093383788
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,64,power_law_1.2,0.15371392250061036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,64,power_law_1.2,0.16726015090942384
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,64,power_law_1.2,0.1836684799194336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,64,power_law_1.2,0.21278335571289064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,64,power_law_1.2,0.2727807998657227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,64,power_law_1.2,0.3543756866455078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,64,power_law_1.2,0.4826867294311524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,64,power_law_1.2,0.16023551940917968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,2,64,power_law_1.01,0.06188032150268554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,64,power_law_1.2,0.6499878692626953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,64,power_law_1.2,0.9627442932128906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,64,power_law_1.2,0.0397158408164978
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,2,64,power_law_1.01,0.08301952362060547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,64,power_law_1.2,1.2697241973876952
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,64,power_law_1.2,0.04321663856506348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,64,power_law_1.2,0.043751678466796874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,64,power_law_1.2,0.04144512176513672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,64,power_law_1.2,1.8172991943359373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,64,power_law_1.2,0.04768511772155762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,64,power_law_1.2,0.04819968223571777
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,64,power_law_1.2,0.049964799880981445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,64,power_law_1.2,0.05224319934844971
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,64,power_law_1.2,0.05501440048217774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,64,power_law_1.2,0.053034238815307624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,64,power_law_1.2,0.05594880104064941
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,64,power_law_1.2,0.06135680198669433
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,64,power_law_1.2,0.06758912086486817
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,64,power_law_1.2,0.06815231800079345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,64,power_law_1.2,0.07227519989013673
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,64,power_law_1.2,2.5513587951660157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,64,power_law_1.2,0.07582208156585693
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,64,power_law_1.2,0.08680447578430175
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,64,power_law_1.2,0.09912832260131836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,64,power_law_1.2,0.11858559608459474
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,2,64,power_law_1.01,0.09496576309204101
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,64,power_law_1.2,0.15958656311035155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,64,power_law_1.2,0.21565311431884765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,64,power_law_1.2,0.2737881660461426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,64,power_law_1.2,0.4194073486328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,64,power_law_1.2,0.5952320098876953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,2,64,power_law_1.2,0.022650880813598635
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,2,64,power_law_1.2,0.022828800678253176
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,2,64,power_law_1.2,0.02228096008300781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,64,power_law_1.2,0.05721471786499024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,64,power_law_1.2,0.9320909118652343
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,2,64,power_law_1.2,0.02230015993118286
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,2,64,power_law_1.2,0.021032960414886476
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,2,64,power_law_1.2,0.021441280841827393
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,2,64,power_law_1.2,0.022293760776519775
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,2,64,power_law_1.2,0.022672638893127442
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,2,64,power_law_1.2,0.022855679988861084
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,2,64,power_law_1.2,0.02230655908584595
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,2,64,power_law_1.2,0.022700800895690917
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,64,power_law_1.2,1.2750809478759766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,2,64,power_law_1.2,0.022274560928344726
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,2,64,power_law_1.2,0.022328319549560546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,2,64,power_law_1.2,0.022649600505828857
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,2,64,power_law_1.2,0.02303999900817871
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,2,64,power_law_1.2,0.02914432048797607
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,2,64,power_law_1.2,0.022296319007873537
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,2,64,power_law_1.01,0.14481663703918457
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,2,64,power_law_1.01,0.17433343887329103
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,2,64,power_law_1.2,0.04844543933868408
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,2,64,power_law_1.2,0.02473599910736084
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,2,64,power_law_1.2,0.033282558917999264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,32,balanced,0.03252736091613769
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,2,64,power_law_1.2,0.07527552127838136
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,32,balanced,0.03217407941818237
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,2,64,power_law_1.2,0.04108672142028809
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,32,balanced,0.03256704092025757
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,32,balanced,0.03583103895187378
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,32,balanced,0.060497918128967286
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,32,balanced,0.075764479637146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,32,balanced,0.12605567932128908
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,2,64,power_law_1.2,0.06439167976379394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,32,balanced,0.12905856132507326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,32,balanced,0.13287424087524413
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,32,balanced,0.1319487953186035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,32,balanced,0.13411456108093261
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,32,balanced,0.13610112190246582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,32,balanced,0.14077695846557617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,32,balanced,0.14337151527404784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,32,balanced,0.15078399658203126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,32,balanced,0.14209792137145996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,32,balanced,0.1531430435180664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,32,balanced,0.1610163116455078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,32,balanced,0.1744576072692871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,32,balanced,0.18781055450439452
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,32,balanced,0.22491519927978515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,32,balanced,0.2557107162475586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,32,balanced,0.320318717956543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,32,balanced,0.13099007606506347
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,32,balanced,0.033807361125946046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,32,balanced,0.3852249526977539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,32,balanced,0.03418368101119995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,32,balanced,0.5264281463623047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,32,balanced,0.03377408027648926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,32,balanced,0.03584383964538575
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,32,balanced,0.036601600646972654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,32,balanced,0.04649472236633301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,32,balanced,0.6983679962158204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,32,balanced,0.053143038749694824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,32,balanced,0.05554687976837158
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,32,balanced,0.058922238349914544
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,32,balanced,0.06131328105926513
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,32,balanced,0.060684800148010254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,32,balanced,0.05874815940856933
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,32,balanced,0.06033152103424072
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,32,balanced,0.066909441947937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,32,balanced,0.06660607814788819
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,32,balanced,0.07383935928344727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,32,balanced,0.06539135932922363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,32,balanced,0.07289855957031251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,32,balanced,0.07837823867797852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,32,balanced,0.08486399650573731
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,32,balanced,0.09262720108032227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,32,balanced,0.11388159751892091
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,32,balanced,0.13086463928222655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,32,balanced,0.17869184494018556
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,32,balanced,0.2158758354187012
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,32,balanced,0.3018380737304688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,32,balanced,0.38670337677001954
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,2,64,power_law_1.2,0.10447360038757325
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,2,64,power_law_1.2,0.14285696029663086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,2,64,power_law_1.2,0.20282880783081056
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,2,64,power_law_1.2,0.2767705535888672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,4,32,balanced,0.01986304044723511
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,4,32,balanced,0.024774401187896727
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,4,32,balanced,0.025120000839233398
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,4,32,balanced,0.025160961151123047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,4,32,balanced,0.019759360551834106
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,4,32,balanced,0.02550784111022949
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,4,32,balanced,0.025596160888671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,4,32,balanced,0.025614080429077146
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,4,32,balanced,0.024747519493103026
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,4,32,balanced,0.02607232093811035
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,4,32,balanced,0.02482304096221924
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,4,32,balanced,0.026350080966949463
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,4,32,balanced,0.02518784046173096
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,4,32,balanced,0.02877568006515503
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,4,32,balanced,0.019622399806976318
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,4,32,balanced,0.028671998977661133
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,4,32,balanced,0.03210880041122437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,4,32,balanced,0.03216511964797973
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,4,32,balanced,0.020257279872894288
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,4,32,balanced,0.021923840045928955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,32,power_law_1.01,0.07428224086761474
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,4,32,balanced,0.025519359111785888
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,32,power_law_1.01,0.07405439853668214
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,32,power_law_1.01,0.07376512050628661
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,32,power_law_1.01,0.07696640014648437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,32,power_law_1.01,0.1240563201904297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,32,power_law_1.01,0.12581248283386232
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,32,power_law_1.01,0.12675583839416504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,32,power_law_1.01,0.1322175979614258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,32,power_law_1.01,0.13196928024291993
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,32,power_law_1.01,0.13362303733825684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,32,power_law_1.01,0.14008447647094727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,32,power_law_1.01,0.136495361328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,32,power_law_1.01,0.14197759628295897
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,32,power_law_1.01,0.15000831604003906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,32,power_law_1.01,0.15164799690246583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,32,power_law_1.01,0.15554816246032716
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,32,power_law_1.01,0.17509632110595702
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,32,power_law_1.01,0.14910592079162596
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,32,power_law_1.01,0.19263999938964843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,32,power_law_1.01,0.2164659118652344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,32,power_law_1.01,0.2698419189453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,32,power_law_1.01,0.32272384643554686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,32,power_law_1.01,0.44312961578369137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,32,power_law_1.01,0.6557234954833985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,32,power_law_1.01,0.7579456329345703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,32,power_law_1.01,1.2323187255859376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,32,power_law_1.01,0.043783679008483886
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,32,power_law_1.01,0.04519807815551758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,32,power_law_1.01,1.536267547607422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,32,power_law_1.01,0.04611455917358399
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,32,power_law_1.01,0.050348801612854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,32,power_law_1.01,0.04704895973205566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,32,power_law_1.01,0.056419839859008794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,32,power_law_1.01,0.054552321434021
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,32,power_law_1.01,0.06131072044372558
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,32,power_law_1.01,0.057208318710327145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,32,power_law_1.01,0.05885824203491211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,32,power_law_1.01,0.060861439704895014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,32,power_law_1.01,0.0649292802810669
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,32,power_law_1.01,0.06717055797576904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,32,power_law_1.01,0.07293568134307862
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,32,power_law_1.01,0.06778624057769775
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,32,power_law_1.01,0.07173120021820069
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,32,power_law_1.01,0.07213056087493896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,32,power_law_1.01,0.0847116756439209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,32,power_law_1.01,0.09447808265686035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,32,power_law_1.01,0.11902079582214356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,32,power_law_1.01,0.12904576301574705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,32,power_law_1.01,0.17650432586669923
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,32,power_law_1.01,0.22040191650390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,32,power_law_1.01,0.33785472869873046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,32,power_law_1.01,0.4279321670532227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,32,power_law_1.01,0.6794022369384766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,4,32,balanced,0.024529919624328614
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,4,32,power_law_1.01,0.024788479804992675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,32,power_law_1.01,0.8873331451416016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,4,32,power_law_1.01,0.024302079677581787
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,4,32,power_law_1.01,0.025022718906402584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,4,32,power_law_1.01,0.02430720090866089
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,4,32,power_law_1.01,0.024496641159057617
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,4,32,power_law_1.01,0.024138240814208983
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,4,32,power_law_1.01,0.023865599632263184
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,4,32,power_law_1.01,0.02390399932861328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,4,32,power_law_1.01,0.024686079025268554
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,4,32,balanced,0.02476288080215454
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,4,32,power_law_1.01,0.024528639316558836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,4,32,power_law_1.01,0.024977920055389406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,4,32,power_law_1.01,0.02551039934158325
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,4,32,power_law_1.01,0.024779520034790038
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,4,32,power_law_1.01,0.025401599407196045
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,4,32,power_law_1.01,0.025108480453491212
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,4,32,power_law_1.01,0.027310080528259277
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,4,32,power_law_1.01,0.026784000396728513
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,4,32,balanced,0.04748415946960449
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,4,32,balanced,0.0441484785079956
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,4,32,balanced,0.074301438331604
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,4,32,balanced,0.06460415840148925
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,4,32,power_law_1.01,0.04589312076568604
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,4,32,power_law_1.01,0.0294323205947876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,4,32,power_law_1.01,0.039121921062469485
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,4,32,power_law_1.01,0.03258239984512329
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,32,power_law_1.2,0.07359744071960449
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,32,power_law_1.2,0.07369344234466553
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,32,power_law_1.2,0.07361152172088622
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,4,32,power_law_1.01,0.059758081436157226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,32,power_law_1.2,0.07404543876647948
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,32,power_law_1.2,0.12501631736755373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,32,power_law_1.2,0.1223910427093506
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,32,power_law_1.2,0.13025152206420898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,32,power_law_1.2,0.12941696166992187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,32,power_law_1.2,0.13230976104736328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,32,power_law_1.2,0.1383795166015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,4,32,power_law_1.01,0.07318272113800048
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,32,power_law_1.2,0.14171520233154297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,32,power_law_1.2,0.14509440422058106
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,32,power_law_1.2,0.1475980758666992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,32,power_law_1.2,0.1507520008087158
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,32,power_law_1.2,0.1539097595214844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,32,power_law_1.2,0.1634867286682129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,32,power_law_1.2,0.18436992645263672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,32,power_law_1.2,0.1980556869506836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,32,power_law_1.2,0.2387494468688965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,32,power_law_1.2,0.1328819179534912
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,32,power_law_1.2,0.29039615631103516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,4,32,power_law_1.01,0.09974783897399903
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,32,power_law_1.2,0.4089664077758789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,4,32,power_law_1.01,0.1148646354675293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,32,power_law_1.2,0.6169152069091797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,32,power_law_1.2,0.04767104148864746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,32,power_law_1.2,0.7610470581054687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,32,power_law_1.2,0.0448089599609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,32,power_law_1.2,0.04729216098785401
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,32,power_law_1.2,0.04353663921356201
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,32,power_law_1.2,1.166165771484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,32,power_law_1.2,0.05164927959442138
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,32,power_law_1.2,0.05717631816864014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,32,power_law_1.2,0.05755904197692872
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,32,power_law_1.2,0.06035456180572509
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,32,power_law_1.2,0.05591807842254639
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,32,power_law_1.2,0.06137472152709961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,32,power_law_1.2,0.06431488037109376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,32,power_law_1.2,1.7672320556640624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,32,power_law_1.2,0.07114880084991455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,32,power_law_1.2,0.07117184162139892
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,32,power_law_1.2,0.07286399841308594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,32,power_law_1.2,0.07497087955474854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,32,power_law_1.2,0.07217152118682861
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,32,power_law_1.2,0.08906368255615235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,32,power_law_1.2,0.10380160331726074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,32,power_law_1.2,0.11691264152526855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,32,power_law_1.2,0.14505984306335448
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,32,power_law_1.2,2.3453721618652343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,32,power_law_1.2,0.20002559661865232
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,32,power_law_1.2,0.29164159774780274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,32,power_law_1.2,0.06002048015594482
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,32,power_law_1.2,0.3988300704956055
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,4,32,power_law_1.2,0.02483583927154541
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,4,32,power_law_1.2,0.025626881122589113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,32,power_law_1.2,0.5700019073486329
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,4,32,power_law_1.2,0.024408318996429444
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,4,32,power_law_1.2,0.02450687885284424
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,4,32,power_law_1.2,0.02513535976409912
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,4,32,power_law_1.2,0.0239846396446228
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,4,32,power_law_1.2,0.023915519714355467
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,4,32,power_law_1.2,0.02436352014541626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,32,power_law_1.2,0.8922969818115234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,4,32,power_law_1.2,0.025804800987243654
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,4,32,power_law_1.2,0.025304319858551028
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,4,32,power_law_1.2,0.024698879718780518
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,4,32,power_law_1.2,0.025183999538421632
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,4,32,power_law_1.2,0.026638081073760988
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,32,power_law_1.2,1.1395558166503907
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,4,32,power_law_1.2,0.02595455884933472
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,4,32,power_law_1.2,0.026835200786590574
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,4,32,power_law_1.2,0.02956928014755249
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,4,32,power_law_1.2,0.05301760196685791
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,4,32,power_law_1.2,0.024321279525756835
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,4,32,power_law_1.01,0.15897600173950194
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,4,32,power_law_1.01,0.1917888069152832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,4,32,power_law_1.2,0.03765376091003418
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,4,32,power_law_1.2,0.09027711868286133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,8,16,balanced,0.032065279483795166
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,4,32,power_law_1.2,0.07489664077758788
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,8,16,balanced,0.03175807952880859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,8,16,balanced,0.033447680473327634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,8,16,balanced,0.03615744113922119
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,4,32,power_law_1.2,0.045516800880432126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,8,16,balanced,0.06279551982879639
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,4,32,power_law_1.2,0.03206144094467163
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,8,16,balanced,0.08084735870361329
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,8,16,balanced,0.13415040016174315
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,8,16,balanced,0.13852160453796386
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,8,16,balanced,0.1331059169769287
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,8,16,balanced,0.141911039352417
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,8,16,balanced,0.1464268779754639
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,8,16,balanced,0.15056127548217774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,8,16,balanced,0.1434124755859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,8,16,balanced,0.14868608474731446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,8,16,balanced,0.15269248008728026
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,4,32,power_law_1.2,0.12782591819763184
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,8,16,balanced,0.1611622428894043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,8,16,balanced,0.1722483253479004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,8,16,balanced,0.1831488037109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,8,16,balanced,0.21304576873779296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,8,16,balanced,0.24389375686645506
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,8,16,balanced,0.3046937561035156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,8,16,balanced,0.13815423965454102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,8,16,balanced,0.36532737731933596
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,8,16,balanced,0.4851532745361328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,8,16,balanced,0.033822720050811765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,8,16,balanced,0.6061606216430664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,8,16,balanced,0.033786880970001223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,8,16,balanced,0.036266241073608395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,8,16,balanced,0.8532685089111329
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,8,16,balanced,0.038691840171813964
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,8,16,balanced,0.04199039936065674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,8,16,balanced,0.05881728172302246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,8,16,balanced,0.06692863941192627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,8,16,balanced,1.133712615966797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,8,16,balanced,0.06729983806610107
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,8,16,balanced,0.06983424186706542
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,8,16,balanced,0.06928256034851074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,8,16,balanced,0.07248256206512452
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,8,16,balanced,0.07958144187927246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,8,16,balanced,0.07220223903656006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,8,16,balanced,0.07666304111480712
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,8,16,balanced,0.07918591976165772
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,8,16,balanced,0.08428799629211425
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,8,16,balanced,0.09016575813293456
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,8,16,balanced,0.09881983757019043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,8,16,balanced,0.11619839668273926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,8,16,balanced,0.13341952323913575
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,8,16,balanced,0.174201602935791
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,8,16,balanced,0.20489471435546874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,8,16,balanced,0.29591552734375004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,8,16,balanced,0.3666099166870117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,8,16,balanced,0.07621888160705567
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,8,16,balanced,0.518979835510254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,8,16,balanced,0.6703372955322265
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,4,32,power_law_1.2,0.16725120544433594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,4,32,power_law_1.2,0.2716633605957031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,4,32,power_law_1.2,0.3431257629394531
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,8,16,balanced,0.02061568021774292
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,8,16,balanced,0.031741440296173096
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,8,16,balanced,0.032131841182708745
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,8,16,balanced,0.03206144094467163
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,8,16,balanced,0.03175168037414551
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,8,16,balanced,0.02025343894958496
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,8,16,balanced,0.03182463884353638
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,8,16,balanced,0.03204735994338989
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,8,16,balanced,0.03167232036590576
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,8,16,balanced,0.021030399799346924
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,8,16,balanced,0.0323801589012146
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,8,16,balanced,0.032533760070800784
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,8,16,balanced,0.034391040802001956
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,8,16,balanced,0.034708480834960934
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,8,16,balanced,0.03699199914932251
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,8,16,balanced,0.039710719585418705
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,8,16,balanced,0.022855679988861084
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,8,16,balanced,0.04293504238128662
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,8,16,balanced,0.04763008117675781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,8,16,balanced,0.025530879497528074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,8,16,power_law_1.01,0.058388481140136714
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,8,16,power_law_1.01,0.07474944114685059
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,8,16,power_law_1.01,0.07403647899627686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,8,16,power_law_1.01,0.07673088073730469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,8,16,power_law_1.01,0.07851391792297363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,8,16,power_law_1.01,0.13207679748535156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,8,16,power_law_1.01,0.13445247650146483
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,8,16,power_law_1.01,0.1337292766571045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,8,16,power_law_1.01,0.13563008308410646
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,8,16,power_law_1.01,0.1410086441040039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,8,16,power_law_1.01,0.1452787208557129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,8,16,power_law_1.01,0.14843775749206542
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,8,16,power_law_1.01,0.15399168014526368
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,8,16,power_law_1.01,0.15164416313171386
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,8,16,power_law_1.01,0.15194751739501952
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,8,16,power_law_1.01,0.15329279899597167
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,8,16,power_law_1.01,0.1648896026611328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,8,16,power_law_1.01,0.18428031921386717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,8,16,power_law_1.01,0.20618751525878903
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,8,16,power_law_1.01,0.2507315254211426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,8,16,power_law_1.01,0.2925811195373535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,8,16,power_law_1.01,0.37985279083251955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,8,16,power_law_1.01,0.46869758605957035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,8,16,power_law_1.01,0.6510310363769531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,8,16,power_law_1.01,0.8592499542236329
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,8,16,power_law_1.01,1.2650355529785156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,8,16,balanced,0.03209856033325195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,8,16,power_law_1.01,0.03993727922439575
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,8,16,balanced,0.03201920032501221
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,8,16,power_law_1.01,0.04865407943725586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,8,16,power_law_1.01,0.05306623935699463
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,8,16,power_law_1.01,0.053858561515808104
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,8,16,power_law_1.01,1.7144998168945313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,8,16,power_law_1.01,0.0587007999420166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,8,16,power_law_1.01,0.06068096160888672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,8,16,power_law_1.01,0.0668940782546997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,8,16,power_law_1.01,0.06614528179168702
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,8,16,power_law_1.01,0.06347008228302002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,8,16,power_law_1.01,0.06767871856689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,8,16,power_law_1.01,0.07334015846252442
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,8,16,power_law_1.01,0.07492095947265624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,8,16,power_law_1.01,0.07892735958099364
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,8,16,power_law_1.01,0.07548672199249268
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,8,16,power_law_1.01,0.0770739221572876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,8,16,power_law_1.01,0.07766655921936036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,8,16,power_law_1.01,0.08668671607971191
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,8,16,power_law_1.01,0.09858816146850587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,8,16,power_law_1.01,0.10699135780334472
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,8,16,power_law_1.01,0.14771967887878418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,8,16,power_law_1.01,0.16441984176635743
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,8,16,power_law_1.01,0.23453439712524413
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,8,16,power_law_1.01,0.2805196762084961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,8,16,power_law_1.01,0.3534694290161133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,8,16,power_law_1.01,0.5763916778564453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,8,16,power_law_1.01,0.7517747497558593
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,8,16,power_law_1.01,0.024364800453186036
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,8,16,power_law_1.01,0.03217407941818237
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,8,16,power_law_1.01,0.0316761589050293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,8,16,power_law_1.01,0.03199615955352783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,8,16,power_law_1.01,0.9963302612304688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,8,16,power_law_1.01,0.029665279388427734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,8,16,power_law_1.01,0.030181119441986082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,8,16,power_law_1.01,0.030444800853729248
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,8,16,power_law_1.01,0.03083904027938843
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,8,16,power_law_1.01,0.031031041145324706
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,8,16,power_law_1.01,0.03082495927810669
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,8,16,power_law_1.01,0.031239678859710695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,8,16,power_law_1.01,0.029216001033782957
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,8,16,power_law_1.01,0.031836159229278564
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,8,16,power_law_1.01,0.032250878810882566
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,8,16,power_law_1.01,0.032956159114837645
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,8,16,power_law_1.01,0.03371648073196411
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,8,16,power_law_1.01,0.0362175989151001
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,8,16,balanced,0.03126528024673462
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,8,16,balanced,0.109367036819458
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,8,16,balanced,0.09110015869140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,8,16,balanced,0.06206208229064941
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,8,16,balanced,0.06761216163635254
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,8,16,power_law_1.01,0.056337919235229496
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,8,16,power_law_1.01,0.039519999027252194
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,8,16,power_law_1.01,0.06248447895050049
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,8,16,power_law_1.2,0.06100096225738525
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,8,16,power_law_1.01,0.04614399909973145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,8,16,power_law_1.2,0.07350143909454346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,8,16,power_law_1.2,0.07336959838867188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,8,16,power_law_1.2,0.07636096000671386
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,8,16,power_law_1.2,0.07825407981872559
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,8,16,power_law_1.2,0.1316479969024658
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,8,16,power_law_1.2,0.13462271690368652
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,8,16,power_law_1.2,0.13606143951416017
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,8,16,power_law_1.2,0.1360204792022705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,8,16,power_law_1.2,0.1401638412475586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,8,16,power_law_1.2,0.14492287635803222
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,8,16,power_law_1.2,0.15254783630371094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,8,16,power_law_1.2,0.15666175842285157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,8,16,power_law_1.2,0.1474713611602783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,8,16,power_law_1.2,0.1533580780029297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,8,16,power_law_1.2,0.1586739158630371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,8,16,power_law_1.2,0.16962560653686523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,8,16,power_law_1.2,0.19274112701416016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,8,16,power_law_1.2,0.20709888458251954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,8,16,power_law_1.2,0.2746073532104492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,8,16,power_law_1.2,0.28839296340942383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,8,16,power_law_1.2,0.4317337417602539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,8,16,power_law_1.2,0.5220108795166015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,8,16,power_law_1.2,0.7784114837646484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,8,16,power_law_1.2,0.9831795501708985
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,8,16,power_law_1.01,0.08129152297973632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,8,16,power_law_1.2,0.03988991975784302
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,8,16,power_law_1.2,0.05002624034881592
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,8,16,power_law_1.2,1.4113165283203126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,8,16,power_law_1.2,0.04990975856781006
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,8,16,power_law_1.01,0.1061081600189209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,8,16,power_law_1.2,0.05394815921783447
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,8,16,power_law_1.2,0.055095038414001464
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,8,16,power_law_1.2,0.059989762306213376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,8,16,power_law_1.2,0.06354559898376465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,8,16,power_law_1.2,0.06718592166900635
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,8,16,power_law_1.2,0.06582143783569336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,8,16,power_law_1.2,0.06550911903381348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,8,16,power_law_1.2,0.07228928089141846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,8,16,power_law_1.2,2.0900595092773435
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,8,16,power_law_1.2,0.07635968208312989
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,8,16,power_law_1.01,0.1523366355895996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,8,16,power_law_1.2,0.07261184215545655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,8,16,power_law_1.2,0.0787391996383667
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,8,16,power_law_1.2,0.0804646396636963
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,8,16,power_law_1.2,0.10175488471984864
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,8,16,power_law_1.01,0.1325209617614746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,8,16,power_law_1.2,0.11661439895629884
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,8,16,power_law_1.2,0.15135616302490235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,8,16,power_law_1.2,0.19336959838867188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,8,16,power_law_1.2,0.2505036735534668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,8,16,power_law_1.2,0.3044633674621582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,8,16,power_law_1.2,0.07712639808654785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,8,16,power_law_1.2,0.5003839874267578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,8,16,power_law_1.2,0.5776972961425781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,8,16,power_law_1.2,0.02395519971847534
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,8,16,power_law_1.2,0.0880345630645752
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,8,16,power_law_1.2,0.030760960578918455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,8,16,power_law_1.2,0.7821414184570312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,8,16,power_law_1.2,0.031027200222015383
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,8,16,power_law_1.2,0.030859520435333253
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,8,16,power_law_1.2,0.02830591917037964
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,8,16,power_law_1.2,0.029150719642639163
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,8,16,power_law_1.2,1.167293472290039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,8,16,power_law_1.2,0.029847040176391604
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,8,16,power_law_1.2,0.030178558826446534
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,8,16,power_law_1.2,0.03029759883880615
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,8,16,power_law_1.2,0.03094016075134277
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,8,16,power_law_1.2,0.031639039516448975
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,8,16,power_law_1.2,0.032117760181427
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,8,16,power_law_1.2,0.033002240657806395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,8,16,power_law_1.2,0.034213120937347415
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,8,16,power_law_1.2,0.037873919010162356
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,8,16,power_law_1.2,0.043400959968566896
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,8,16,power_law_1.2,0.030271999835968018
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,8,16,power_law_1.2,0.035345919132232666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,8,16,power_law_1.01,0.21967487335205077
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,8,16,power_law_1.01,0.31575679779052734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,8,16,power_law_1.2,0.1091430377960205
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,8,16,power_law_1.2,0.06519423961639405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,16,8,balanced,0.03481215953826904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,16,8,balanced,0.03331968069076538
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,16,8,balanced,0.036807680130004884
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,8,16,power_law_1.2,0.12804096221923827
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,8,16,power_law_1.2,0.21275903701782228
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,8,16,power_law_1.2,0.08002304077148438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,16,8,balanced,0.04201216220855713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,16,8,balanced,0.06639872074127198
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,16,8,balanced,0.09545215606689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,16,8,balanced,0.1478809642791748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,16,8,balanced,0.14855168342590333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,16,8,balanced,0.1559500789642334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,16,8,balanced,0.16078847885131836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,16,8,balanced,0.15279359817504884
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,16,8,balanced,0.1584447956085205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,16,8,balanced,0.16180992126464844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,16,8,balanced,0.1702911949157715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,16,8,balanced,0.17250431060791016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,16,8,balanced,0.1799577522277832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,16,8,balanced,0.19316736221313477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,16,8,balanced,0.2211404800415039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,8,16,power_law_1.2,0.049927678108215336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,16,8,balanced,0.3074713516235351
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,16,8,balanced,0.365588493347168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,16,8,balanced,0.4833087921142578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,16,8,balanced,0.5962623977661133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,16,8,balanced,0.8265856170654298
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,16,8,balanced,0.03422976016998291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,16,8,balanced,0.2502630424499512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,16,8,balanced,0.037500159740448
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,16,8,balanced,1.0512537384033203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,16,8,balanced,0.043587841987609864
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,16,8,balanced,0.04697728157043457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,16,8,balanced,0.05265151977539062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,16,8,balanced,0.0715776014328003
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,16,8,balanced,0.07885312080383301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,16,8,balanced,0.08576255798339844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,16,8,balanced,0.08556544303894043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,16,8,balanced,0.09306624412536621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,16,8,balanced,1.5170906066894532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,16,8,balanced,0.08528639793395995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,16,8,balanced,0.08905471801757812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,16,8,balanced,0.09236607551574708
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,16,8,balanced,0.09710592269897461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,16,8,balanced,0.09705216407775878
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,16,8,balanced,0.10286848068237305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,16,8,balanced,0.11060223579406739
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,16,8,balanced,0.12927488327026365
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,16,8,balanced,0.14538111686706542
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,16,8,balanced,0.18177152633666993
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,16,8,balanced,2.0692185974121093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,16,8,balanced,0.2161075210571289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,16,8,balanced,0.28966527938842773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,16,8,balanced,0.35514881134033205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,16,8,balanced,0.5176755142211914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,16,8,balanced,0.6521190643310547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,16,8,balanced,0.944879379272461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,16,8,balanced,1.2391948699951172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,8,16,power_law_1.2,0.23719808578491214
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,8,16,power_law_1.2,0.3634239959716797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,8,16,power_law_1.2,0.5337062454223632
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,16,8,balanced,0.023595519065856933
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,16,8,balanced,0.04529151916503906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,16,8,balanced,0.04524415969848633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,16,8,balanced,0.045841917991638184
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,16,8,balanced,0.031662080287933346
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,16,8,balanced,0.04564735889434814
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,16,8,balanced,0.04565120220184326
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,16,8,balanced,0.02399359941482544
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,16,8,balanced,0.02719360113143921
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,16,8,balanced,0.04671999931335449
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,16,8,balanced,0.04745215892791748
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,16,8,balanced,0.048636159896850585
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,16,8,balanced,0.05014783859252929
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,16,8,balanced,0.025556480884552
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,16,8,balanced,0.05158783912658691
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,16,8,balanced,0.055171837806701665
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,16,8,balanced,0.057657599449157715
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,16,8,balanced,0.06512383937835693
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,16,8,balanced,0.07099520206451417
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,16,8,balanced,0.04631936073303223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,16,8,power_law_1.01,0.03904383897781372
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,16,8,power_law_1.01,0.0622105598449707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,16,8,power_law_1.01,0.06657023906707764
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,16,8,power_law_1.01,0.07512959957122803
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,16,8,power_law_1.01,0.07777535915374756
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,16,8,power_law_1.01,0.0833523178100586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,16,8,power_law_1.01,0.1425599956512451
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,16,8,power_law_1.01,0.145982723236084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,16,8,power_law_1.01,0.15024895668029786
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,16,8,power_law_1.01,0.15270912170410156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,16,8,power_law_1.01,0.1581107234954834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,16,8,power_law_1.01,0.15673727989196778
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,16,8,power_law_1.01,0.15409536361694337
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,16,8,power_law_1.01,0.1603558349609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,16,8,power_law_1.01,0.165666561126709
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,16,8,power_law_1.01,0.17221887588500978
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,16,8,power_law_1.01,0.18363519668579104
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,16,8,power_law_1.01,0.208276481628418
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,16,8,balanced,0.04493567943572998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,16,8,power_law_1.01,0.22239360809326172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,16,8,balanced,0.04557951927185059
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,16,8,power_law_1.01,0.34144512176513675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,16,8,power_law_1.01,0.4369241714477539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,16,8,power_law_1.01,0.5538918304443359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,16,8,power_law_1.01,0.27356031417846677
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,16,8,power_law_1.01,0.7476428985595703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,16,8,power_law_1.01,0.9844556427001953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,16,8,power_law_1.01,0.040344319343566894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,16,8,power_law_1.01,1.4634597778320313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,16,8,power_law_1.01,0.05017983913421631
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,16,8,power_law_1.01,1.8784165954589844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,16,8,power_law_1.01,0.05658751964569092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,16,8,power_law_1.01,0.06565248012542725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,16,8,power_law_1.01,0.06544000148773194
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,16,8,power_law_1.01,0.04700160026550293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,16,8,power_law_1.01,0.07733888149261474
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,16,8,power_law_1.01,0.0753868818283081
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,16,8,power_law_1.01,0.08071807861328126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,16,8,power_law_1.01,0.08385024070739747
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,16,8,power_law_1.01,0.09136639595031738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,16,8,power_law_1.01,0.08719231605529784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,16,8,power_law_1.01,0.08669183731079103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,16,8,power_law_1.01,0.09020159721374513
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,16,8,power_law_1.01,0.09252351760864258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,16,8,power_law_1.01,0.0971008014678955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,16,8,power_law_1.01,0.1028108787536621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,16,8,power_law_1.01,0.11790592193603515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,16,8,power_law_1.01,0.14215295791625976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,16,8,power_law_1.01,0.1652672004699707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,16,8,power_law_1.01,0.21115648269653323
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,16,8,power_law_1.01,0.27780736923217775
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,16,8,power_law_1.01,0.34248191833496094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,16,8,power_law_1.01,0.4531135940551758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,16,8,power_law_1.01,0.6046809768676757
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,16,8,power_law_1.01,0.02717952013015747
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,16,8,power_law_1.01,0.9083148956298828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,16,8,power_law_1.01,0.03042815923690796
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,16,8,power_law_1.01,0.034355199337005614
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,16,8,power_law_1.01,0.03700608015060425
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,16,8,power_law_1.01,0.041565442085266115
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,16,8,power_law_1.01,1.225239028930664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,16,8,power_law_1.01,0.043749117851257326
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,16,8,power_law_1.01,0.04138239860534668
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,16,8,power_law_1.01,0.03970943927764893
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,16,8,power_law_1.01,0.04188799858093262
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,16,8,power_law_1.01,0.04262271881103515
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,16,8,power_law_1.01,0.04391039848327637
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,16,8,power_law_1.01,0.043424000740051275
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,16,8,power_law_1.01,0.04516736030578613
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,16,8,power_law_1.01,0.04646783828735351
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,16,8,power_law_1.01,0.04786687850952148
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,16,8,power_law_1.01,0.04915584087371826
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,16,8,power_law_1.01,0.05260672092437744
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,16,8,balanced,0.04510975837707519
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,16,8,balanced,0.1584563159942627
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,16,8,balanced,0.09838208198547363
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,16,8,balanced,0.19803007125854494
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,16,8,balanced,0.1111193561553955
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,16,8,power_law_1.01,0.06157824039459229
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,16,8,power_law_1.01,0.07013887882232665
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,16,8,power_law_1.01,0.08568703651428222
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,16,8,power_law_1.2,0.036665599346160885
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,16,8,power_law_1.2,0.05759488105773926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,16,8,power_law_1.2,0.05797760009765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,16,8,power_law_1.2,0.07269887924194336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,16,8,power_law_1.01,0.10125823974609376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,16,8,power_law_1.2,0.07762688159942627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,16,8,power_law_1.2,0.08703616142272949
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,16,8,power_law_1.2,0.14128128051757813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,16,8,power_law_1.2,0.14484095573425293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,16,8,power_law_1.2,0.14679295539855958
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,16,8,power_law_1.2,0.15209983825683593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,16,8,power_law_1.2,0.16057472229003905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,16,8,power_law_1.2,0.15761152267456055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,16,8,power_law_1.2,0.15420672416687012
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,16,8,power_law_1.2,0.16278400421142578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,16,8,power_law_1.2,0.16638591766357422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,16,8,power_law_1.2,0.17098880767822267
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,16,8,power_law_1.2,0.18665216445922853
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,16,8,power_law_1.2,0.2082508850097656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,16,8,power_law_1.2,0.23317888259887692
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,16,8,power_law_1.2,0.28279296875000004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,16,8,power_law_1.2,0.33563262939453126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,16,8,power_law_1.2,0.4474444961547851
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,16,8,power_law_1.2,0.5699750518798827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,16,8,power_law_1.2,0.8131571197509766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,16,8,power_law_1.2,1.0795481872558592
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,16,8,power_law_1.2,1.5822848510742187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,16,8,power_law_1.2,0.039847679138183593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,16,8,power_law_1.2,0.04690944194793702
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,16,8,power_law_1.2,0.0485964822769165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,16,8,power_law_1.2,2.0801600646972656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,16,8,power_law_1.2,0.057057280540466306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,16,8,power_law_1.2,0.06614143848419189
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,16,8,power_law_1.2,0.06819968223571778
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,16,8,power_law_1.01,0.13213695526123045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,16,8,power_law_1.2,0.07581183910369874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,16,8,power_law_1.2,0.08186367988586427
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,16,8,power_law_1.01,0.2169856071472168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,16,8,power_law_1.2,0.0840652847290039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,16,8,power_law_1.2,0.08962431907653809
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,16,8,power_law_1.2,0.08898303985595704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,16,8,power_law_1.2,0.0848243236541748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,16,8,power_law_1.2,0.09170047760009765
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,16,8,power_law_1.01,0.1506726360321045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,16,8,power_law_1.2,0.09430784225463867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,16,8,power_law_1.2,0.10429311752319337
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,16,8,power_law_1.2,0.1246617603302002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,16,8,power_law_1.2,0.1447385597229004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,16,8,power_law_1.2,0.07545983791351318
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,16,8,power_law_1.2,0.17947263717651368
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,16,8,power_law_1.2,0.20396928787231444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,16,8,power_law_1.2,0.2881356811523438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,16,8,power_law_1.2,0.3551385498046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,16,8,power_law_1.2,0.5156825637817383
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,16,8,power_law_1.01,0.3096908760070801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,16,8,power_law_1.2,0.09166208267211914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,16,8,power_law_1.2,0.026776320934295654
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,16,8,power_law_1.2,0.03145983934402466
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,16,8,power_law_1.2,0.6800089263916016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,16,8,power_law_1.2,0.031432960033416744
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,16,8,power_law_1.2,0.03607167959213257
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,16,8,power_law_1.2,0.04124159812927246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,16,8,power_law_1.2,0.950298843383789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,16,8,power_law_1.2,0.041520638465881346
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,16,8,power_law_1.2,0.04165760040283203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,16,8,power_law_1.2,0.04195583820343017
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,16,8,power_law_1.2,0.0436902379989624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,16,8,power_law_1.2,0.04372096061706543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,16,8,power_law_1.2,0.04519167900085449
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,16,8,power_law_1.2,0.04914815902709961
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,16,8,power_law_1.2,0.05121664047241211
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,16,8,power_law_1.2,0.058928642272949215
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,16,8,power_law_1.2,0.041855998039245605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,16,8,power_law_1.2,1.2519769287109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,16,8,power_law_1.2,0.04674431800842285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,16,8,power_law_1.2,0.0365068793296814
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,16,8,power_law_1.01,0.4328972625732422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,16,8,power_law_1.01,0.5529497528076173
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,16,8,power_law_1.2,0.4910502243041992
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,16,8,power_law_1.2,0.07002495765686036
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,16,8,power_law_1.2,0.10082176208496094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,16,8,power_law_1.2,0.12488320350646973
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,32,4,balanced,0.030905599594116213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,32,4,balanced,0.037698559761047364
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,32,4,balanced,0.03379584074020386
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,16,8,power_law_1.2,0.07655424118041992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,32,4,balanced,0.05270400047302246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,32,4,balanced,0.07521664142608643
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,32,4,balanced,0.09228287696838379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,32,4,balanced,0.09511167526245116
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,16,8,power_law_1.2,0.1783283233642578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,32,4,balanced,0.16527231216430666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,32,4,balanced,0.16852352142333984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,32,4,balanced,0.17604736328124998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,32,4,balanced,0.17310592651367188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,32,4,balanced,0.18132608413696288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,32,4,balanced,0.1875174331665039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,32,4,balanced,0.19141376495361329
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,32,4,balanced,0.19748607635498047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,32,4,balanced,0.2113996887207031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,32,4,balanced,0.23968383789062503
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,32,4,balanced,0.26835712432861325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,32,4,balanced,0.09406847953796386
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,32,4,balanced,0.3233548736572266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,16,8,power_law_1.2,0.2303436851501465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,32,4,balanced,0.3841651153564453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,32,4,balanced,0.49991294860839847
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,32,4,balanced,0.6113049697875976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,32,4,balanced,0.03381376028060913
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,32,4,balanced,0.8362329864501954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,32,4,balanced,0.03788671970367431
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,32,4,balanced,1.065945587158203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,32,4,balanced,1.5260263061523438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,32,4,balanced,0.06706431865692139
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,32,4,balanced,0.09212160110473633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,32,4,balanced,0.0924287986755371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,32,4,balanced,0.058329601287841794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,32,4,balanced,2.087394561767578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,32,4,balanced,0.09433343887329101
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,32,4,balanced,0.11073151588439942
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,32,4,balanced,0.1119923210144043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,32,4,balanced,0.11469568252563475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,32,4,balanced,0.11882880210876465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,32,4,balanced,0.121015043258667
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,32,4,balanced,0.12702207565307616
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,32,4,balanced,0.1264691162109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,32,4,balanced,0.13147775650024413
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,32,4,balanced,0.1395404815673828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,32,4,balanced,0.15774592399597168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,32,4,balanced,0.17545600891113283
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,32,4,balanced,0.21007999420166015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,32,4,balanced,0.0453004789352417
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,32,4,balanced,0.24371711730957032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,32,4,balanced,0.3125004768371582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,32,4,balanced,0.37847553253173827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,32,4,balanced,0.5573286437988282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,32,4,balanced,0.6899571228027344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,32,4,balanced,1.0242508697509767
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,32,4,balanced,1.3434355163574219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,16,8,power_law_1.2,0.33467391967773436
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,16,8,power_law_1.2,0.6339020919799805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,16,8,power_law_1.2,0.7569062042236328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,32,4,balanced,0.03338752031326294
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,32,4,balanced,0.06628096103668213
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,32,4,balanced,0.047744002342224125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,32,4,balanced,0.06603263854980469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,32,4,balanced,0.0666099214553833
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,32,4,balanced,0.031756799221038814
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,32,4,balanced,0.06669439792633056
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,32,4,balanced,0.06697728157043456
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,32,4,balanced,0.06863232135772705
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,32,4,balanced,0.06840447902679443
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,32,4,balanced,0.07007232189178467
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,32,4,balanced,0.07106688022613525
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,32,4,balanced,0.07460351943969726
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,32,4,balanced,0.07721216201782226
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,32,4,balanced,0.08392704010009766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,32,4,balanced,0.03093247890472412
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,32,4,balanced,0.038492159843444826
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,32,4,balanced,0.10046208381652832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,32,4,balanced,0.11250816345214845
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,32,4,balanced,0.08892928123474121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,32,4,power_law_1.01,0.03458559989929199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,32,4,power_law_1.01,0.038632960319519044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,32,4,power_law_1.01,0.06374527931213378
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,32,4,power_law_1.01,0.07169023990631104
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,32,4,power_law_1.01,0.0751091194152832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,32,4,power_law_1.01,0.08211711883544923
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,32,4,balanced,0.06567552089691162
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,32,4,power_law_1.01,0.08886143684387207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,32,4,power_law_1.01,0.0908351993560791
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,32,4,power_law_1.01,0.1628505516052246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,32,4,power_law_1.01,0.16448511123657225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,32,4,power_law_1.01,0.16860031127929687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,32,4,power_law_1.01,0.17390207290649412
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,32,4,power_law_1.01,0.17796607971191408
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,32,4,power_law_1.01,0.18515199661254883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,32,4,power_law_1.01,0.18900863647460936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,32,4,power_law_1.01,0.19651199340820313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,32,4,power_law_1.01,0.2105446434020996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,32,4,power_law_1.01,0.23949695587158204
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,32,4,balanced,0.06566400051116943
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,32,4,power_law_1.01,0.269171199798584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,32,4,power_law_1.01,0.3290572738647461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,32,4,power_law_1.01,0.3762623977661133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,32,4,power_law_1.01,0.49348480224609376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,32,4,power_law_1.01,0.6146752166748046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,32,4,power_law_1.01,0.8036326599121093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,32,4,power_law_1.01,1.0706034851074218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,32,4,power_law_1.01,0.03790080070495606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,32,4,power_law_1.01,0.045670399665832515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,32,4,power_law_1.01,1.587667236328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,32,4,power_law_1.01,0.049583358764648436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,32,4,power_law_1.01,0.06861695766448975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,32,4,power_law_1.01,0.07079040050506592
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,32,4,power_law_1.01,0.08242303848266601
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,32,4,power_law_1.01,0.08261247634887695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,32,4,power_law_1.01,2.0637863159179686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,32,4,power_law_1.01,0.08525952339172363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,32,4,power_law_1.01,0.09641983985900879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,32,4,power_law_1.01,0.10122112274169921
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,32,4,power_law_1.01,0.10269311904907226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,32,4,power_law_1.01,0.107892484664917
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,32,4,power_law_1.01,0.11251456260681152
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,32,4,power_law_1.01,0.12051327705383301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,32,4,power_law_1.01,0.12096896171569824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,32,4,power_law_1.01,0.1261465644836426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,32,4,power_law_1.01,0.13588224411010744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,32,4,power_law_1.01,0.15376511573791504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,32,4,power_law_1.01,0.1739673614501953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,32,4,power_law_1.01,0.2185766410827637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,32,4,power_law_1.01,0.2551359939575195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,32,4,power_law_1.01,0.33582080841064454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,32,4,power_law_1.01,0.4111360168457031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,32,4,power_law_1.01,0.5764364624023438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,32,4,power_law_1.01,0.7432307434082032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,32,4,power_law_1.01,1.0879808044433594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,32,4,power_law_1.01,0.03413503885269165
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,32,4,power_law_1.01,0.03745023965835571
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,32,4,power_law_1.01,0.04273536205291748
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,32,4,power_law_1.01,0.045076479911804204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,32,4,power_law_1.01,1.4285874938964844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,32,4,power_law_1.01,0.0516428804397583
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,32,4,power_law_1.01,0.05267583847045898
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,32,4,power_law_1.01,0.06261631965637207
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,32,4,power_law_1.01,0.06219520092010498
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,32,4,power_law_1.01,0.06066815853118897
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,32,4,power_law_1.01,0.06171904087066651
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,32,4,power_law_1.01,0.06566271781921387
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,32,4,power_law_1.01,0.06777088165283203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,32,4,power_law_1.01,0.05773952007293701
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,32,4,power_law_1.01,0.07071104049682617
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,32,4,power_law_1.01,0.07390463829040528
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,32,4,power_law_1.01,0.077358078956604
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,32,4,power_law_1.01,0.06340479850769043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,32,4,balanced,0.0664844799041748
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,32,4,balanced,0.2813068771362305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,32,4,balanced,0.17105152130126952
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,32,4,balanced,0.19604608535766602
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,32,4,balanced,0.36315902709960934
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,32,4,power_law_1.01,0.10463104248046876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,32,4,power_law_1.01,0.09588224411010741
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,32,4,power_law_1.01,0.1287065601348877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,32,4,power_law_1.2,0.03496959924697876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,32,4,power_law_1.2,0.037808640003204344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,32,4,power_law_1.2,0.06169983863830566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,32,4,power_law_1.2,0.07040639877319335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,32,4,power_law_1.2,0.0739302396774292
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,32,4,power_law_1.2,0.07991936206817626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,32,4,power_law_1.2,0.08873472213745118
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,32,4,power_law_1.2,0.0908620834350586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,32,4,power_law_1.2,0.16289535522460938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,32,4,power_law_1.01,0.16286975860595704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,32,4,power_law_1.2,0.1644812774658203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,32,4,power_law_1.2,0.16785919189453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,32,4,power_law_1.2,0.17495807647705078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,32,4,power_law_1.2,0.17887104034423829
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,32,4,power_law_1.2,0.18726272583007814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,32,4,power_law_1.2,0.18936832427978517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,32,4,power_law_1.2,0.19675264358520508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,32,4,power_law_1.2,0.21156991958618163
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,32,4,power_law_1.2,0.24108543395996093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,32,4,power_law_1.2,0.27089920043945315
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,32,4,power_law_1.2,0.32874622344970705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,32,4,power_law_1.2,0.3840857696533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,32,4,power_law_1.2,0.5096704101562499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,32,4,power_law_1.2,0.6152972793579101
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,32,4,power_law_1.2,0.8531686401367187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,32,4,power_law_1.2,1.1139110565185546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,32,4,power_law_1.2,1.634984893798828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,32,4,power_law_1.2,0.03831295967102051
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,32,4,power_law_1.2,0.04609663963317871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,32,4,power_law_1.2,0.04750336170196533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,32,4,power_law_1.2,2.1173298645019534
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,32,4,power_law_1.2,0.06312191963195801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,32,4,power_law_1.2,0.07640831947326661
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,32,4,power_law_1.2,0.07646592140197754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,32,4,power_law_1.2,0.08032256126403808
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,32,4,power_law_1.2,0.08304639816284179
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,32,4,power_law_1.2,0.09674752235412598
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,32,4,power_law_1.2,0.0985536003112793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,32,4,power_law_1.2,0.1028287982940674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,32,4,power_law_1.2,0.10898943901062012
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,32,4,power_law_1.2,0.11488639831542968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,32,4,power_law_1.01,0.21682432174682614
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,32,4,power_law_1.2,0.12132224082946777
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,32,4,power_law_1.2,0.1212992000579834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,32,4,power_law_1.2,0.12816767692565917
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,32,4,power_law_1.2,0.1366374397277832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,32,4,power_law_1.2,0.1565542411804199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,32,4,power_law_1.2,0.1788096046447754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,32,4,power_law_1.2,0.22668415069580078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,32,4,power_law_1.2,0.2572313690185547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,32,4,power_law_1.01,0.4053311920166015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,32,4,power_law_1.01,0.28084096908569334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,32,4,power_law_1.2,0.33751678466796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,32,4,power_law_1.2,0.43173248291015626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,32,4,power_law_1.2,0.03419903993606567
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,32,4,power_law_1.01,0.5127257537841797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,32,4,power_law_1.2,0.03731967926025391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,32,4,power_law_1.2,0.7759500885009766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,32,4,power_law_1.2,0.04119552135467529
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,32,4,power_law_1.2,0.04384768009185791
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,32,4,power_law_1.2,0.047713279724121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,32,4,power_law_1.2,0.04943615913391113
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,32,4,power_law_1.2,0.060140800476074216
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,32,4,power_law_1.2,0.06252799987792969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,32,4,power_law_1.2,0.05912960052490235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,32,4,power_law_1.2,1.4509657287597657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,32,4,power_law_1.2,0.06017024040222167
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,32,4,power_law_1.2,0.5932979202270507
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,32,4,power_law_1.2,0.06244863986968994
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,32,4,power_law_1.2,0.06410496234893799
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,32,4,power_law_1.2,0.06831232070922852
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,32,4,power_law_1.2,0.07005695819854736
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,32,4,power_law_1.2,0.07489535808563233
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,32,4,power_law_1.2,0.0785152006149292
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,32,4,power_law_1.2,1.1487897491455077
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,32,4,power_law_1.01,0.769085464477539
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,32,4,power_law_1.01,0.995700454711914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,32,4,power_law_1.2,0.08678784370422363
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,32,4,power_law_1.2,0.12911104202270507
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,32,4,power_law_1.2,0.10525312423706054
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,32,4,power_law_1.2,0.21505664825439452
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,32,4,power_law_1.2,0.16965248107910155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,128,balanced,0.03796864032745361
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,128,balanced,0.040378880500793454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,128,balanced,0.037914879322052
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,128,balanced,0.04044288158416748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,128,balanced,0.038227200508117676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,128,balanced,0.07018752098083496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,128,balanced,0.12370816230773926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,128,balanced,0.12314111709594726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,128,balanced,0.12395135879516601
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,128,balanced,0.12372480392456056
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,128,balanced,0.12473983764648437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,128,balanced,0.12530176162719725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,128,balanced,0.12648960113525393
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,128,balanced,0.12683903694152832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,128,balanced,0.1294489574432373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,128,balanced,0.13046143531799317
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,128,balanced,0.13232768058776856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,128,balanced,0.13548288345336915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,128,balanced,0.14200575828552248
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,32,4,power_law_1.2,0.2811673545837402
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,128,balanced,0.14461440086364746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,128,balanced,0.15622912406921388
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,128,balanced,0.16652799606323243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,128,balanced,0.18437503814697265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,128,balanced,0.20766719818115234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,128,balanced,0.2803264045715332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,128,balanced,0.34362239837646485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,128,balanced,0.03580032110214233
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,128,balanced,0.15110272407531739
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,128,balanced,0.03576832056045533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,128,balanced,0.035861759185791015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,128,balanced,0.038736639022827146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,128,balanced,0.035856640338897704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,128,balanced,0.03702399969100952
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,128,balanced,0.04538879871368408
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,128,balanced,0.04530303955078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,128,balanced,0.04608895778656006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,128,balanced,0.046050558090209956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,128,balanced,0.04659711837768555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,128,balanced,0.04765056133270264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,128,balanced,0.04703487873077393
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,128,balanced,0.0484991979598999
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,128,balanced,0.050173439979553223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,128,balanced,0.0488921594619751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,128,balanced,0.05183616161346436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,128,balanced,0.052007679939270024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,128,balanced,0.05349631786346436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,128,balanced,0.06632192134857177
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,128,balanced,0.05888256072998047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,128,balanced,0.06953472137451172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,128,balanced,0.0768230390548706
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,128,balanced,0.08818816184997559
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,128,balanced,0.09951359748840331
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,128,balanced,0.12554112434387207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,128,balanced,0.1570470428466797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,32,4,power_law_1.2,0.40786560058593746
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,32,4,power_law_1.2,0.555752944946289
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,32,4,power_law_1.2,0.6980838775634766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,32,4,power_law_1.2,1.0831168365478514
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,32,4,power_law_1.2,1.662488250732422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,1,128,balanced,0.01941632032394409
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,1,128,balanced,0.01955072045326233
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,1,128,balanced,0.01988479971885681
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,1,128,balanced,0.019788800477981566
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,1,128,balanced,0.019791359901428225
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,1,128,balanced,0.019362560510635375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,1,128,balanced,0.019882240295410157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,1,128,balanced,0.01989248037338257
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,1,128,balanced,0.019866880178451538
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,1,128,balanced,0.02021631956100464
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,1,128,balanced,0.019819519519805907
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,1,128,balanced,0.019793920516967774
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,1,128,balanced,0.02014336109161377
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,1,128,balanced,0.020288000106811525
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,1,128,balanced,0.0202892804145813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,1,128,balanced,0.02071552038192749
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,1,128,balanced,0.02105855941772461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,1,128,balanced,0.021071360111236573
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,1,128,balanced,0.019383039474487305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,1,128,balanced,0.019134720563888548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,128,power_law_1.01,0.0700057601928711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,128,power_law_1.01,0.06986623764038086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,128,power_law_1.01,0.07033984184265137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,128,power_law_1.01,0.06987391948699952
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,128,power_law_1.01,0.1232038402557373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,128,power_law_1.01,0.12440832138061524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,128,power_law_1.01,0.12456831932067872
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,128,power_law_1.01,0.1258176040649414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,128,power_law_1.01,0.12804863929748536
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,128,power_law_1.01,0.12874239921569824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,128,power_law_1.01,0.13259903907775877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,128,power_law_1.01,0.1332147216796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,128,power_law_1.01,0.13533056259155274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,128,power_law_1.01,0.13974783897399903
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,128,power_law_1.01,0.14784128189086915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,128,power_law_1.01,0.15158656120300293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,128,power_law_1.01,0.15078399658203126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,128,power_law_1.01,0.16241920471191407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,128,power_law_1.01,0.17635839462280273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,128,power_law_1.01,0.2109542465209961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,128,power_law_1.01,0.24380800247192386
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,128,power_law_1.01,0.3354470443725586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,128,power_law_1.01,0.4584115219116211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,128,power_law_1.01,0.6270323181152344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,128,power_law_1.01,0.8766207885742187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,1,128,balanced,0.019763200283050536
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,1,128,balanced,0.0201907205581665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,128,power_law_1.01,0.036962559223175054
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,128,power_law_1.01,1.4549388122558593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,128,power_law_1.01,0.03786240100860595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,128,power_law_1.01,0.03705087900161743
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,128,power_law_1.01,0.042547202110290526
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,128,power_law_1.01,1.8134259033203126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,128,power_law_1.01,0.04584320068359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,128,power_law_1.01,0.0482700777053833
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,128,power_law_1.01,0.048348159790039064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,128,power_law_1.01,0.04692992210388184
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,1,128,balanced,0.020696320533752442
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,128,power_law_1.01,0.05261568069458008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,128,power_law_1.01,0.0511411190032959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,128,power_law_1.01,0.04938496112823486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,128,power_law_1.01,0.05230591773986817
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,128,power_law_1.01,0.05289087772369385
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,128,power_law_1.01,0.05374591827392579
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,128,power_law_1.01,0.05670527935028076
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,128,power_law_1.01,0.06753791809082031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,128,power_law_1.01,0.07039487838745118
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,128,power_law_1.01,0.08082176208496093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,128,power_law_1.01,0.09775103569030762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,128,power_law_1.01,0.11348223686218262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,128,power_law_1.01,0.16784000396728516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,128,power_law_1.01,0.19978368759155274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,128,power_law_1.01,0.24546560287475586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,128,power_law_1.01,0.38431873321533205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,128,power_law_1.01,0.06342271804809571
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,1,128,power_law_1.01,0.019520000219345093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,128,power_law_1.01,0.6058201599121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,1,128,power_law_1.01,0.019829759597778322
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,1,128,power_law_1.01,0.019816960096359255
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,1,128,power_law_1.01,0.019811840057373048
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,1,128,power_law_1.01,0.019980800151824952
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,128,power_law_1.01,0.8334540557861327
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,1,128,power_law_1.01,0.019805439710617066
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,1,128,power_law_1.01,0.019824639558792115
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,1,128,power_law_1.01,0.019930880069732666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,1,128,power_law_1.01,0.020240640640258788
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,1,128,power_law_1.01,0.02020479917526245
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,1,128,power_law_1.01,0.019856640100479127
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,1,128,power_law_1.01,0.020419840812683103
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,1,128,power_law_1.01,0.01990399956703186
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,1,128,power_law_1.01,0.02065279960632324
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,1,128,power_law_1.01,0.020904960632324218
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,1,128,power_law_1.01,0.019957760572433474
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,1,128,power_law_1.01,0.02108288049697876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,1,128,balanced,0.026932480335235598
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,1,128,balanced,0.0363097596168518
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,1,128,balanced,0.02605567932128906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,1,128,balanced,0.0429260778427124
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,1,128,power_law_1.01,0.023724799156188962
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,1,128,power_law_1.01,0.023585278987884522
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,1,128,power_law_1.01,0.03279232025146485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,128,power_law_1.2,0.07026048183441162
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,1,128,power_law_1.01,0.03624191999435425
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,128,power_law_1.2,0.07001471996307373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,128,power_law_1.2,0.0704089593887329
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,128,power_law_1.2,0.12111231803894043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,128,power_law_1.2,0.12408063888549806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,128,power_law_1.2,0.12430080413818359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,128,power_law_1.2,0.12820480346679688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,128,power_law_1.2,0.12557312011718752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,128,power_law_1.2,0.13219200134277345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,128,power_law_1.2,0.13001855850219726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,128,power_law_1.2,0.13452544212341308
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,128,power_law_1.2,0.13463040351867678
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,128,power_law_1.2,0.13805055618286133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,128,power_law_1.2,0.14508416175842284
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,128,power_law_1.2,0.15119872093200684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,128,power_law_1.2,0.1513267230987549
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,128,power_law_1.2,0.15670144081115722
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,128,power_law_1.2,0.17442047119140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,1,128,power_law_1.01,0.05017471790313721
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,128,power_law_1.2,0.20829824447631834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,128,power_law_1.2,0.2760268783569336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,128,power_law_1.2,0.33783294677734377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,128,power_law_1.2,0.5859596633911133
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,1,128,power_law_1.01,0.05969535827636718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,128,power_law_1.2,0.6009638214111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,1,128,power_law_1.01,0.0697279977798462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,128,power_law_1.2,1.0116223907470703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,128,power_law_1.2,0.03636480093002319
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,128,power_law_1.2,1.3585638427734374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,128,power_law_1.2,0.037079041004180906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,128,power_law_1.2,0.03776384115219116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,128,power_law_1.2,0.04550271987915039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,128,power_law_1.2,0.04688767910003662
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,128,power_law_1.2,0.04690432071685791
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,1,128,power_law_1.01,0.09669631958007813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,128,power_law_1.2,0.04929791927337647
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,128,power_law_1.2,0.04815360069274902
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,128,power_law_1.2,0.052774400711059564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,128,power_law_1.2,0.04960896015167236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,128,power_law_1.2,0.05369215965270996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,128,power_law_1.2,2.46495361328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,128,power_law_1.2,0.05288832187652588
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,128,power_law_1.2,0.06067840099334717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,128,power_law_1.2,0.06776959896087646
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,128,power_law_1.2,0.06625152111053467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,128,power_law_1.2,0.07146368026733399
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,128,power_law_1.2,0.07891456127166749
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,128,power_law_1.2,0.08822527885437012
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,128,power_law_1.2,0.1141312026977539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,128,power_law_1.2,2.8621542358398435
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,128,power_law_1.2,0.1460966396331787
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,128,power_law_1.2,0.2240345573425293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,128,power_law_1.2,0.26732160568237306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,128,power_law_1.2,0.4123737716674805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,1,128,power_law_1.2,0.0196070396900177
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,128,power_law_1.2,0.5762841415405273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,128,power_law_1.2,0.05106560230255127
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,1,128,power_law_1.2,0.019819519519805907
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,1,128,power_law_1.2,0.019846400022506712
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,1,128,power_law_1.2,0.019711999893188475
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,1,128,power_law_1.2,0.019799040555953978
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,1,128,power_law_1.2,0.019809279441833496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,128,power_law_1.2,0.9274406433105469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,1,128,power_law_1.2,0.019973119497299192
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,1,128,power_law_1.2,0.019792640209198
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,1,128,power_law_1.2,0.01980031967163086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,1,128,power_law_1.2,0.020281600952148437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,1,128,power_law_1.2,0.02009471893310547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,1,128,power_law_1.2,0.020273919105529788
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,1,128,power_law_1.2,0.019911680221557617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,128,power_law_1.2,1.2303501129150392
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,1,128,power_law_1.2,0.020710399150848387
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,1,128,power_law_1.2,0.02130431890487671
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,1,128,power_law_1.2,0.01982079982757568
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,1,128,power_law_1.01,0.14603903770446777
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,1,128,power_law_1.01,0.18884096145629883
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,1,128,power_law_1.2,0.022753279209136962
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,1,128,power_law_1.2,0.02922624111175537
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,1,128,power_law_1.2,0.037634561061859126
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,1,128,power_law_1.2,0.07064064025878906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,2,64,balanced,0.03213696002960205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,2,64,balanced,0.03172863960266113
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,1,128,power_law_1.2,0.02271359920501709
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,2,64,balanced,0.0315827202796936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,2,64,balanced,0.032104959487915044
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,1,128,power_law_1.2,0.043681278228759765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,2,64,balanced,0.06533120155334474
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,2,64,balanced,0.11367039680480957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,2,64,balanced,0.11341183662414552
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,2,64,balanced,0.11529088020324707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,2,64,balanced,0.11527039527893065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,2,64,balanced,0.11617152214050293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,2,64,balanced,0.1194431972503662
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,2,64,balanced,0.11834879875183106
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,2,64,balanced,0.12274304389953614
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,2,64,balanced,0.12290687561035156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,2,64,balanced,0.12445311546325684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,2,64,balanced,0.034586880207061764
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,2,64,balanced,0.12831232070922852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,2,64,balanced,0.1351731204986572
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,2,64,balanced,0.1316211223602295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,2,64,balanced,0.14189184188842774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,2,64,balanced,0.14819968223571778
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,2,64,balanced,0.1644313621520996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,2,64,balanced,0.17922687530517578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,2,64,balanced,0.21653120040893556
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,2,64,balanced,0.2524812889099121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,2,64,balanced,0.033786880970001223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,2,64,balanced,0.32891006469726564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,2,64,balanced,0.03337599992752075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,2,64,balanced,0.035774719715118405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,2,64,balanced,0.033720319271087644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,2,64,balanced,0.03332864046096802
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,2,64,balanced,0.03703039884567261
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,2,64,balanced,0.04286208152770996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,2,64,balanced,0.044526081085205074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,2,64,balanced,0.04358016014099121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,2,64,balanced,0.04607744216918945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,2,64,balanced,0.044248318672180174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,2,64,balanced,0.05137919902801513
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,2,64,balanced,0.046443519592285154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,2,64,balanced,0.4318886566162109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,2,64,balanced,0.05588607788085938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,2,64,balanced,0.048325119018554685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,2,64,balanced,0.049049601554870606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,2,64,balanced,0.05221888065338135
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,2,64,balanced,0.057991681098937986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,2,64,balanced,0.054329600334167484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,2,64,balanced,0.06330239772796631
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,2,64,balanced,0.06974207878112793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,2,64,balanced,0.07636096000671386
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,2,64,balanced,0.08625408172607421
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,2,64,balanced,0.10956416130065919
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,2,64,balanced,0.12923647880554198
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,2,64,balanced,0.17366655349731447
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,2,64,balanced,0.21790847778320313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,1,128,power_law_1.2,0.07698431968688965
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,1,128,power_law_1.2,0.11786111831665039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,1,128,power_law_1.2,0.1505945587158203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,1,128,power_law_1.2,0.3362675094604492
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,1,128,power_law_1.2,0.23141504287719727
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,2,64,balanced,0.019384319782257083
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,2,64,balanced,0.021459200382232667
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,2,64,balanced,0.02172287940979004
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,2,64,balanced,0.02147455930709839
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,2,64,balanced,0.022288639545440674
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,2,64,balanced,0.01942271947860718
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,2,64,balanced,0.021870079040527343
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,2,64,balanced,0.022332160472869872
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,2,64,balanced,0.02229759931564331
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,2,64,balanced,0.022298879623413086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,2,64,balanced,0.022302720546722412
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,2,64,balanced,0.02233344078063965
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,2,64,balanced,0.022305281162261964
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,2,64,balanced,0.022731521129608155
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,2,64,balanced,0.022787840366363527
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,2,64,balanced,0.022256639003753662
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,2,64,balanced,0.024353280067443847
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,2,64,balanced,0.01898751974105835
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,2,64,balanced,0.018969600200653074
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,2,64,balanced,0.01957759976387024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,2,64,power_law_1.01,0.06472447872161866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,2,64,power_law_1.01,0.06510208129882813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,2,64,power_law_1.01,0.0656115198135376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,2,64,power_law_1.01,0.06610176086425781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,2,64,power_law_1.01,0.11213952064514161
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,2,64,power_law_1.01,0.11456383705139159
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,2,64,power_law_1.01,0.11558400154113771
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,2,64,power_law_1.01,0.11778688430786133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,2,64,power_law_1.01,0.11889151573181152
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,2,64,power_law_1.01,0.1195315170288086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,2,64,power_law_1.01,0.12159872055053711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,2,64,power_law_1.01,0.12604672431945801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,2,64,power_law_1.01,0.13030400276184081
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,2,64,power_law_1.01,0.13595520019531251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,2,64,power_law_1.01,0.13563263893127442
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,2,64,power_law_1.01,0.14087295532226562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,2,64,power_law_1.01,0.14179840087890624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,2,64,power_law_1.01,0.15652864456176757
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,2,64,power_law_1.01,0.16607744216918946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,2,64,power_law_1.01,0.20344703674316406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,2,64,power_law_1.01,0.24514303207397461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,2,64,power_law_1.01,0.2965657615661621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,2,64,power_law_1.01,0.490041618347168
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,2,64,balanced,0.0214847993850708
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,2,64,power_law_1.01,0.6373811340332031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,2,64,power_law_1.01,0.8552166748046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,2,64,balanced,0.02184704065322876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,2,64,power_law_1.01,1.1118195343017578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,2,64,power_law_1.01,0.03544960021972656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,2,64,balanced,0.02192512035369873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,2,64,power_law_1.01,0.035873279571533204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,2,64,power_law_1.01,0.03602432012557984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,2,64,power_law_1.01,0.03583872079849243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,2,64,power_law_1.01,0.04325632095336914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,2,64,power_law_1.01,0.04348031997680664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,2,64,power_law_1.01,0.04408959865570068
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,2,64,power_law_1.01,1.4405952453613282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,2,64,power_law_1.01,0.04695551872253418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,2,64,power_law_1.01,0.0463321590423584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,2,64,power_law_1.01,0.04865151882171631
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,2,64,power_law_1.01,0.049456639289855955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,2,64,power_law_1.01,0.05072512149810791
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,2,64,power_law_1.01,0.05369344234466553
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,2,64,power_law_1.01,0.059362559318542486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,2,64,power_law_1.01,0.060014081001281736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,2,64,power_law_1.01,0.060921602249145515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,2,64,power_law_1.01,0.06474112033843994
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,2,64,power_law_1.01,0.07219967842102051
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,2,64,power_law_1.01,0.08063360214233398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,2,64,power_law_1.01,0.09592703819274903
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,2,64,power_law_1.01,0.10965760231018067
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,2,64,power_law_1.01,0.1541324806213379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,2,64,power_law_1.01,0.2316595268249512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,2,64,power_law_1.01,0.30204160690307613
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,2,64,power_law_1.01,0.3746752166748047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,2,64,power_law_1.01,0.021827840805053712
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,2,64,power_law_1.01,0.5267788696289062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,2,64,power_law_1.01,0.021868801116943358
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,2,64,power_law_1.01,0.02158207893371582
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,2,64,power_law_1.01,0.021948161125183104
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,2,64,power_law_1.01,0.021036798954010012
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,2,64,power_law_1.01,0.021057279109954835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,2,64,power_law_1.01,0.7795661163330079
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,2,64,power_law_1.01,0.021756160259246825
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,2,64,power_law_1.01,0.02187648057937622
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,2,64,power_law_1.01,0.021908481121063233
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,2,64,power_law_1.01,0.02191871881484985
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,2,64,power_law_1.01,0.02183295965194702
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,2,64,power_law_1.01,0.022069759368896484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,2,64,power_law_1.01,0.021893119812011717
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,2,64,power_law_1.01,0.021863679885864257
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,2,64,power_law_1.01,0.021868801116943358
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,2,64,power_law_1.01,0.02223360061645508
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,2,64,power_law_1.01,0.022371840476989747
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,2,64,balanced,0.030105600357055663
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,2,64,balanced,0.029254400730133058
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,2,64,balanced,0.04322559833526611
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,2,64,balanced,0.052043519020080566
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,2,64,power_law_1.01,0.025913600921630857
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,2,64,power_law_1.01,0.024360959529876706
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,2,64,power_law_1.01,0.029937920570373533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,2,64,power_law_1.2,0.06524032115936279
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,2,64,power_law_1.2,0.06508416175842285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,2,64,power_law_1.01,0.03943039894104004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,2,64,power_law_1.2,0.0654207992553711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,2,64,power_law_1.2,0.06554495811462402
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,2,64,power_law_1.2,0.11315072059631348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,2,64,power_law_1.2,0.11682304382324218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,2,64,power_law_1.2,0.11699456214904784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,2,64,power_law_1.2,0.12174079895019532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,2,64,power_law_1.2,0.12022784233093262
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,2,64,power_law_1.01,0.04816639900207519
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,2,64,power_law_1.2,0.12049663543701172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,2,64,power_law_1.2,0.12641152381896972
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,2,64,power_law_1.2,0.1259404754638672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,2,64,power_law_1.2,0.1341196823120117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,2,64,power_law_1.2,0.13673215866088867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,2,64,power_law_1.2,0.13822336196899415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,2,64,power_law_1.2,0.13729408264160156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,2,64,power_law_1.01,0.06373248100280762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,2,64,power_law_1.2,0.14823040008544922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,2,64,power_law_1.2,0.16182783126831055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,2,64,power_law_1.2,0.18640256881713868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,2,64,power_law_1.2,0.22774656295776366
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,2,64,power_law_1.01,0.07333375930786132
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,2,64,power_law_1.2,0.3091084861755371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,2,64,power_law_1.2,0.4565836715698242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,2,64,power_law_1.2,0.5307916641235352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,2,64,power_law_1.2,0.7279692840576172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,2,64,power_law_1.2,0.9988441467285156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,2,64,power_law_1.2,0.03599616050720215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,2,64,power_law_1.2,0.03587584018707275
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,2,64,power_law_1.2,0.03736191987991333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,2,64,power_law_1.2,0.043368959426879884
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,2,64,power_law_1.2,0.04379263877868652
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,2,64,power_law_1.2,1.7127719116210938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,2,64,power_law_1.2,0.036172800064086914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,2,64,power_law_1.2,0.04634111881256103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,2,64,power_law_1.2,0.048465919494628903
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,2,64,power_law_1.2,0.051823358535766605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,2,64,power_law_1.2,0.051502079963684076
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,2,64,power_law_1.2,0.051299839019775396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,2,64,power_law_1.2,0.055870718955993645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,2,64,power_law_1.2,0.0579968023300171
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,2,64,power_law_1.2,2.2587046813964844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,2,64,power_law_1.2,0.06286464214324952
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,2,64,power_law_1.2,0.06846464157104493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,2,64,power_law_1.2,0.07468671798706054
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,2,64,power_law_1.2,0.08871552467346192
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,2,64,power_law_1.2,0.11662976264953613
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,2,64,power_law_1.2,0.04528895854949951
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,2,64,power_law_1.2,0.13393792152404785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,2,64,power_law_1.2,0.20901248931884764
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,2,64,power_law_1.2,0.2433024024963379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,2,64,power_law_1.2,0.33998462677001956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,2,64,power_law_1.2,0.06568192005157471
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,2,64,power_law_1.2,0.4504601669311524
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,2,64,power_law_1.2,0.02191871881484985
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,2,64,power_law_1.2,0.02150144100189209
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,2,64,power_law_1.2,0.021845760345458983
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,2,64,power_law_1.2,0.021098239421844484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,2,64,power_law_1.2,0.7800537872314453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,2,64,power_law_1.2,0.021030399799346924
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,2,64,power_law_1.2,0.020695040225982668
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,2,64,power_law_1.2,0.021831679344177245
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,2,64,power_law_1.2,0.021425919532775877
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,2,64,power_law_1.2,0.021493759155273438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,2,64,power_law_1.2,1.0395801544189454
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,2,64,power_law_1.2,0.021446399688720703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,2,64,power_law_1.2,0.021512959003448486
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,2,64,power_law_1.2,0.021512959003448486
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,2,64,power_law_1.2,0.022261760234832763
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,2,64,power_law_1.2,0.02186624050140381
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,2,64,power_law_1.2,0.022312960624694823
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,2,64,power_law_1.2,0.02595328092575073
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,2,64,power_law_1.2,0.02189824104309082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,2,64,power_law_1.01,0.09275903701782226
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,2,64,power_law_1.01,0.12247424125671387
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,2,64,power_law_1.01,0.16384000778198243
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,2,64,power_law_1.2,0.023557119369506836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,2,64,power_law_1.2,0.03002624034881592
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,2,64,power_law_1.2,0.061795840263366694
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,2,64,power_law_1.2,0.0412556791305542
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,4,32,balanced,0.03043584108352661
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,4,32,balanced,0.034237439632415774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,4,32,balanced,0.033799679279327394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,4,32,balanced,0.03170815944671631
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,4,32,balanced,0.051199998855590824
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,2,64,power_law_1.2,0.0771891212463379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,4,32,balanced,0.0655999994277954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,4,32,balanced,0.11290111541748046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,4,32,balanced,0.11101183891296387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,4,32,balanced,0.11673983573913574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,4,32,balanced,0.11413248062133789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,2,64,power_law_1.2,0.046327037811279295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,4,32,balanced,0.12123135566711425
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,4,32,balanced,0.11715968132019043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,4,32,balanced,0.11970560073852539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,4,32,balanced,0.1229644775390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,4,32,balanced,0.12641663551330568
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,4,32,balanced,0.13333375930786134
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,4,32,balanced,0.1253990364074707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,4,32,balanced,0.1348799991607666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,4,32,balanced,0.1420595169067383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,4,32,balanced,0.15399423599243164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,4,32,balanced,0.16491519927978515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,4,32,balanced,0.19784320831298827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,4,32,balanced,0.22525184631347656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,4,32,balanced,0.28279552459716795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,4,32,balanced,0.3409075164794922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,4,32,balanced,0.035041279792785644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,4,32,balanced,0.4630515289306641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,4,32,balanced,0.03207551956176758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,4,32,balanced,0.03500799894332886
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,4,32,balanced,0.032207360267639165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,4,32,balanced,0.6164300918579102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,4,32,balanced,0.035100159645080564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,4,32,balanced,0.03880448102951049
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,4,32,balanced,0.05028736114501953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,4,32,balanced,0.05058047771453857
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,4,32,balanced,0.054261760711669924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,4,32,balanced,0.050140161514282224
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,4,32,balanced,0.05635327816009521
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,4,32,balanced,0.0513651180267334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,4,32,balanced,0.05417984008789063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,4,32,balanced,0.05700096130371094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,4,32,balanced,0.06456192016601563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,4,32,balanced,0.05608704090118408
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,4,32,balanced,0.06404352188110352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,4,32,balanced,0.07001471996307373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,4,32,balanced,0.0742796802520752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,4,32,balanced,0.08349823951721191
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,4,32,balanced,0.1015321636199951
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,4,32,balanced,0.11816831588745118
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,4,32,balanced,0.16046720504760742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,4,32,balanced,0.19335552215576174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,4,32,balanced,0.2734041595458984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,4,32,balanced,0.05059840202331543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,4,32,balanced,0.34832641601562503
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,2,64,power_law_1.2,0.10010111808776856
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,2,64,power_law_1.2,0.11876352310180664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,2,64,power_law_1.2,0.2838105583190918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,2,64,power_law_1.2,0.21961984634399415
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,4,32,balanced,0.01944831967353821
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,4,32,balanced,0.02274816036224365
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,4,32,balanced,0.023104000091552734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,4,32,balanced,0.02308095932006836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,4,32,balanced,0.018981120586395263
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,4,32,balanced,0.023144960403442383
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,4,32,balanced,0.023481600284576416
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,4,32,balanced,0.023479039669036868
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,4,32,balanced,0.023211519718170166
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,4,32,balanced,0.023543040752410888
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,4,32,balanced,0.02346751928329468
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,4,32,balanced,0.0235263991355896
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,4,32,balanced,0.02476032018661499
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,4,32,balanced,0.019477759599685666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,4,32,balanced,0.023612160682678223
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,4,32,balanced,0.027159039974212644
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,4,32,balanced,0.026440958976745605
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,4,32,balanced,0.01980672001838684
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,4,32,balanced,0.027232000827789305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,4,32,balanced,0.02142848014831543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,4,32,power_law_1.01,0.06410496234893799
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,4,32,power_law_1.01,0.06433023929595946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,4,32,power_law_1.01,0.06569983959197997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,4,32,power_law_1.01,0.06762112140655517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,4,32,power_law_1.01,0.11029120445251465
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,4,32,balanced,0.02270335912704468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,4,32,power_law_1.01,0.1120486354827881
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,4,32,power_law_1.01,0.11139455795288086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,4,32,power_law_1.01,0.11535103797912598
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,4,32,power_law_1.01,0.11928319931030273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,4,32,power_law_1.01,0.12165760040283202
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,4,32,power_law_1.01,0.11572735786437989
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,4,32,power_law_1.01,0.12436351776123047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,4,32,power_law_1.01,0.12736384391784666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,4,32,power_law_1.01,0.13246848106384276
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,4,32,power_law_1.01,0.13179519653320312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,4,32,power_law_1.01,0.13109760284423827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,4,32,power_law_1.01,0.13764863967895508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,4,32,power_law_1.01,0.15575424194335938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,4,32,power_law_1.01,0.17626239776611327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,4,32,power_law_1.01,0.19308416366577147
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,4,32,power_law_1.01,0.22970239639282228
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,4,32,power_law_1.01,0.30672256469726566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,4,32,power_law_1.01,0.37744640350341796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,4,32,power_law_1.01,0.5733158493041992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,4,32,power_law_1.01,0.7355084991455078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,4,32,power_law_1.01,0.9649433898925782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,4,32,power_law_1.01,0.03840255975723267
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,4,32,power_law_1.01,1.395061798095703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,4,32,power_law_1.01,0.03880448102951049
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,4,32,power_law_1.01,0.03802495956420898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,4,32,power_law_1.01,0.04174335956573486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,4,32,power_law_1.01,0.04735743999481201
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,4,32,power_law_1.01,0.050481920242309575
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,4,32,power_law_1.01,0.04894591808319092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,4,32,power_law_1.01,0.056770558357238765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,4,32,power_law_1.01,0.05088511943817139
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,4,32,power_law_1.01,0.05460351943969727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,4,32,power_law_1.01,0.05384704113006592
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,4,32,power_law_1.01,0.057523198127746575
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,4,32,power_law_1.01,0.05761536121368408
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,4,32,power_law_1.01,0.06392320156097411
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,4,32,power_law_1.01,0.060070400238037114
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,4,32,balanced,0.02311295986175537
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,4,32,balanced,0.02275455951690674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,4,32,power_law_1.01,0.06744063854217529
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,4,32,power_law_1.01,0.0731379222869873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,4,32,power_law_1.01,0.08916864395141602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,4,32,power_law_1.01,0.09849984169006347
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,4,32,power_law_1.01,0.1225062370300293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,4,32,power_law_1.01,0.16097024917602537
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,4,32,power_law_1.01,0.1930227279663086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,4,32,power_law_1.01,0.3071590423583984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,4,32,power_law_1.01,0.062471680641174324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,4,32,power_law_1.01,0.40124160766601563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,4,32,power_law_1.01,0.022858240604400636
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,4,32,power_law_1.01,0.02475264072418213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,4,32,power_law_1.01,0.5551961517333984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,4,32,power_law_1.01,0.02305919885635376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,4,32,power_law_1.01,0.024561920166015626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,4,32,power_law_1.01,0.02271104097366333
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,4,32,power_law_1.01,0.02205183982849121
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,4,32,power_law_1.01,0.023906559944152833
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,4,32,power_law_1.01,0.8478553771972657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,4,32,power_law_1.01,0.02255359888076782
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,4,32,power_law_1.01,0.02231679916381836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,4,32,power_law_1.01,0.02413055896759033
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,4,32,power_law_1.01,0.02308864116668701
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,4,32,power_law_1.01,0.023147521018981935
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,4,32,power_law_1.01,0.02321919918060303
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,4,32,power_law_1.01,0.023891201019287107
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,4,32,power_law_1.01,0.023531520366668703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,4,32,power_law_1.01,0.02422271966934204
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,4,32,power_law_1.01,0.023554561138153077
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,4,32,balanced,0.03864831924438476
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,4,32,balanced,0.04294400215148926
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,4,32,balanced,0.057393918037414546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,4,32,balanced,0.06635519981384277
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,4,32,power_law_1.01,0.03581056118011475
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,4,32,power_law_1.01,0.026822400093078614
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,4,32,power_law_1.01,0.03048703908920288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,4,32,power_law_1.2,0.06348031997680664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,4,32,power_law_1.2,0.06408448219299316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,4,32,power_law_1.2,0.06479872226715087
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,4,32,power_law_1.2,0.06528768062591553
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,4,32,power_law_1.01,0.039450879096984866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,4,32,power_law_1.2,0.10713855743408203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,4,32,power_law_1.01,0.05255807876586914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,4,32,power_law_1.2,0.10964096069335938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,4,32,power_law_1.2,0.11246591567993165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,4,32,power_law_1.2,0.11593728065490723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,4,32,power_law_1.2,0.1180569553375244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,4,32,power_law_1.2,0.11854463577270509
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,4,32,power_law_1.2,0.12473471641540528
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,4,32,power_law_1.01,0.10431232452392578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,4,32,power_law_1.2,0.1313203239440918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,4,32,power_law_1.2,0.13180288314819336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,4,32,power_law_1.2,0.12996607780456543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,4,32,power_law_1.2,0.13743743896484376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,4,32,power_law_1.2,0.14427007675170897
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,4,32,power_law_1.2,0.15959039688110352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,4,32,power_law_1.2,0.17592191696166992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,4,32,power_law_1.2,0.2054899215698242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,4,32,power_law_1.01,0.07092735767364503
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,4,32,power_law_1.2,0.25884544372558593
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,4,32,power_law_1.01,0.08288384437561035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,4,32,power_law_1.2,0.12419967651367188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,4,32,power_law_1.2,0.40686080932617186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,4,32,power_law_1.2,0.4610790252685547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,4,32,power_law_1.2,0.03752831935882568
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,4,32,power_law_1.2,0.5768896102905273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,4,32,power_law_1.2,0.03871999979019165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,4,32,power_law_1.2,0.038672640323638915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,4,32,power_law_1.2,0.03891072034835815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,4,32,power_law_1.2,0.9410470581054687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,4,32,power_law_1.2,0.04851840019226074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,4,32,power_law_1.2,0.04620672225952148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,4,32,power_law_1.2,0.050024957656860346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,4,32,power_law_1.2,0.055505919456481936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,4,32,power_law_1.2,1.3325491333007813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,4,32,power_law_1.2,0.05439743995666504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,4,32,power_law_1.2,0.05058944225311279
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,4,32,power_law_1.2,0.05200384140014649
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,4,32,power_law_1.2,0.05990272045135499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,4,32,power_law_1.2,0.06380799770355225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,4,32,power_law_1.2,0.06706048011779785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,4,32,power_law_1.2,0.06281856060028077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,4,32,power_law_1.2,0.06447743892669677
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,4,32,power_law_1.2,0.0700108814239502
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,4,32,power_law_1.2,1.7531097412109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,4,32,power_law_1.2,0.07663616180419922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,4,32,power_law_1.2,0.08775424003601075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,4,32,power_law_1.2,0.10004223823547362
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,4,32,power_law_1.2,0.14184320449829102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,4,32,power_law_1.2,0.22307071685791016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,4,32,power_law_1.2,0.24038143157958985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,4,32,power_law_1.2,0.3571072006225586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,4,32,power_law_1.2,0.022828800678253176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,4,32,power_law_1.2,0.4542195129394531
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,4,32,power_law_1.2,0.022731521129608155
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,4,32,power_law_1.2,0.02232320070266724
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,4,32,power_law_1.2,0.022874879837036132
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,4,32,power_law_1.2,0.74700927734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,4,32,power_law_1.2,0.02228096008300781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,4,32,power_law_1.2,0.021544959545135498
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,4,32,power_law_1.2,0.023137280941009523
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,4,32,power_law_1.2,0.022499840259552
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,4,32,power_law_1.2,0.02269695997238159
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,4,32,power_law_1.2,1.0153382110595703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,4,32,power_law_1.2,0.022760961055755615
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,4,32,power_law_1.2,0.022983679771423342
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,4,32,power_law_1.2,0.023095040321350097
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,4,32,power_law_1.2,0.02367232084274292
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,4,32,power_law_1.2,0.02403712034225464
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,4,32,power_law_1.2,0.02407680034637451
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,4,32,power_law_1.2,0.025280001163482668
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,4,32,power_law_1.2,0.022726399898529055
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,4,32,power_law_1.01,0.13362175941467286
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,4,32,power_law_1.01,0.19272703170776367
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,4,32,power_law_1.2,0.05007743835449219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,4,32,power_law_1.2,0.06742911815643311
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,4,32,power_law_1.2,0.030686719417572023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,8,16,balanced,0.030858240127563476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,8,16,balanced,0.0316044807434082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,4,32,power_law_1.2,0.03362688064575196
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,4,32,power_law_1.2,0.039792640209197996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,8,16,balanced,0.02962048053741455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,8,16,balanced,0.03998591899871826
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,8,16,balanced,0.05552896022796631
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,8,16,balanced,0.06995200157165528
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,8,16,balanced,0.11783295631408693
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,8,16,balanced,0.11984640121459962
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,8,16,balanced,0.12168319702148438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,8,16,balanced,0.12103808403015137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,8,16,balanced,0.12348031997680664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,8,16,balanced,0.127455997467041
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,4,32,power_law_1.2,0.08109824180603027
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,8,16,balanced,0.1319040012359619
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,8,16,balanced,0.12559616088867187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,8,16,balanced,0.1315814399719238
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,8,16,balanced,0.1416051197052002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,8,16,balanced,0.15083519935607909
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,8,16,balanced,0.16083967208862304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,8,16,balanced,0.1869990348815918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,8,16,balanced,0.21382400512695315
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,8,16,balanced,0.26756992340087893
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,8,16,balanced,0.3213350296020508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,8,16,balanced,0.4288678359985352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,8,16,balanced,0.5344204711914062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,8,16,balanced,0.13484288215637208
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,8,16,balanced,0.031366400718688965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,8,16,balanced,0.03168384075164795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,8,16,balanced,0.7495257568359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,8,16,balanced,0.03658879995346069
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,8,16,balanced,0.038420479297637936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,8,16,balanced,0.048102397918701176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,8,16,balanced,0.05585919857025147
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,8,16,balanced,0.9991487884521485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,8,16,balanced,0.06253056049346924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,8,16,balanced,0.057669119834899904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,8,16,balanced,0.05738880157470703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,8,16,balanced,0.0640998411178589
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,8,16,balanced,0.06312575817108154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,8,16,balanced,0.07131648063659668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,8,16,balanced,0.061528320312499996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,8,16,balanced,0.06703616142272949
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,8,16,balanced,0.0702348804473877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,8,16,balanced,0.07555967807769776
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,8,16,balanced,0.033733119964599614
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,8,16,balanced,0.07895808219909668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,8,16,balanced,0.08898431777954101
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,8,16,balanced,0.10394623756408691
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,8,16,balanced,0.11935999870300293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,8,16,balanced,0.15528191566467284
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,8,16,balanced,0.18284671783447265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,8,16,balanced,0.2631961631774902
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,8,16,balanced,0.3292812728881836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,8,16,balanced,0.46405502319335934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,8,16,balanced,0.5964710235595703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,4,32,power_law_1.2,0.11610112190246583
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,4,32,power_law_1.2,0.13275391578674317
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,4,32,power_law_1.2,0.23872768402099612
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,4,32,power_law_1.2,0.29081087112426757
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,8,16,balanced,0.02028287887573242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,8,16,balanced,0.028428800106048584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,8,16,balanced,0.028465919494628907
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,8,16,balanced,0.028444159030914306
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,8,16,balanced,0.02839168071746826
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,8,16,balanced,0.028431360721588132
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,8,16,balanced,0.019799040555953978
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,8,16,balanced,0.02885119915008545
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,8,16,balanced,0.028816640377044678
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,8,16,balanced,0.02887295961380005
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,8,16,balanced,0.029214720726013183
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,8,16,balanced,0.02892416000366211
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,8,16,balanced,0.030949120521545408
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,8,16,balanced,0.032048640251159666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,8,16,balanced,0.022269439697265626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,8,16,balanced,0.020689918994903564
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,8,16,balanced,0.03422847986221313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,8,16,balanced,0.036426239013671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,8,16,balanced,0.04117631912231445
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,8,16,balanced,0.0232422399520874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,8,16,power_law_1.01,0.03699455976486206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,8,16,power_law_1.01,0.06579840183258057
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,8,16,power_law_1.01,0.0669209623336792
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,8,16,power_law_1.01,0.06822271823883057
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,8,16,power_law_1.01,0.07097087860107422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,8,16,power_law_1.01,0.1178598403930664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,8,16,power_law_1.01,0.11986432075500489
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,8,16,power_law_1.01,0.11895296096801758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,8,16,power_law_1.01,0.11935232162475587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,8,16,power_law_1.01,0.12156160354614258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,8,16,power_law_1.01,0.12697983741760255
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,8,16,power_law_1.01,0.13195008277893067
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,8,16,power_law_1.01,0.1309657573699951
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,8,16,power_law_1.01,0.13248767852783203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,8,16,balanced,0.02820224046707153
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,8,16,power_law_1.01,0.13597567558288576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,8,16,power_law_1.01,0.13802240371704103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,8,16,power_law_1.01,0.14836480140686034
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,8,16,power_law_1.01,0.16721536636352538
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,8,16,power_law_1.01,0.18624256134033204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,8,16,power_law_1.01,0.23583871841430665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,8,16,power_law_1.01,0.27533824920654293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,8,16,power_law_1.01,0.32630401611328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,8,16,power_law_1.01,0.40896255493164063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,8,16,power_law_1.01,0.5854950332641602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,8,16,power_law_1.01,0.7681126403808594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,8,16,power_law_1.01,1.1010470581054688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,8,16,power_law_1.01,0.03706624031066895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,8,16,power_law_1.01,0.04262527942657471
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,8,16,power_law_1.01,1.4831846618652345
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,8,16,balanced,0.028414719104766846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,8,16,power_law_1.01,0.04312064170837403
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,8,16,power_law_1.01,0.04608895778656006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,8,16,power_law_1.01,0.04656383991241455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,8,16,power_law_1.01,0.05306496143341064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,8,16,power_law_1.01,0.05670271873474121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,8,16,power_law_1.01,0.06001023769378662
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,8,16,power_law_1.01,0.05705088138580322
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,8,16,power_law_1.01,0.06107135772705078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,8,16,power_law_1.01,0.06343039989471436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,8,16,power_law_1.01,0.06641664028167724
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,8,16,balanced,0.028116478919982913
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,8,16,power_law_1.01,0.06648064136505127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,8,16,power_law_1.01,0.07025152206420898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,8,16,power_law_1.01,0.07203072071075439
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,8,16,power_law_1.01,0.07596543788909912
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,8,16,power_law_1.01,0.08624896049499511
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,8,16,power_law_1.01,0.1011520004272461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,8,16,power_law_1.01,0.12414719581604003
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,8,16,power_law_1.01,0.15273088455200196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,8,16,power_law_1.01,0.18802560806274413
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,8,16,power_law_1.01,0.24435199737548827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,8,16,power_law_1.01,0.31966720581054686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,8,16,power_law_1.01,0.06980095863342285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,8,16,power_law_1.01,0.022722558975219725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,8,16,power_law_1.01,0.4419136047363281
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,8,16,power_law_1.01,0.028531200885772705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,8,16,power_law_1.01,0.6056332778930664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,8,16,power_law_1.01,0.0281279993057251
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,8,16,power_law_1.01,0.028463358879089358
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,8,16,power_law_1.01,0.028454399108886717
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,8,16,power_law_1.01,0.02721152067184448
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,8,16,power_law_1.01,0.8526297760009764
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,8,16,power_law_1.01,0.028096001148223877
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,8,16,power_law_1.01,0.028061439990997316
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,8,16,power_law_1.01,0.028419840335845947
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,8,16,power_law_1.01,0.028409600257873535
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,8,16,power_law_1.01,0.02900736093521118
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,8,16,power_law_1.01,0.0289574408531189
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,8,16,power_law_1.01,0.028980479240417478
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,8,16,power_law_1.01,0.029198079109191893
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,8,16,power_law_1.01,0.030495998859405515
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,8,16,power_law_1.01,0.03174527883529663
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,8,16,power_law_1.01,0.028349440097808837
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,8,16,balanced,0.08090239524841308
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,8,16,balanced,0.06111616134643555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,8,16,balanced,0.05578368186950684
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,8,16,balanced,0.09619711875915528
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,8,16,power_law_1.01,0.037182719707489015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,8,16,power_law_1.01,0.049675521850585935
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,8,16,power_law_1.01,0.03918335914611816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,8,16,power_law_1.2,0.03462271928787232
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,8,16,power_law_1.01,0.05904767990112305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,8,16,power_law_1.2,0.06601727962493896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,8,16,power_law_1.2,0.06280960083007812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,8,16,power_law_1.2,0.06685823917388915
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,8,16,power_law_1.01,0.07801727771759033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,8,16,power_law_1.2,0.0700928020477295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,8,16,power_law_1.2,0.11276927947998047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,8,16,power_law_1.2,0.11598464012145997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,8,16,power_law_1.2,0.11772671699523926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,8,16,power_law_1.2,0.11883647918701172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,8,16,power_law_1.2,0.12310400009155273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,8,16,power_law_1.2,0.12909055709838868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,8,16,power_law_1.2,0.12966527938842773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,8,16,power_law_1.2,0.1361574363708496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,8,16,power_law_1.2,0.13255295753479004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,8,16,power_law_1.2,0.13664511680603025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,8,16,power_law_1.2,0.14604800224304199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,8,16,power_law_1.2,0.17271808624267576
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,8,16,power_law_1.01,0.0834227180480957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,8,16,power_law_1.2,0.19924863815307617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,8,16,power_law_1.2,0.2367692756652832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,8,16,power_law_1.2,0.2920063972473145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,8,16,power_law_1.2,0.39275135040283204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,8,16,power_law_1.2,0.14058752059936525
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,8,16,power_law_1.2,0.4737139129638672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,8,16,power_law_1.01,0.11314432144165039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,8,16,power_law_1.2,0.7227891540527344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,8,16,power_law_1.2,0.0398963189125061
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,8,16,power_law_1.2,0.04334080219268799
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,8,16,power_law_1.2,0.9312614440917969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,8,16,power_law_1.2,0.043613438606262205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,8,16,power_law_1.2,0.04721536159515381
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,8,16,power_law_1.2,0.052413439750671385
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,8,16,power_law_1.2,1.3504396057128907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,8,16,power_law_1.2,0.056720638275146486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,8,16,power_law_1.2,0.055221757888793944
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,8,16,power_law_1.2,0.0600870418548584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,8,16,power_law_1.2,0.06129792213439942
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,8,16,power_law_1.2,1.8092160034179687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,8,16,power_law_1.2,0.04259200096130371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,8,16,power_law_1.2,0.06288640022277832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,8,16,power_law_1.2,0.07053696155548096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,8,16,power_law_1.2,0.06777088165283203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,8,16,power_law_1.2,0.06944767951965332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,8,16,power_law_1.2,0.07117568016052246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,8,16,power_law_1.2,0.07743872165679931
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,8,16,power_law_1.2,0.09002880096435548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,8,16,power_law_1.2,0.10659968376159668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,8,16,power_law_1.2,0.13459327697753906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,8,16,power_law_1.2,0.15714048385620116
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,8,16,power_law_1.01,0.15014016151428222
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,8,16,power_law_1.2,0.20567552566528322
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,8,16,power_law_1.2,0.2779430389404297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,8,16,power_law_1.2,0.4184307098388672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,8,16,power_law_1.2,0.023079679012298585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,8,16,power_law_1.2,0.5493465423583984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,8,16,power_law_1.2,0.0697331190109253
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,8,16,power_law_1.2,0.028076798915863038
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,8,16,power_law_1.2,0.027403519153594968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,8,16,power_law_1.2,0.7627954864501953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,8,16,power_law_1.2,0.029267199039459228
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,8,16,power_law_1.2,0.028367359638214112
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,8,16,power_law_1.2,0.027729918956756593
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,8,16,power_law_1.2,0.02560256004333496
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,8,16,power_law_1.2,0.028440320491790773
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,8,16,power_law_1.2,0.028447999954223636
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,8,16,power_law_1.2,0.028407039642333987
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,8,16,power_law_1.2,0.029747200012207032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,8,16,power_law_1.2,1.2127474975585937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,8,16,power_law_1.2,0.02964735984802246
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,8,16,power_law_1.2,0.02961024045944214
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,8,16,power_law_1.2,0.030918400287628173
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,8,16,power_law_1.2,0.027156479358673095
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,8,16,power_law_1.2,0.028546559810638427
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,8,16,power_law_1.01,0.24096511840820312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,8,16,power_law_1.01,0.28798847198486327
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,8,16,power_law_1.2,0.034031360149383544
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,8,16,power_law_1.2,0.05860991954803467
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,8,16,power_law_1.2,0.04121600151062012
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,8,16,power_law_1.2,0.04774784088134766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,16,8,balanced,0.03134464025497437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,16,8,balanced,0.03005824089050293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,16,8,balanced,0.032572159767150874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,16,8,balanced,0.03631871938705444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,16,8,balanced,0.05900415897369384
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,16,8,balanced,0.0763366413116455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,16,8,balanced,0.12953215599060058
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,8,16,power_law_1.2,0.07148287773132325
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,8,16,power_law_1.2,0.09810688018798827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,16,8,balanced,0.1327180767059326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,16,8,balanced,0.13716352462768555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,16,8,balanced,0.14158720016479492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,16,8,balanced,0.13298048019409178
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,16,8,balanced,0.13837311744689942
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,16,8,balanced,0.1416870403289795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,16,8,balanced,0.14853759765625002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,16,8,balanced,0.15148032188415528
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,16,8,balanced,0.15722240447998045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,16,8,balanced,0.16816255569458008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,16,8,balanced,0.19298944473266602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,16,8,balanced,0.21856256484985354
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,16,8,balanced,0.2690342330932617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,16,8,balanced,0.3204108810424805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,16,8,balanced,0.42305152893066406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,16,8,balanced,0.5248921585083008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,16,8,balanced,0.7218649291992187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,16,8,balanced,0.9234496307373046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,16,8,balanced,0.032112638950347897
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,16,8,balanced,0.03457792043685913
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,16,8,balanced,0.03948287963867188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,16,8,balanced,1.3286872863769532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,16,8,balanced,0.043564801216125486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,16,8,balanced,0.04765952110290527
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,16,8,balanced,0.06708352088928222
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,16,8,balanced,0.0695423984527588
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,16,8,balanced,0.07101568222045898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,16,8,balanced,0.07417856216430664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,16,8,balanced,0.07607295989990234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,16,8,balanced,1.7781158447265626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,16,8,balanced,0.07344895839691162
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,16,8,balanced,0.07831424236297607
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,16,8,balanced,0.08050687789916992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,16,8,balanced,0.08571264266967774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,16,8,balanced,0.08653056144714355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,16,8,balanced,0.09013504028320313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,16,8,balanced,0.09885184288024902
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,16,8,balanced,0.11430399894714356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,16,8,balanced,0.12989184379577637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,16,8,balanced,0.1614963150024414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,16,8,balanced,0.19154943466186522
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,16,8,balanced,0.2570790481567383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,16,8,balanced,0.31632768630981445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,16,8,balanced,0.46090240478515626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,16,8,balanced,0.5795916748046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,16,8,balanced,0.8388019561767578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,16,8,balanced,1.1008013153076173
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,8,16,power_law_1.2,0.11313535690307616
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,8,16,power_law_1.2,0.1674777603149414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,8,16,power_law_1.2,0.20563968658447265
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,8,16,power_law_1.2,0.3104243278503418
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,8,16,power_law_1.2,0.4935769653320312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,16,8,balanced,0.023100159168243407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,16,8,balanced,0.040307202339172364
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,16,8,balanced,0.040422401428222655
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,16,8,balanced,0.04082943916320801
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,16,8,balanced,0.040839681625366206
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,16,8,balanced,0.041323518753051756
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,16,8,balanced,0.02350208044052124
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,16,8,balanced,0.0414412784576416
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,16,8,balanced,0.04128128051757812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,16,8,balanced,0.04160384178161621
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,16,8,balanced,0.04362368106842041
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,16,8,balanced,0.04451839923858643
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,16,8,balanced,0.024665598869323728
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,16,8,balanced,0.04948863983154297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,16,8,balanced,0.0525708818435669
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,16,8,balanced,0.05928703784942627
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,16,8,balanced,0.06394368171691894
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,16,8,balanced,0.04701056003570557
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,16,8,balanced,0.025699839591979978
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,16,8,balanced,0.02931328058242798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,16,8,power_law_1.01,0.03338624000549316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,16,8,power_law_1.01,0.037079041004180906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,16,8,power_law_1.01,0.06097919940948486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,16,8,power_law_1.01,0.0640012788772583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,16,8,power_law_1.01,0.06853759765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,16,8,power_law_1.01,0.07212543964385987
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,16,8,power_law_1.01,0.12599807739257812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,16,8,power_law_1.01,0.12705280303955077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,16,8,power_law_1.01,0.1285696029663086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,16,8,power_law_1.01,0.13329024314880372
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,16,8,power_law_1.01,0.13721856117248535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,16,8,power_law_1.01,0.13820287704467774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,16,8,power_law_1.01,0.13746432304382322
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,16,8,power_law_1.01,0.13848064422607423
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,16,8,power_law_1.01,0.14376319885253905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,16,8,power_law_1.01,0.1485414409637451
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,16,8,power_law_1.01,0.15812352180480957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,16,8,power_law_1.01,0.185163516998291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,16,8,power_law_1.01,0.2049113655090332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,16,8,power_law_1.01,0.24712448120117186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,16,8,power_law_1.01,0.2855936050415039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,16,8,balanced,0.040088319778442384
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,16,8,balanced,0.04052864074707031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,16,8,power_law_1.01,0.3928934478759766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,16,8,power_law_1.01,0.4671039962768555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,16,8,power_law_1.01,0.5727129745483398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,16,8,power_law_1.01,0.8604876708984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,16,8,balanced,0.04061439990997314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,16,8,power_law_1.01,0.03675008058547974
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,16,8,power_law_1.01,0.04201344013214111
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,16,8,power_law_1.01,0.04755839824676514
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,16,8,power_law_1.01,1.2603033447265626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,16,8,power_law_1.01,0.05103744029998779
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,16,8,power_law_1.01,0.05722239971160888
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,16,8,power_law_1.01,0.06942207813262939
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,16,8,power_law_1.01,0.0700057601928711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,16,8,power_law_1.01,1.6693901062011718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,16,8,power_law_1.01,0.0754918384552002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,16,8,power_law_1.01,0.07319680213928223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,16,8,power_law_1.01,0.08001919746398925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,16,8,power_law_1.01,0.07976064205169678
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,16,8,power_law_1.01,0.07516287803649903
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,16,8,power_law_1.01,0.08005632400512695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,16,8,power_law_1.01,0.0664793586730957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,16,8,power_law_1.01,0.08261759757995606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,16,8,power_law_1.01,0.08326144218444824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,16,8,power_law_1.01,0.08946175575256347
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,16,8,power_law_1.01,0.10485631942749023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,16,8,power_law_1.01,0.1208512020111084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,16,8,power_law_1.01,0.1469375991821289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,16,8,power_law_1.01,0.17007360458374024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,16,8,power_law_1.01,0.23957887649536133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,16,8,power_law_1.01,0.304007682800293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,16,8,power_law_1.01,0.41810302734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,16,8,power_law_1.01,0.025530879497528074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,16,8,power_law_1.01,0.5712857437133789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,16,8,power_law_1.01,0.02879487991333008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,16,8,power_law_1.01,0.7837273406982421
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,16,8,power_law_1.01,0.03314815998077393
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,16,8,power_law_1.01,0.033729279041290285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,16,8,power_law_1.01,0.03500927925109863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,16,8,power_law_1.01,1.0649958038330078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,16,8,power_law_1.01,0.0374399995803833
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,16,8,power_law_1.01,0.03707135915756225
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,16,8,power_law_1.01,0.04025728225708008
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,16,8,power_law_1.01,0.03729536056518555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,16,8,power_law_1.01,0.0385971188545227
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,16,8,power_law_1.01,0.03950079917907715
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,16,8,power_law_1.01,0.04048895835876465
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,16,8,power_law_1.01,0.041553921699523926
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,16,8,power_law_1.01,0.04343167781829834
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,16,8,power_law_1.01,0.045171198844909666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,16,8,power_law_1.01,0.04523007869720459
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,16,8,power_law_1.01,0.03691263914108277
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,16,8,balanced,0.087258882522583
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,16,8,balanced,0.09792256355285645
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,16,8,balanced,0.13845503807067872
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,16,8,balanced,0.17337215423583985
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,16,8,power_law_1.01,0.05558144092559815
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,16,8,power_law_1.01,0.07040128231048584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,16,8,power_law_1.01,0.05875584125518799
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,16,8,power_law_1.01,0.08881152153015137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,16,8,power_law_1.2,0.033381121158599855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,16,8,power_law_1.2,0.041877760887145996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,16,8,power_law_1.2,0.05806079864501953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,16,8,power_law_1.2,0.0630079984664917
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,16,8,power_law_1.2,0.06923136234283447
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,16,8,power_law_1.2,0.07514624118804931
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,16,8,power_law_1.2,0.12512127876281737
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,16,8,power_law_1.2,0.12547967910766603
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,16,8,power_law_1.2,0.12891776084899903
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,16,8,power_law_1.2,0.13848959922790527
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,16,8,power_law_1.2,0.13406463623046877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,16,8,power_law_1.2,0.133953275680542
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,16,8,power_law_1.2,0.1267430400848389
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,16,8,power_law_1.2,0.1422041606903076
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,16,8,power_law_1.2,0.14453887939453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,16,8,power_law_1.2,0.14946047782897948
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,16,8,power_law_1.2,0.15927680015563966
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,16,8,power_law_1.2,0.18155647277832032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,16,8,power_law_1.2,0.20821632385253905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,16,8,power_law_1.2,0.256367359161377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,16,8,power_law_1.2,0.29879936218261716
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,16,8,power_law_1.01,0.10949631690979005
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,16,8,power_law_1.2,0.3972454452514648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,16,8,power_law_1.2,0.49338623046874996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,16,8,power_law_1.2,0.6874636840820313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,16,8,power_law_1.01,0.12733311653137208
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,16,8,power_law_1.2,0.95683837890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,16,8,power_law_1.2,0.03668992042541504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,16,8,power_law_1.2,0.04180607795715332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,16,8,power_law_1.2,0.04528895854949951
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,16,8,power_law_1.2,1.3921676635742188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,16,8,power_law_1.2,0.05194496154785157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,16,8,power_law_1.2,0.05746560096740723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,16,8,power_law_1.2,0.061286401748657224
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,16,8,power_law_1.2,0.06975488185882568
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,16,8,power_law_1.2,0.07101183891296386
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,16,8,power_law_1.01,0.19148672103881836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,16,8,power_law_1.2,0.07280384063720703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,16,8,power_law_1.2,0.07386879920959473
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,16,8,power_law_1.2,0.07993343830108643
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,16,8,power_law_1.2,1.692200927734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,16,8,power_law_1.2,0.08012543678283691
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,16,8,power_law_1.2,0.07990272045135498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,16,8,power_law_1.2,0.08286720275878906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,16,8,power_law_1.2,0.08379648208618165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,16,8,power_law_1.2,0.09249407768249511
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,16,8,power_law_1.2,0.10753151893615723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,16,8,power_law_1.2,0.12421248435974122
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,16,8,power_law_1.2,0.15873408317565918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,16,8,power_law_1.2,0.18931711196899415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,16,8,power_law_1.2,0.25168256759643554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,16,8,power_law_1.2,0.3214080047607422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,16,8,power_law_1.2,0.07585919857025146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,16,8,power_law_1.2,0.451146240234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,16,8,power_law_1.01,0.2424575996398926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,16,8,power_law_1.2,0.564318733215332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,16,8,power_law_1.2,0.025530879497528074
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,16,8,power_law_1.2,0.028488960266113285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,16,8,power_law_1.2,0.8329804992675781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,16,8,power_law_1.2,0.0329036808013916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,16,8,power_law_1.2,0.03661184072494507
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,16,8,power_law_1.2,0.03715712070465088
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,16,8,power_law_1.2,0.03440256118774414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,16,8,power_law_1.2,1.2043762969970704
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,16,8,power_law_1.2,0.03770112037658692
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,16,8,power_law_1.2,0.03724800109863281
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,16,8,power_law_1.2,0.039815680980682375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,16,8,power_law_1.2,0.03917311906814575
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,16,8,power_law_1.2,0.04065408229827881
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,16,8,power_law_1.2,0.042201600074768066
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,16,8,power_law_1.2,0.04450816154479981
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,16,8,power_law_1.2,0.04650112152099609
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,16,8,power_law_1.2,0.030536959171295164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,16,8,power_law_1.2,0.03901695966720581
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,16,8,power_law_1.01,0.348136978149414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,16,8,power_law_1.01,0.5107455825805663
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,16,8,power_law_1.2,0.05126016139984131
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,16,8,power_law_1.2,0.06932608127593995
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,16,8,power_law_1.2,0.06650752067565918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,16,8,power_law_1.2,0.10299263954162599
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,16,8,power_law_1.2,0.14158207893371583
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,16,8,power_law_1.2,0.09153792381286621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,32,4,balanced,0.028847360610961915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,32,4,balanced,0.029912319183349613
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,32,4,balanced,0.03413631916046143
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,32,4,balanced,0.03953408002853394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,32,4,balanced,0.0625331211090088
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,32,4,balanced,0.08123135566711426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,32,4,balanced,0.08204416275024415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,32,4,balanced,0.08343040466308593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,32,4,balanced,0.14372223854064942
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,32,4,balanced,0.14719743728637696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,32,4,balanced,0.150577917098999
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,32,4,balanced,0.15387136459350587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,32,4,balanced,0.1597760009765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,32,4,balanced,0.1640127944946289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,32,4,balanced,0.16663423538208008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,32,4,balanced,0.17291648864746095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,32,4,balanced,0.1845964813232422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,32,4,balanced,0.20860288619995115
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,32,4,balanced,0.23404415130615233
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,32,4,balanced,0.2833472061157226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,32,4,balanced,0.33424129486083987
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,32,4,balanced,0.43772033691406254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,32,4,balanced,0.5363942337036133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,32,4,balanced,0.731883544921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,32,4,balanced,0.9338406372070314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,32,4,balanced,0.03170047998428345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,32,4,balanced,0.03457407951354981
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,32,4,balanced,0.04117248058319092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,32,4,balanced,1.3329061889648437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,32,4,balanced,0.052572159767150885
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,32,4,balanced,0.059827198982238775
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,32,4,balanced,0.07598464012145996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,32,4,balanced,0.07770880222320556
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,32,4,balanced,0.07884928226470947
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,32,4,balanced,0.09445119857788085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,32,4,balanced,1.7823858642578123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,32,4,balanced,0.09593983650207519
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,32,4,balanced,0.09823360443115234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,32,4,balanced,0.10259072303771972
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,32,4,balanced,0.10586496353149415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,32,4,balanced,0.11141375541687011
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,32,4,balanced,0.11093248367309569
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,32,4,balanced,0.11497599601745605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,32,4,balanced,0.12363519668579101
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,32,4,balanced,0.13903103828430177
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,32,4,balanced,0.1543449592590332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,32,4,balanced,0.18507904052734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,32,4,balanced,0.21503360748291014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,32,4,balanced,0.2770688056945801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,32,4,balanced,0.3361484909057617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,32,4,balanced,0.4942668914794922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,32,4,balanced,0.6123353576660155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,32,4,balanced,0.9076211547851563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,32,4,balanced,1.1874752044677734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,16,8,power_law_1.2,0.20124288558959963
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,16,8,power_law_1.2,0.2554956817626953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,16,8,power_law_1.2,0.3779430389404297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,16,8,power_law_1.2,0.5503488159179687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,16,8,power_law_1.2,0.7619174194335938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,32,4,balanced,0.030110719203948977
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,32,4,balanced,0.06017663955688477
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,32,4,balanced,0.031168000698089598
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,32,4,balanced,0.06057600021362305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,32,4,balanced,0.06009471893310547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,32,4,balanced,0.060501761436462395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,32,4,balanced,0.06104447841644287
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,32,4,balanced,0.0623692798614502
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,32,4,balanced,0.06302591800689697
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,32,4,balanced,0.0631987190246582
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,32,4,balanced,0.06489088058471679
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,32,4,balanced,0.06777984142303467
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,32,4,balanced,0.07022463798522949
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,32,4,balanced,0.03211519956588745
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,32,4,balanced,0.07589119911193848
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,32,4,balanced,0.08050304412841797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,32,4,balanced,0.09017472267150879
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,32,4,balanced,0.09960960388183594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,32,4,balanced,0.035412480831146245
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,32,4,balanced,0.044200959205627444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,32,4,power_law_1.01,0.03173504114151001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,32,4,power_law_1.01,0.03526015996932984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,32,4,power_law_1.01,0.05346176147460937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,32,4,power_law_1.01,0.06320000171661377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,32,4,power_law_1.01,0.06818175792694092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,32,4,power_law_1.01,0.07465856075286866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,32,4,power_law_1.01,0.07827328205108643
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,32,4,power_law_1.01,0.07859072208404541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,32,4,power_law_1.01,0.14140671730041504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,32,4,power_law_1.01,0.14304384231567382
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,32,4,power_law_1.01,0.14685312271118164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,32,4,power_law_1.01,0.15094528198242188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,32,4,power_law_1.01,0.1544320011138916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,32,4,power_law_1.01,0.16220031738281251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,32,4,power_law_1.01,0.16408447265624998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,32,4,power_law_1.01,0.16928512573242188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,32,4,power_law_1.01,0.18420991897583008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,32,4,power_law_1.01,0.20661376953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,32,4,power_law_1.01,0.23612415313720705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,32,4,power_law_1.01,0.2864883232116699
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,32,4,power_law_1.01,0.33547134399414064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,32,4,power_law_1.01,0.42523647308349605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,32,4,power_law_1.01,0.5232371139526367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,32,4,power_law_1.01,0.7176640319824219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,32,4,power_law_1.01,0.926935043334961
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,32,4,balanced,0.05940351963043213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,32,4,power_law_1.01,1.412095947265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,32,4,power_law_1.01,0.03551615953445435
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,32,4,balanced,0.05994624137878418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,32,4,power_law_1.01,1.7678477478027343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,32,4,power_law_1.01,0.04109055995941162
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,32,4,power_law_1.01,0.046108160018920895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,32,4,power_law_1.01,0.05994880199432373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,32,4,power_law_1.01,0.06467584133148194
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,32,4,power_law_1.01,0.06707839965820313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,32,4,power_law_1.01,0.07667200088500976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,32,4,power_law_1.01,0.0742412805557251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,32,4,power_law_1.01,0.0889459228515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,32,4,power_law_1.01,0.08666111946105956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,32,4,power_law_1.01,0.0927359962463379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,32,4,power_law_1.01,0.09544960021972657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,32,4,power_law_1.01,0.09896703720092773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,32,4,power_law_1.01,0.10729984283447265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,32,4,power_law_1.01,0.10644479751586915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,32,4,power_law_1.01,0.10882431983947753
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,32,4,balanced,0.0605734395980835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,32,4,power_law_1.01,0.11706368446350097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,32,4,power_law_1.01,0.13455871582031248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,32,4,power_law_1.01,0.1553548812866211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,32,4,power_law_1.01,0.19092479705810547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,32,4,power_law_1.01,0.22771711349487306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,32,4,power_law_1.01,0.29674367904663085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,32,4,power_law_1.01,0.36927745819091795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,32,4,power_law_1.01,0.5070003128051758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,32,4,power_law_1.01,0.6524134063720703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,32,4,power_law_1.01,0.032581119537353515
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,32,4,power_law_1.01,0.035741438865661626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,32,4,power_law_1.01,0.04020736217498779
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,32,4,power_law_1.01,0.04315392017364502
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,32,4,power_law_1.01,0.048382720947265624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,32,4,power_law_1.01,0.05220863819122314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,32,4,power_law_1.01,1.2350220489501953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,32,4,power_law_1.01,0.05711999893188476
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,32,4,power_law_1.01,0.05707520008087158
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,32,4,power_law_1.01,0.05828735828399658
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,32,4,power_law_1.01,0.059101438522338866
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,32,4,power_law_1.01,0.06023680210113526
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,32,4,power_law_1.01,0.06025728225708008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,32,4,power_law_1.01,0.955999984741211
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,32,4,power_law_1.01,0.06432767868041991
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,32,4,power_law_1.01,0.06691455841064453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,32,4,power_law_1.01,0.06740608215332031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,32,4,power_law_1.01,0.05632512092590332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,32,4,power_law_1.01,0.07128960132598877
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,32,4,balanced,0.14872063636779784
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,32,4,balanced,0.17185535430908203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,32,4,balanced,0.24784639358520505
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,32,4,balanced,0.315729923248291
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,32,4,power_law_1.01,0.08517760276794434
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,32,4,power_law_1.01,0.14544639587402344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,32,4,power_law_1.01,0.09662848472595215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,32,4,power_law_1.2,0.0341759991645813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,32,4,power_law_1.2,0.03429120063781738
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,32,4,power_law_1.01,0.12199935913085938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,32,4,power_law_1.2,0.054271998405456545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,32,4,power_law_1.2,0.062414078712463385
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,32,4,power_law_1.2,0.06528768062591553
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,32,4,power_law_1.2,0.07354623794555663
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,32,4,power_law_1.2,0.07678080081939698
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,32,4,power_law_1.2,0.07845632076263429
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,32,4,power_law_1.2,0.14214912414550782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,32,4,power_law_1.2,0.14191743850708008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,32,4,power_law_1.2,0.145666561126709
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,32,4,power_law_1.2,0.15061504364013673
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,32,4,power_law_1.2,0.15458304405212403
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,32,4,power_law_1.01,0.1811814308166504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,32,4,power_law_1.2,0.16327936172485352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,32,4,power_law_1.2,0.16528127670288087
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,32,4,power_law_1.2,0.1709657669067383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,32,4,power_law_1.2,0.18371456146240234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,32,4,power_law_1.2,0.20847232818603514
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,32,4,power_law_1.2,0.23452032089233396
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,32,4,power_law_1.01,0.22315904617309573
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,32,4,power_law_1.2,0.2829644775390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,32,4,power_law_1.01,0.33189247131347654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,32,4,power_law_1.2,0.34080383300781253
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,32,4,power_law_1.2,0.4449574279785156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,32,4,power_law_1.2,0.5479372787475586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,32,4,power_law_1.2,0.03538815975189209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,32,4,power_law_1.2,0.7442355346679688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,32,4,power_law_1.2,0.04070528030395508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,32,4,power_law_1.2,0.9755123138427735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,32,4,power_law_1.2,0.04486144065856933
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,32,4,power_law_1.2,0.05771008014678956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,32,4,power_law_1.2,0.0633523178100586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,32,4,power_law_1.2,0.06721536159515382
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,32,4,power_law_1.2,1.3795980834960937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,32,4,power_law_1.2,0.07564159870147705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,32,4,power_law_1.2,0.07395840167999268
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,32,4,power_law_1.2,0.0859391975402832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,32,4,power_law_1.2,0.08772864341735839
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,32,4,power_law_1.01,0.4374489593505859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,32,4,power_law_1.2,0.0911411190032959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,32,4,power_law_1.2,0.09521663665771485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,32,4,power_law_1.2,0.09797632217407226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,32,4,power_law_1.2,0.10529919624328614
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,32,4,power_law_1.2,0.1054361629486084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,32,4,power_law_1.2,1.8252954101562502
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,32,4,power_law_1.2,0.10874624252319336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,32,4,power_law_1.2,0.12053759574890137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,32,4,power_law_1.2,0.13693056106567383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,32,4,power_law_1.2,0.1586790370941162
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,32,4,power_law_1.2,0.19747583389282225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,32,4,power_law_1.2,0.23836927413940429
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,32,4,power_law_1.2,0.31120384216308594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,32,4,power_law_1.2,0.39840766906738284
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,32,4,power_law_1.2,0.03251840114593506
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,32,4,power_law_1.2,0.5595865631103516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,32,4,power_law_1.2,0.035411200523376464
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,32,4,power_law_1.2,0.6833894348144531
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,32,4,power_law_1.2,0.03828991889953613
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,32,4,power_law_1.2,0.04364031791687012
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,32,4,power_law_1.2,0.04765696048736572
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,32,4,power_law_1.2,0.04728576183319092
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,32,4,power_law_1.2,0.058442239761352544
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,32,4,power_law_1.2,0.9948774719238281
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,32,4,power_law_1.2,0.05812863826751709
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,32,4,power_law_1.2,0.057505278587341314
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,32,4,power_law_1.2,0.05896063804626465
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,32,4,power_law_1.2,0.059991040229797364
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,32,4,power_law_1.2,0.05919616222381592
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,32,4,power_law_1.2,0.06205952167510986
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,32,4,power_law_1.2,0.06568960189819337
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,32,4,power_law_1.2,1.3155020141601563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,32,4,power_law_1.2,0.0687769603729248
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,32,4,power_law_1.2,0.07191040039062499
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,32,4,power_law_1.01,0.6303372955322265
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,32,4,power_law_1.01,0.8301145935058594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,32,4,power_law_1.2,0.08054528236389161
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,32,4,power_law_1.2,0.09306752204895019
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,32,4,power_law_1.2,0.12023679733276367
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,32,4,power_law_1.2,0.27052543640136717
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,32,4,power_law_1.2,0.16426624298095702
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,128,balanced,0.020743680000305176
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,32,4,power_law_1.2,0.1790438461303711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,128,balanced,0.021478400230407715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,128,balanced,0.021029119491577146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,128,balanced,0.019471360445022585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,128,balanced,0.03047679901123047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,128,balanced,0.03209088087081909
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,128,balanced,0.031701760292053224
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,128,balanced,0.03301503896713257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,128,balanced,0.03350271940231323
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,128,balanced,0.0337280011177063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,128,balanced,0.03414783954620361
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,128,balanced,0.03410943984985352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,128,balanced,0.034970879554748535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,128,balanced,0.03584511995315552
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,128,balanced,0.035797760486602784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,128,balanced,0.03824127912521362
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,128,balanced,0.04006015777587891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,128,balanced,0.021534719467163087
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,128,balanced,0.04447231769561767
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,128,balanced,0.0407206392288208
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,128,balanced,0.04521344184875488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,128,balanced,0.048149762153625486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,128,balanced,0.05346432209014893
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,128,balanced,0.06077184200286865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,128,balanced,0.07722623825073242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,128,balanced,0.09295488357543945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,128,balanced,0.023975679874420168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,128,balanced,0.024723200798034667
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,128,balanced,0.038314239978790285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,128,balanced,0.02432255983352661
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,128,balanced,0.023644158840179445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,128,balanced,0.023960320949554442
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,128,balanced,0.028890879154205323
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,128,balanced,0.028881919384002686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,128,balanced,0.02971519947052002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,128,balanced,0.029638400077819826
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,128,balanced,0.029806079864501955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,128,balanced,0.030103039741516114
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,128,balanced,0.030178558826446534
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,128,balanced,0.02923775911331177
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,128,balanced,0.03069823980331421
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,128,balanced,0.030557439327239987
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,128,balanced,0.03176192045211792
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,128,balanced,0.031756799221038814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,128,balanced,0.03215744018554688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,128,balanced,0.034115839004516604
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,128,balanced,0.03618175983428955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,128,balanced,0.03339263916015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,128,balanced,0.03537791967391968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,128,balanced,0.03661312103271484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,128,balanced,0.040990719795227046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,128,balanced,0.04520959854125976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,128,balanced,0.052908802032470705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,128,balanced,0.06485375881195068
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,32,4,power_law_1.2,0.3309823989868164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,32,4,power_law_1.2,0.5326873779296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,32,4,power_law_1.2,0.6557734680175782
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,1,128,balanced,0.014109439849853517
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,1,128,balanced,0.0140774405002594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,1,128,balanced,0.014074879884719848
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,1,128,balanced,0.01413375973701477
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,1,128,balanced,0.014062080383300781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,1,128,balanced,0.014160640239715578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,1,128,balanced,0.013980159759521485
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,32,4,power_law_1.2,0.93106689453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,1,128,balanced,0.014337279796600341
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,1,128,balanced,0.01449087977409363
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,1,128,balanced,0.014083839654922487
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,1,128,balanced,0.014155520200729372
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,1,128,balanced,0.014232319593429566
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,1,128,balanced,0.014211839437484742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,1,128,balanced,0.014545919895172118
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,1,128,balanced,0.014493440389633178
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,1,128,balanced,0.014566400051116942
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,32,4,power_law_1.2,1.4011328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,1,128,balanced,0.014460159540176393
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,1,128,balanced,0.01410688042640686
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,1,128,balanced,0.014136320352554322
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,1,128,balanced,0.01444991946220398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,128,power_law_1.01,0.019440640211105347
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,128,power_law_1.01,0.019779839515686036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,128,power_law_1.01,0.030219519138336183
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,1,128,balanced,0.014438400268554688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,128,power_law_1.01,0.031247360706329348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,128,power_law_1.01,0.03210112094879151
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,128,power_law_1.01,0.03389440059661865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,128,power_law_1.01,0.03419647932052612
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,128,power_law_1.01,0.035842559337615966
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,128,power_law_1.01,0.03543551921844482
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,128,power_law_1.01,0.0382259202003479
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,1,128,balanced,0.014654719829559326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,128,power_law_1.01,0.03704960107803344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,128,power_law_1.01,0.03868671894073487
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,128,power_law_1.01,0.03923199892044067
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,128,power_law_1.01,0.043526401519775396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,128,power_law_1.01,0.04502399921417236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,128,power_law_1.01,0.044939517974853516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,1,128,balanced,0.015416320562362671
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,128,power_law_1.01,0.04519040107727051
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,128,power_law_1.01,0.050993919372558594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,1,128,balanced,0.0149619197845459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,128,power_law_1.01,0.05789055824279785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,128,power_law_1.01,0.08034560203552246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,128,power_law_1.01,0.0796134376525879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,128,power_law_1.01,0.16107135772705078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,128,power_law_1.01,0.25719039916992187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,128,power_law_1.01,0.3445375823974609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,128,power_law_1.01,0.02392832040786743
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,128,power_law_1.01,0.4852262496948242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,128,power_law_1.01,0.023953919410705564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,128,power_law_1.01,0.023508479595184324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,128,power_law_1.01,0.028463358879089358
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,128,power_law_1.01,0.7677043151855469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,128,power_law_1.01,0.02927615880966187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,128,power_law_1.01,0.029635839462280274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,128,power_law_1.01,0.029290239810943603
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,128,power_law_1.01,0.030524160861968998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,128,power_law_1.01,0.031123199462890626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,128,power_law_1.01,0.03173120021820068
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,128,power_law_1.01,0.03298943996429443
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,128,power_law_1.01,0.03465856075286865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,128,power_law_1.01,0.03703808069229126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,128,power_law_1.01,0.03789184093475342
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,1,128,balanced,0.015854079723358155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,128,power_law_1.01,0.03752576112747193
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,128,power_law_1.01,0.041418237686157225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,128,power_law_1.01,0.05437568187713623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,128,power_law_1.01,0.06049407958984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,128,power_law_1.01,0.12813055992126465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,128,power_law_1.01,0.03215104103088379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,128,power_law_1.01,0.0861683177947998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,128,power_law_1.01,0.1041215991973877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,128,power_law_1.01,0.15805695533752442
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,128,power_law_1.01,0.20154239654541017
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,128,power_law_1.01,0.03538304090499878
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,128,power_law_1.01,0.2790195274353027
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,1,128,power_law_1.01,0.013752319812774657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,1,128,power_law_1.01,0.014461439847946168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,128,power_law_1.01,0.42535678863525395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,1,128,power_law_1.01,0.014136320352554322
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,128,power_law_1.01,0.030521600246429442
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,1,128,power_law_1.01,0.014149119853973388
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,1,128,power_law_1.01,0.014131200313568116
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,1,128,power_law_1.01,0.014467840194702148
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,1,128,power_law_1.01,0.014513920545578003
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,1,128,power_law_1.01,0.014470399618148805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,1,128,power_law_1.01,0.014160640239715578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,1,128,power_law_1.01,0.014237439632415772
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,1,128,power_law_1.01,0.014471679925918579
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,1,128,power_law_1.01,0.014529279470443726
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,1,128,power_law_1.01,0.014533120393753051
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,1,128,power_law_1.01,0.01454848051071167
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,1,128,power_law_1.01,0.014888960123062133
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,1,128,power_law_1.01,0.014114559888839723
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,1,128,balanced,0.016925439834594727
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,1,128,balanced,0.017791999578475954
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,1,128,power_law_1.01,0.013766399621963502
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,1,128,power_law_1.01,0.014863359928131103
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,1,128,power_law_1.01,0.015352319478988647
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,1,128,power_law_1.01,0.020606720447540285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,128,power_law_1.2,0.01935744047164917
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,1,128,power_law_1.01,0.017008639574050903
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,128,power_law_1.2,0.019811840057373048
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,128,power_law_1.2,0.03261183977127075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,128,power_law_1.2,0.03215359926223755
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,1,128,power_law_1.01,0.023157761096954346
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,1,128,power_law_1.01,0.01737215995788574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,128,power_law_1.2,0.03347968101501465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,128,power_law_1.2,0.035352320671081544
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,128,power_law_1.2,0.03537791967391968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,128,power_law_1.2,0.0362175989151001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,128,power_law_1.2,0.03826688051223755
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,128,power_law_1.2,0.03825023889541626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,128,power_law_1.2,0.040769281387329104
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,1,128,power_law_1.01,0.02808448076248169
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,128,power_law_1.2,0.043795199394226075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,128,power_law_1.2,0.045194239616394044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,128,power_law_1.2,0.044313597679138186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,128,power_law_1.2,0.04584959983825684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,128,power_law_1.2,0.050599679946899415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,128,power_law_1.2,0.05960319995880127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,128,power_law_1.2,0.07975296020507813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,128,power_law_1.2,0.09725184440612793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,128,power_law_1.2,0.13287936210632326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,128,power_law_1.2,0.2156902313232422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,128,power_law_1.2,0.03419903993606567
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,128,power_law_1.2,0.2874803161621094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,128,power_law_1.2,0.025189120769500733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,128,power_law_1.2,0.024797439575195312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,128,power_law_1.2,0.029224960803985595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,128,power_law_1.2,0.029413120746612547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,128,power_law_1.2,0.0296396803855896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,128,power_law_1.2,0.4406617736816406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,128,power_law_1.2,0.030040318965911868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,128,power_law_1.2,0.0304422402381897
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,128,power_law_1.2,0.03055232048034668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,128,power_law_1.2,0.031741440296173096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,128,power_law_1.2,0.03191296100616455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,128,power_law_1.2,0.031681280136108395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,128,power_law_1.2,0.033354239463806154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,128,power_law_1.2,0.03495039939880371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,128,power_law_1.2,0.037391359806060794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,128,power_law_1.2,0.035778560638427735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,128,power_law_1.2,0.034713599681854244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,128,power_law_1.2,0.036669440269470215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,128,power_law_1.2,0.042817277908325194
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,128,power_law_1.2,0.0642419195175171
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,128,power_law_1.2,0.08653056144714355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,128,power_law_1.2,0.13181695938110352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,128,power_law_1.2,0.0547046422958374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,128,power_law_1.2,0.15986047744750978
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,128,power_law_1.2,0.25861888885498047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,128,power_law_1.2,0.3717670440673828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,128,power_law_1.2,0.5850777435302734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,1,128,power_law_1.2,0.014126080274581908
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,1,128,power_law_1.2,0.014135040044784547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,1,128,power_law_1.2,0.01410688042640686
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,1,128,power_law_1.2,0.014135040044784547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,128,power_law_1.2,1.1585971069335936
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,1,128,power_law_1.2,0.014097919464111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,1,128,power_law_1.2,0.01428607940673828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,1,128,power_law_1.2,0.01444991946220398
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,1,128,power_law_1.2,0.014465279579162599
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,1,128,power_law_1.2,0.01410688042640686
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,1,128,power_law_1.2,0.01445631980895996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,1,128,power_law_1.2,0.014213119745254516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,1,128,power_law_1.2,0.014511359930038454
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,1,128,power_law_1.2,0.014723839759826659
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,1,128,power_law_1.2,0.014606080055236817
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,128,power_law_1.2,0.5440371322631836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,1,128,power_law_1.2,0.014645760059356689
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,1,128,power_law_1.01,0.03468415975570679
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,1,128,power_law_1.01,0.04657663822174073
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,128,power_law_1.2,0.7311244964599609
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,1,128,power_law_1.01,0.06110720157623291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,128,power_law_1.2,0.901861114501953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,1,128,power_law_1.2,0.014520319700241088
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,1,128,power_law_1.2,0.015353599786758423
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,1,128,power_law_1.2,0.016962560415267943
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,1,128,power_law_1.2,0.014973440170288085
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,1,128,power_law_1.2,0.0181222403049469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,64,balanced,0.017748479843139646
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,64,balanced,0.017429759502410887
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,1,128,power_law_1.2,0.021619200706481934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,64,balanced,0.017337599992752077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,64,balanced,0.0173580801486969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,64,balanced,0.01875584006309509
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,64,balanced,0.028245759010314942
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,64,balanced,0.02967679977416992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,64,balanced,0.03004544019699097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,64,balanced,0.030263040065765377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,64,balanced,0.031188480854034424
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,1,128,power_law_1.2,0.026362879276275637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,64,balanced,0.032565760612487796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,64,balanced,0.03250943899154663
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,64,balanced,0.03460864067077637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,64,balanced,0.03363071918487549
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,64,balanced,0.034190080165863036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,64,balanced,0.037402880191802976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,64,balanced,0.03976576089859009
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,64,balanced,0.03635967969894409
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,64,balanced,0.041159682273864746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,64,balanced,0.04343552112579346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,64,balanced,0.04721920013427734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,64,balanced,0.05363711833953857
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,64,balanced,0.06672255992889405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,64,balanced,0.08168576240539552
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,64,balanced,0.03133824110031128
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,64,balanced,0.1120627212524414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,64,balanced,0.13780351638793947
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,64,balanced,0.02268160104751587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,64,balanced,0.022293760776519775
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,64,balanced,0.022653439044952393
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,64,balanced,0.023120639324188234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,64,balanced,0.027761919498443605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,64,balanced,0.028395519256591794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,64,balanced,0.02827775955200195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,64,balanced,0.02896768093109131
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,64,balanced,0.028812799453735348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,64,balanced,0.029304320812225344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,64,balanced,0.02882431983947754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,64,balanced,0.030186240673065183
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,64,balanced,0.030443520545959474
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,64,balanced,0.030432000160217288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,64,balanced,0.030935039520263673
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,64,balanced,0.032152318954467775
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,64,balanced,0.02234112024307251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,64,balanced,0.03462271928787232
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,64,balanced,0.03088000059127808
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,64,balanced,0.033016319274902346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,64,balanced,0.03545216083526611
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,64,balanced,0.038077440261840824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,64,balanced,0.04032896041870117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,64,balanced,0.04754432201385498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,64,balanced,0.05575424194335937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,64,balanced,0.07635968208312989
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,64,balanced,0.09446016311645508
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,1,128,power_law_1.2,0.03177472114562988
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,1,128,power_law_1.2,0.042686719894409184
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,1,128,power_law_1.2,0.07188735961914063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,1,128,power_law_1.2,0.09700736045837402
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,2,64,balanced,0.014474240541458131
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,2,64,balanced,0.014881279468536377
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,2,64,balanced,0.014865920543670655
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,2,64,balanced,0.014466559886932373
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,2,64,balanced,0.014944000244140627
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,2,64,balanced,0.01496448040008545
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,2,64,balanced,0.014944000244140627
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,2,64,balanced,0.0145305597782135
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,2,64,balanced,0.014924800395965577
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,2,64,balanced,0.01494655966758728
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,2,64,balanced,0.014969600439071657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,2,64,balanced,0.015334399938583373
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,2,64,balanced,0.014859520196914671
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,2,64,balanced,0.015054080486297607
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,2,64,balanced,0.01534208059310913
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,2,64,balanced,0.015311360359191895
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,2,64,balanced,0.014856959581375123
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,2,64,balanced,0.014442239999771117
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,1,128,power_law_1.2,0.054656000137329096
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,2,64,balanced,0.014620159864425658
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,2,64,balanced,0.01450111985206604
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,64,power_law_1.01,0.017418240308761594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,64,power_law_1.01,0.018236160278320312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,64,power_law_1.01,0.01788159966468811
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,64,power_law_1.01,0.029250559806823728
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,64,power_law_1.01,0.029702401161193846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,64,power_law_1.01,0.030908160209655762
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,2,64,balanced,0.014991359710693358
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,64,power_law_1.01,0.031198720932006836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,64,power_law_1.01,0.03209343910217285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,2,64,balanced,0.015947519540786742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,64,power_law_1.01,0.03280384063720703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,64,power_law_1.01,0.03350912094116211
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,2,64,balanced,0.015805439949035643
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,64,power_law_1.01,0.03421695947647095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,64,power_law_1.01,0.03501183986663818
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,2,64,balanced,0.01692288041114807
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,64,power_law_1.01,0.039637761116027834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,64,power_law_1.01,0.03922431945800781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,64,power_law_1.01,0.04049536228179932
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,64,power_law_1.01,0.0393395209312439
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,64,power_law_1.01,0.04166656017303467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,64,power_law_1.01,0.05023231983184815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,64,power_law_1.01,0.06895616054534912
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,64,power_law_1.01,0.08415231704711915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,64,power_law_1.01,0.11400447845458986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,64,power_law_1.01,0.14722559928894044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,64,power_law_1.01,0.20378240585327148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,64,power_law_1.01,0.2782143974304199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,64,power_law_1.01,0.022790400981903075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,64,power_law_1.01,0.458519058227539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,64,power_law_1.01,0.02310784101486206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,64,power_law_1.01,0.5636851119995117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,64,power_law_1.01,0.023535358905792236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,64,power_law_1.01,0.04576640129089356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,64,power_law_1.01,0.027619841098785403
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,64,power_law_1.01,0.02813568115234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,64,power_law_1.01,0.02871936082839966
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,64,power_law_1.01,0.029214720726013183
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,64,power_law_1.01,0.029667840003967282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,64,power_law_1.01,0.029653759002685548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,64,power_law_1.01,0.030480639934539793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,64,power_law_1.01,0.03247999906539917
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,64,power_law_1.01,0.03167360067367554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,64,power_law_1.01,0.03426431894302368
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,64,power_law_1.01,0.03342080116271973
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,64,power_law_1.01,0.034931199550628664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,64,power_law_1.01,0.03378943920135498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,64,power_law_1.01,0.03996031999588013
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,64,power_law_1.01,0.050109438896179206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,64,power_law_1.01,0.06207744121551514
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,64,power_law_1.01,0.08076416015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,2,64,balanced,0.01761664032936096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,64,power_law_1.01,0.11266559600830077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,64,power_law_1.01,0.15850367546081542
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,2,64,balanced,0.01977728009223938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,64,power_law_1.01,0.2206553649902344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,2,64,power_law_1.01,0.014506239891052247
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,64,power_law_1.01,0.2900364875793457
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,2,64,power_law_1.01,0.014873600006103516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,2,64,power_law_1.01,0.014933760166168214
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,64,power_law_1.01,0.39036544799804684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,64,power_law_1.01,0.033724160194396974
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,2,64,power_law_1.01,0.015010559558868408
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,2,64,power_law_1.01,0.015301120281219483
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,2,64,power_law_1.01,0.014888960123062133
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,2,64,power_law_1.01,0.014949120283126833
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,64,power_law_1.01,0.036735999584198
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,2,64,power_law_1.01,0.015037440061569212
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,2,64,power_law_1.01,0.01529728055000305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,2,64,power_law_1.01,0.015256320238113403
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,2,64,power_law_1.01,0.015329279899597167
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,2,64,power_law_1.01,0.015267839431762695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,2,64,power_law_1.01,0.014970879554748535
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,2,64,power_law_1.01,0.015298559665679931
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,2,64,power_law_1.01,0.014955519437789916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,2,64,balanced,0.02126336097717285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,2,64,power_law_1.01,0.01495296001434326
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,2,64,power_law_1.01,0.015356160402297975
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,2,64,power_law_1.01,0.01787775993347168
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,2,64,power_law_1.01,0.016117759943008424
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,2,64,power_law_1.01,0.016588799953460694
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,64,power_law_1.2,0.017799680233001706
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,64,power_law_1.2,0.01901952028274536
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,64,power_law_1.2,0.0191103994846344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,2,64,power_law_1.01,0.019819519519805907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,64,power_law_1.2,0.02926208019256592
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,64,power_law_1.2,0.03048703908920288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,64,power_law_1.2,0.030903038978576658
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,64,power_law_1.2,0.03175168037414551
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,2,64,power_law_1.01,0.020288000106811525
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,64,power_law_1.2,0.03256704092025757
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,64,power_law_1.2,0.032924160957336426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,64,power_law_1.2,0.0342195200920105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,64,power_law_1.2,0.03502079963684082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,64,power_law_1.2,0.037904639244079594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,64,power_law_1.2,0.03906816005706787
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,64,power_law_1.2,0.0415667200088501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,64,power_law_1.2,0.039888639450073246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,64,power_law_1.2,0.041562881469726566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,64,power_law_1.2,0.04526080131530762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,64,power_law_1.2,0.05521279811859131
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,64,power_law_1.2,0.0686131191253662
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,64,power_law_1.2,0.08742015838623048
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,64,power_law_1.2,0.1150284767150879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,64,power_law_1.2,0.1740518379211426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,64,power_law_1.2,0.22903039932250974
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,64,power_law_1.2,0.35544193267822266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,2,64,power_law_1.01,0.02599551916122437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,64,power_law_1.2,0.02352895975112915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,64,power_law_1.2,0.02352895975112915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,64,power_law_1.2,0.023463680744171142
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,64,power_law_1.2,0.02839423894882202
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,64,power_law_1.2,0.028896000385284427
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,64,power_law_1.2,0.02969856023788452
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,64,power_law_1.2,0.030507519245147708
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,64,power_law_1.2,0.0324838399887085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,64,power_law_1.2,0.03133440017700195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,64,power_law_1.2,0.03132927894592285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,64,power_law_1.2,0.032578558921813966
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,64,power_law_1.2,0.03247616052627563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,64,power_law_1.2,0.033780479431152345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,64,power_law_1.2,0.033480958938598634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,64,power_law_1.2,0.034937601089477535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,64,power_law_1.2,0.03453952074050903
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,64,power_law_1.2,0.03700223922729492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,64,power_law_1.2,0.04034304141998291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,64,power_law_1.2,0.046219520568847657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,64,power_law_1.2,0.062439680099487305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,64,power_law_1.2,0.08303232192993164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,64,power_law_1.2,0.10626432418823242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,64,power_law_1.2,0.14640000343322754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,64,power_law_1.2,0.4612083053588867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,64,power_law_1.2,0.219736328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,64,power_law_1.2,0.4132723236083985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,64,power_law_1.2,0.9342412567138672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,2,64,power_law_1.01,0.030086400508880617
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,2,64,power_law_1.2,0.014923520088195802
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,2,64,power_law_1.2,0.014954240322113038
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,2,64,power_law_1.2,0.014932479858398438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,2,64,power_law_1.2,0.014945280551910401
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,2,64,power_law_1.2,0.014805760383605957
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,2,64,power_law_1.01,0.039659519195556645
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,2,64,power_law_1.2,0.014949120283126833
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,2,64,power_law_1.2,0.014865920543670655
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,2,64,power_law_1.2,0.014931199550628663
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,2,64,power_law_1.2,0.014926079511642456
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,2,64,power_law_1.2,0.014880000352859496
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,2,64,power_law_1.2,0.014869760274887085
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,2,64,power_law_1.2,0.01452415943145752
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,2,64,power_law_1.2,0.015287040472030639
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,2,64,power_law_1.01,0.05094143867492675
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,2,64,power_law_1.2,0.015360000133514403
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,2,64,power_law_1.01,0.06763391971588135
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,64,power_law_1.2,0.31091455459594725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,64,power_law_1.2,0.5719142532348632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,64,power_law_1.2,0.7227839660644532
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,2,64,power_law_1.2,0.01535104036331177
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,2,64,power_law_1.2,0.01655295968055725
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,2,64,power_law_1.2,0.01613312005996704
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,2,64,power_law_1.2,0.015334399938583373
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,2,64,power_law_1.2,0.017390079498291015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,2,64,power_law_1.2,0.022255361080169678
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,4,32,balanced,0.017420799732208253
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,2,64,power_law_1.2,0.019403519630432128
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,4,32,balanced,0.017771519422531128
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,32,balanced,0.016595200300216675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,32,balanced,0.017730560302734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,32,balanced,0.018648320436477663
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,32,balanced,0.029253120422363284
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,32,balanced,0.030064640045166013
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,32,balanced,0.03130496025085449
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,32,balanced,0.031690239906311035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,32,balanced,0.032979199886322016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,32,balanced,0.03330431938171387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,32,balanced,0.03331199884414673
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,32,balanced,0.034167039394378665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,32,balanced,0.03536384105682373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,32,balanced,0.036901121139526365
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,32,balanced,0.0387059211730957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,32,balanced,0.035539200305938726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,32,balanced,0.03870208024978637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,32,balanced,0.04034687995910645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,32,balanced,0.046071038246154786
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,32,balanced,0.05025792121887207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,32,balanced,0.06035327911376953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,32,balanced,0.07654272079467773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,32,balanced,0.1049728012084961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,32,balanced,0.12824576377868652
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,32,balanced,0.1757708740234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,32,balanced,0.023063039779663085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,32,balanced,0.22274303436279297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,32,balanced,0.023137280941009523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,32,balanced,0.022461440563201904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,32,balanced,0.023123199939727786
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,32,balanced,0.023168001174926758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,32,balanced,0.02879103899002075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,32,balanced,0.02888319969177246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,32,balanced,0.030035200119018557
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,32,balanced,0.02969599962234497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,32,balanced,0.03123456001281738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,32,balanced,0.031239678859710695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,32,balanced,0.030882558822631835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,32,balanced,0.030836479663848875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,32,balanced,0.0321612811088562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,2,64,power_law_1.2,0.02742784023284912
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,32,balanced,0.033070080280303955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,32,balanced,0.0345523190498352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,32,balanced,0.030539519786834717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,32,balanced,0.032889599800109866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,32,balanced,0.03457407951354981
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,32,balanced,0.0374015998840332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,32,balanced,0.04029952049255371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,32,balanced,0.04567679882049561
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,32,balanced,0.05218688011169433
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,32,balanced,0.0728768014907837
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,32,balanced,0.08998271942138672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,32,balanced,0.12467200279235839
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,32,balanced,0.1583296012878418
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,2,64,power_law_1.2,0.03904000043869019
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,2,64,power_law_1.2,0.03171711921691894
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,2,64,power_law_1.2,0.05089791774749756
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,2,64,power_law_1.2,0.09807616233825683
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,4,32,balanced,0.014945280551910401
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,4,32,balanced,0.015278079509735107
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,4,32,balanced,0.014949120283126833
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,4,32,balanced,0.014935040473937988
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,4,32,balanced,0.015333119630813598
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,4,32,balanced,0.01534208059310913
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,4,32,balanced,0.015269119739532471
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,4,32,balanced,0.014897919893264772
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,4,32,balanced,0.01527168035507202
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,4,32,balanced,0.015380480289459228
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,4,32,balanced,0.01534208059310913
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,4,32,balanced,0.015805439949035643
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,4,32,balanced,0.01566848039627075
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,4,32,balanced,0.015692800283432007
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,4,32,balanced,0.0157695996761322
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,4,32,balanced,0.016176639795303344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,4,32,balanced,0.014132479429244995
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,4,32,balanced,0.014521600008010865
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,4,32,balanced,0.01501952052116394
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,4,32,balanced,0.01529088020324707
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,2,64,power_law_1.2,0.0720358419418335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,4,32,power_law_1.01,0.017812479734420777
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,4,32,power_law_1.01,0.017876479625701904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,32,power_law_1.01,0.018553600311279297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,32,power_law_1.01,0.018702080249786375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,32,power_law_1.01,0.030126080513000485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,32,power_law_1.01,0.030472960472106937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,4,32,balanced,0.015292160511016844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,32,power_law_1.01,0.031315200328826905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,32,power_law_1.01,0.03299328088760376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,32,power_law_1.01,0.032460799217224126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,32,power_law_1.01,0.03455359935760498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,32,power_law_1.01,0.03585792064666748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,32,power_law_1.01,0.03694720029830932
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,32,power_law_1.01,0.037447679042816165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,32,power_law_1.01,0.03987711906433106
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,32,power_law_1.01,0.03868799924850464
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,32,power_law_1.01,0.0406924819946289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,32,power_law_1.01,0.041969919204711915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,32,power_law_1.01,0.04749311923980713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,32,power_law_1.01,0.06493184089660645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,32,power_law_1.01,0.0644697618484497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,32,power_law_1.01,0.10408960342407228
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,32,power_law_1.01,0.12047488212585448
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,32,power_law_1.01,0.15840895652770998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,32,power_law_1.01,0.23790847778320315
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,32,power_law_1.01,0.3456473541259766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,32,power_law_1.01,0.40081153869628905
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,4,32,balanced,0.01739392042160034
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,32,power_law_1.01,0.022749440670013426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,32,power_law_1.01,0.02354048013687134
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,32,power_law_1.01,0.569958381652832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,32,power_law_1.01,0.02308864116668701
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,32,power_law_1.01,0.028488960266113285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,32,power_law_1.01,0.02846208095550537
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,32,power_law_1.01,0.028825600147247315
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,32,power_law_1.01,0.0312550401687622
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,4,32,balanced,0.017776639461517335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,32,power_law_1.01,0.03049344062805176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,32,power_law_1.01,0.03133055925369262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,32,power_law_1.01,0.03209728002548218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,32,power_law_1.01,0.03384320020675659
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,32,power_law_1.01,0.03615231990814209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,32,power_law_1.01,0.03299200057983399
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,32,power_law_1.01,0.03253632068634033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,32,power_law_1.01,0.03468415975570679
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,32,power_law_1.01,0.03683327913284302
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,32,power_law_1.01,0.041414399147033695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,32,power_law_1.01,0.044766721725463865
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,4,32,balanced,0.019872000217437742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,32,power_law_1.01,0.05391871929168701
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,32,power_law_1.01,0.08578047752380372
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,32,power_law_1.01,0.11464575767517089
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,32,power_law_1.01,0.030693120956420898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,32,power_law_1.01,0.1664409637451172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,32,power_law_1.01,0.22758272171020505
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,4,32,power_law_1.01,0.015306240320205689
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,32,power_law_1.01,0.27585792541503906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,4,32,power_law_1.01,0.014935040473937988
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,4,32,power_law_1.01,0.015304960012435915
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,4,32,power_law_1.01,0.014905600547790526
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,4,32,power_law_1.01,0.015281920433044433
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,4,32,power_law_1.01,0.0153766405582428
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,4,32,power_law_1.01,0.01531391978263855
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,4,32,power_law_1.01,0.0153164803981781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,32,power_law_1.01,0.05953023910522461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,4,32,power_law_1.01,0.015360000133514403
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,4,32,power_law_1.01,0.015352319478988647
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,4,32,power_law_1.01,0.01539072036743164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,4,32,power_law_1.01,0.015353599786758423
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,4,32,power_law_1.01,0.015801600217819213
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,4,32,power_law_1.01,0.015901440382003786
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,4,32,power_law_1.01,0.015737600326538086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,4,32,power_law_1.01,0.016531200408935548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,32,power_law_1.01,0.4341363143920899
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,4,32,balanced,0.021116158962249755
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,4,32,balanced,0.02438271999359131
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,4,32,balanced,0.027051520347595216
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,4,32,power_law_1.01,0.015508480072021486
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,4,32,power_law_1.01,0.018237440586090087
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,4,32,power_law_1.01,0.019514880180358886
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,4,32,power_law_1.01,0.017538559436798096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,4,32,power_law_1.2,0.019813120365142822
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,4,32,power_law_1.2,0.018228479623794556
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,32,power_law_1.2,0.018177280426025393
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,4,32,power_law_1.01,0.02557823896408081
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,32,power_law_1.2,0.028495359420776366
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,32,power_law_1.2,0.029748480319976806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,32,power_law_1.2,0.030933759212493896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,32,power_law_1.2,0.03209088087081909
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,32,power_law_1.2,0.03290240049362182
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,32,power_law_1.2,0.03373568058013916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,4,32,power_law_1.01,0.02022655963897705
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,4,32,power_law_1.01,0.02926208019256592
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,32,power_law_1.2,0.03462912082672119
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,32,power_law_1.2,0.03682559967041016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,32,power_law_1.2,0.04034687995910645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,32,power_law_1.2,0.04117248058319092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,32,power_law_1.2,0.03870975971221924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,32,power_law_1.2,0.038956799507141114
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,32,power_law_1.2,0.04029056072235108
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,32,power_law_1.2,0.04442240238189697
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,32,power_law_1.2,0.06471168041229249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,32,power_law_1.2,0.07227263927459718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,32,power_law_1.2,0.08306303977966309
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,32,power_law_1.2,0.12130047798156737
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,32,power_law_1.2,0.14906880378723145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,32,power_law_1.2,0.20812288284301758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,32,power_law_1.2,0.3075212860107422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,32,power_law_1.2,0.42987648010253904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,32,power_law_1.2,0.022771201133728027
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,32,power_law_1.2,0.023141119480133057
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,32,power_law_1.2,0.023160319328308105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,32,power_law_1.2,0.02823296070098877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,32,power_law_1.2,0.028430080413818358
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,32,power_law_1.2,0.030008320808410642
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,32,power_law_1.2,0.03051136016845703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,32,power_law_1.2,0.03215487957000732
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,32,power_law_1.2,0.0308735990524292
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,32,power_law_1.2,0.031718399524688724
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,32,power_law_1.2,0.033323519229888916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,32,power_law_1.2,0.035825920104980466
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,32,power_law_1.2,0.033712639808654785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,32,power_law_1.2,0.03335551977157593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,32,power_law_1.2,0.03469696044921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,32,power_law_1.2,0.03543040037155151
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,32,power_law_1.2,0.03742464065551758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,32,power_law_1.2,0.04188416004180908
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,32,power_law_1.2,0.04943359851837158
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,32,power_law_1.2,0.059599361419677734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,32,power_law_1.2,0.0787827205657959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,32,power_law_1.2,0.11908351898193359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,32,power_law_1.2,0.13365376472473145
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,4,32,power_law_1.01,0.03582463979721069
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,32,power_law_1.2,0.2921126365661621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,32,power_law_1.2,0.20924032211303709
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,4,32,power_law_1.2,0.015360000133514403
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,4,32,power_law_1.2,0.01533951997756958
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,4,32,power_law_1.2,0.014941439628601075
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,4,32,power_law_1.2,0.015278079509735107
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,4,32,power_law_1.2,0.015249919891357423
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,4,32,power_law_1.2,0.015322879552841187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,4,32,power_law_1.2,0.015260159969329834
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,4,32,power_law_1.2,0.015412479639053345
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,4,32,power_law_1.2,0.015344640016555786
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,4,32,power_law_1.2,0.015664639472961425
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,4,32,power_law_1.2,0.015338239669799806
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,4,32,power_law_1.2,0.015738879442214963
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,4,32,power_law_1.2,0.015371520519256592
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,4,32,power_law_1.2,0.015742720365524293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,4,32,power_law_1.01,0.044313597679138186
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,4,32,power_law_1.01,0.0660595178604126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,32,power_law_1.2,0.8237375640869141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,32,power_law_1.2,0.45612415313720706
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,4,32,power_law_1.01,0.07404416084289552
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,32,power_law_1.2,0.5912819290161133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,32,power_law_1.2,0.6331084823608398
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,4,32,power_law_1.2,0.016227840185165404
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,4,32,power_law_1.2,0.016563199758529663
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,4,32,power_law_1.2,0.018558720350265505
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,4,32,power_law_1.2,0.01945088028907776
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,4,32,power_law_1.2,0.021447679996490478
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,4,32,power_law_1.2,0.01739776015281677
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,128,balanced,0.027184638977050778
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,4,32,power_law_1.2,0.024156160354614258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,128,balanced,0.027569921016693116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,128,balanced,0.02799488067626953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,128,balanced,0.027242240905761717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,128,balanced,0.02807039976119995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,128,balanced,0.048081917762756346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,128,balanced,0.04927616119384766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,128,balanced,0.05094016075134278
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,128,balanced,0.0522163200378418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,128,balanced,0.05338111877441406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,128,balanced,0.053812479972839354
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,128,balanced,0.05501823902130128
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,128,balanced,0.0551142406463623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,128,balanced,0.056572160720825194
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,128,balanced,0.057894401550292976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,128,balanced,0.06078335762023925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,128,balanced,0.06201087951660156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,128,balanced,0.06097536087036133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,128,balanced,0.06329855918884278
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,128,balanced,0.07080448150634766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,128,balanced,0.06746623992919923
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,128,balanced,0.075764479637146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,128,balanced,0.09710847854614257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,128,balanced,0.11584128379821776
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,128,balanced,0.1320576000213623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,128,balanced,0.16418304443359374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,128,balanced,0.03049855947494507
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,128,balanced,0.1999577522277832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,128,balanced,0.030494720935821534
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,128,balanced,0.02971519947052002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,128,balanced,0.030521600246429442
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,128,balanced,0.02964479923248291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,128,balanced,0.037419519424438476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,128,balanced,0.03767296075820923
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,128,balanced,0.038239998817443846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,128,balanced,0.0382476806640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,128,balanced,0.03866624116897583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,128,balanced,0.0381657600402832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,128,balanced,0.03964287996292114
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,128,balanced,0.03916800022125244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,128,balanced,0.040770559310913085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,128,balanced,0.04034815788269043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,128,balanced,0.0431334400177002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,128,balanced,0.042725119590759274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,128,balanced,0.0407539176940918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,128,balanced,0.04239359855651855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,128,balanced,0.04516863822937012
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,128,balanced,0.043208961486816404
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,4,32,power_law_1.2,0.02880768060684204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,128,balanced,0.050067200660705566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,128,balanced,0.05323264122009277
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,128,balanced,0.06012544155120849
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,128,balanced,0.07002495765686036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,128,balanced,0.08727168083190918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,128,balanced,0.10961152076721192
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,4,32,power_law_1.2,0.03690880060195923
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,4,32,power_law_1.2,0.0643455982208252
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,4,32,power_law_1.2,0.05023104190826416
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,4,32,power_law_1.2,0.0751692819595337
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,1,128,balanced,0.016316159963607788
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,1,128,balanced,0.016743680238723753
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,1,128,balanced,0.016606719493865968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,1,128,balanced,0.0165721595287323
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,1,128,balanced,0.016574720144271853
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,1,128,balanced,0.016495360136032106
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,1,128,balanced,0.016593919992446898
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,1,128,balanced,0.016208640336990356
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,1,128,balanced,0.01617535948753357
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,1,128,balanced,0.016574720144271853
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,1,128,balanced,0.016599040031433105
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,1,128,balanced,0.016522239446640014
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,1,128,balanced,0.016828160285949707
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,1,128,balanced,0.016592639684677123
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,1,128,balanced,0.016641279458999632
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,1,128,balanced,0.016811519861221313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,1,128,balanced,0.017002240419387818
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,1,128,balanced,0.016125439405441283
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,1,128,balanced,0.016212480068206785
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,1,128,balanced,0.016284159421920776
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,1,128,balanced,0.01619328022003174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,128,power_law_1.01,0.026981120109558103
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,4,32,power_law_1.2,0.10797568321228028
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,128,power_law_1.01,0.02725503921508789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,128,power_law_1.01,0.048912639617919924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,128,power_law_1.01,0.04928256034851074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,128,power_law_1.01,0.05259263992309571
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,128,power_law_1.01,0.05221375942230224
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,128,power_law_1.01,0.054453759193420416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,128,power_law_1.01,0.055829758644104
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,128,power_law_1.01,0.05666431903839111
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,128,power_law_1.01,0.05797887802124023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,128,power_law_1.01,0.05935616016387939
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,128,power_law_1.01,0.06133759975433349
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,128,power_law_1.01,0.0639577579498291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,128,power_law_1.01,0.0695091199874878
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,128,power_law_1.01,0.07338880062103272
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,128,power_law_1.01,0.0729036808013916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,128,power_law_1.01,0.09368063926696776
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,128,power_law_1.01,0.11436927795410157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,128,power_law_1.01,0.12075136184692384
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,128,power_law_1.01,0.14988415718078613
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,128,power_law_1.01,0.19552383422851563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,128,power_law_1.01,0.27715967178344725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,128,power_law_1.01,0.3571558380126953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,128,power_law_1.01,0.539051513671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,128,power_law_1.01,0.7534079742431641
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,1,128,balanced,0.01784832000732422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,128,power_law_1.01,1.2273049926757813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,128,power_law_1.01,0.030512640476226805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,1,128,balanced,0.019011839628219604
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,1,128,balanced,0.017437440156936646
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,128,power_law_1.01,1.7315367126464842
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,128,power_law_1.01,0.03792639970779419
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,128,power_law_1.01,0.038229761123657224
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,128,power_law_1.01,0.037914879322052
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,128,power_law_1.01,0.03865344047546387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,128,power_law_1.01,0.03977983951568603
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,128,power_law_1.01,0.03908992052078247
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,128,power_law_1.01,0.030526719093322753
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,128,power_law_1.01,0.03990015983581543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,128,power_law_1.01,0.040398077964782717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,128,power_law_1.01,0.04077824115753174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,128,power_law_1.01,0.041579518318176266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,128,power_law_1.01,0.04347904205322266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,128,power_law_1.01,0.04602240085601807
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,128,power_law_1.01,0.048115200996398926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,128,power_law_1.01,0.048183040618896486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,128,power_law_1.01,0.03043584108352661
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,128,power_law_1.01,0.05841663837432861
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,128,power_law_1.01,0.06436223983764648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,128,power_law_1.01,0.07836287975311279
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,128,power_law_1.01,0.09633664131164552
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,128,power_law_1.01,0.14294272422790527
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,128,power_law_1.01,0.17817855834960938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,128,power_law_1.01,0.26349184036254886
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,128,power_law_1.01,0.36993793487548826
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,1,128,power_law_1.01,0.01659775972366333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,128,power_law_1.01,0.5883020782470704
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,1,128,power_law_1.01,0.01651584029197693
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,1,128,power_law_1.01,0.016542719602584837
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,1,128,power_law_1.01,0.016591360569000242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,128,power_law_1.01,0.8041305541992188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,1,128,power_law_1.01,0.016567039489746093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,1,128,power_law_1.01,0.01650048017501831
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,1,128,power_law_1.01,0.01649407982826233
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,1,128,power_law_1.01,0.016523519754409792
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,1,128,power_law_1.01,0.016640000343322754
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,1,128,power_law_1.01,0.016532479524612426
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,1,128,power_law_1.01,0.01655295968055725
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,1,128,power_law_1.01,0.01665279984474182
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,1,128,power_law_1.01,0.016113920211791994
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,1,128,power_law_1.01,0.016641279458999632
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,1,128,balanced,0.02027776002883911
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,1,128,balanced,0.023074560165405274
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,1,128,balanced,0.027980799674987795
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,1,128,power_law_1.01,0.016131839752197265
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,1,128,power_law_1.01,0.01780351996421814
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,1,128,power_law_1.01,0.01712000012397766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,1,128,power_law_1.01,0.01989375948905945
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,1,128,power_law_1.01,0.018627840280532836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,1,128,power_law_1.01,0.022716159820556643
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,128,power_law_1.2,0.028884479999542235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,128,power_law_1.2,0.028852479457855223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,128,power_law_1.2,0.049015040397644045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,128,power_law_1.2,0.05138432025909424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,128,power_law_1.2,0.05462656021118164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,128,power_law_1.2,0.05437312126159668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,128,power_law_1.2,0.05710207939147949
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,128,power_law_1.2,0.05885695934295654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,128,power_law_1.2,0.06138368129730225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,128,power_law_1.2,0.061839361190795895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,128,power_law_1.2,0.06086912155151367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,128,power_law_1.2,0.06629119873046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,128,power_law_1.2,0.07055615901947021
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,128,power_law_1.2,0.07797887802124023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,128,power_law_1.2,0.08632960319519042
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,128,power_law_1.2,0.09837056159973144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,128,power_law_1.2,0.11268735885620118
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,1,128,power_law_1.01,0.027589120864868165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,128,power_law_1.2,0.12523776054382324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,128,power_law_1.2,0.1520729637145996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,128,power_law_1.2,0.21895296096801758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,128,power_law_1.2,0.2767091178894043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,1,128,power_law_1.01,0.033859839439392095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,128,power_law_1.2,0.48765441894531253
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,128,power_law_1.2,0.9261325073242187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,128,power_law_1.2,0.6456217956542969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,128,power_law_1.2,1.2963250732421874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,128,power_law_1.2,0.03006079912185669
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,128,power_law_1.2,0.030389759540557858
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,128,power_law_1.2,0.03705471992492676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,128,power_law_1.2,0.038037760257720946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,128,power_law_1.2,0.03891455888748169
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,128,power_law_1.2,0.03911295890808105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,128,power_law_1.2,0.04033535957336426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,128,power_law_1.2,0.04146944046020508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,128,power_law_1.2,0.042873601913452145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,128,power_law_1.2,0.043587841987609864
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,128,power_law_1.2,0.040641279220581056
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,128,power_law_1.2,0.04191872119903565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,128,power_law_1.2,1.9488780212402343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,128,power_law_1.2,0.04399487972259521
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,128,power_law_1.2,0.04650879859924316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,128,power_law_1.2,0.047713279724121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,128,power_law_1.2,0.04860288143157959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,128,power_law_1.2,0.057114877700805665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,128,power_law_1.2,0.06654592037200928
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,128,power_law_1.2,0.0787391996383667
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,128,power_law_1.2,0.10146559715270995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,128,power_law_1.2,0.15016703605651854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,128,power_law_1.2,0.2747737693786621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,128,power_law_1.2,0.2292185592651367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,128,power_law_1.2,0.4702272033691406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,128,power_law_1.2,0.9148799896240234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,1,128,power_law_1.01,0.04178431987762451
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,1,128,power_law_1.2,0.01613312005996704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,128,power_law_1.2,0.6125875091552734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,1,128,power_law_1.2,0.016122879981994628
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,1,128,power_law_1.2,0.016217600107192993
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,1,128,power_law_1.2,0.016161279678344725
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,1,128,power_law_1.01,0.05675136089324952
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,1,128,power_law_1.2,0.016240639686584474
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,1,128,power_law_1.2,0.016190719604492188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,1,128,power_law_1.2,0.016606719493865968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,1,128,power_law_1.2,0.016450560092926024
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,1,128,power_law_1.2,0.016565760374069212
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,1,128,power_law_1.2,0.01657855987548828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,1,128,power_law_1.2,0.016579840183258057
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,1,128,power_law_1.2,0.016586240530014038
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,1,128,power_law_1.2,0.016999679803848266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,1,128,power_law_1.2,0.017349120378494263
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,1,128,power_law_1.01,0.10724736213684083
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,1,128,power_law_1.2,0.01650431990623474
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,1,128,power_law_1.01,0.07327744007110595
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,1,128,power_law_1.01,0.14151552200317383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,128,power_law_1.2,1.259736328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,128,power_law_1.2,2.7354443359375002
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,1,128,power_law_1.2,0.01822975993156433
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,1,128,power_law_1.2,0.020614399909973144
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,1,128,power_law_1.2,0.0174835205078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,1,128,power_law_1.2,0.022036480903625488
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,1,128,power_law_1.2,0.0305459189414978
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,64,balanced,0.02756736040115356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,64,balanced,0.028017919063568115
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,1,128,power_law_1.2,0.03682687997817993
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,64,balanced,0.028015360832214353
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,64,balanced,0.02525439977645874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,64,balanced,0.025626881122589113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,64,balanced,0.04397823810577393
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,64,balanced,0.04567679882049561
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,64,balanced,0.04658688068389892
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,64,balanced,0.04873216152191162
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,64,balanced,0.04991615772247314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,64,balanced,0.04987135887145996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,64,balanced,0.05304319858551025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,64,balanced,0.05216127872467041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,64,balanced,0.05589375972747803
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,64,balanced,0.05388800144195557
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,64,balanced,0.05478528022766114
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,64,balanced,0.058234882354736325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,64,balanced,0.06296063899993896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,64,balanced,0.06120704174041748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,64,balanced,0.0682086420059204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,64,balanced,0.07299327850341797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,64,balanced,0.09694335937500001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,64,balanced,0.11462783813476562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,64,balanced,0.1406886386871338
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,64,balanced,0.16338815689086914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,64,balanced,0.20980480194091794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,64,balanced,0.2664627265930176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,64,balanced,0.028010239601135255
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,1,128,power_law_1.2,0.049625601768493656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,64,balanced,0.029646079540252685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,64,balanced,0.027924480438232424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,64,balanced,0.02799232006072998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,64,balanced,0.028446719646453855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,64,balanced,0.03517184019088745
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,64,balanced,0.036252160072326664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,64,balanced,0.03586816072463989
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,64,balanced,0.03623296022415161
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,64,balanced,0.037884159088134764
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,64,balanced,0.03667327880859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,64,balanced,0.03825279951095581
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,64,balanced,0.040755200386047366
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,64,balanced,0.03739392042160034
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,64,balanced,0.03849855899810791
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,64,balanced,0.038618879318237306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,64,balanced,0.03952896118164063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,64,balanced,0.04240640163421631
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,64,balanced,0.040006399154663086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,64,balanced,0.04363135814666748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,64,balanced,0.04775424003601074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,64,balanced,0.055092477798461915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,64,balanced,0.06138879776000976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,64,balanced,0.07665535926818848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,64,balanced,0.09174912452697755
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,64,balanced,0.12284159660339355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,64,balanced,0.15565695762634277
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,1,128,power_law_1.2,0.06703360080718994
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,1,128,power_law_1.2,0.09436927795410156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,2,64,balanced,0.01651968002319336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,2,64,balanced,0.01651584029197693
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,2,64,balanced,0.016724480390548708
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,2,64,balanced,0.016593919992446898
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,2,64,balanced,0.016953599452972413
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,2,64,balanced,0.016577279567718504
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,2,64,balanced,0.016569600105285645
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,2,64,balanced,0.016919039487838745
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,2,64,balanced,0.016547839641571045
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,2,64,balanced,0.016418559551239015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,2,64,balanced,0.016976640224456788
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,2,64,balanced,0.016576000452041627
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,2,64,balanced,0.0170035195350647
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,2,64,balanced,0.017018879652023318
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,2,64,balanced,0.017396479845046997
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,2,64,balanced,0.017390079498291015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,2,64,balanced,0.01731711983680725
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,2,64,balanced,0.01731328010559082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,2,64,balanced,0.016549119949340822
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,1,128,power_law_1.2,0.1190976047515869
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,2,64,balanced,0.0160806405544281
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,1,128,power_law_1.2,0.17380352020263673
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,2,64,balanced,0.01651968002319336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,64,power_law_1.01,0.02554624080657959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,64,power_law_1.01,0.025550079345703126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,64,power_law_1.01,0.02601344108581543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,64,power_law_1.01,0.043970561027526854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,64,power_law_1.01,0.04640511989593506
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,64,power_law_1.01,0.048334078788757326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,64,power_law_1.01,0.050238718986511235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,64,power_law_1.01,0.051404800415039066
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,2,64,balanced,0.019037439823150634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,64,power_law_1.01,0.05418496131896973
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,64,power_law_1.01,0.056892161369323735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,64,power_law_1.01,0.05697792053222657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,64,power_law_1.01,0.059852800369262694
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,2,64,balanced,0.019466240406036377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,64,power_law_1.01,0.06521599769592286
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,64,power_law_1.01,0.05314559936523437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,64,power_law_1.01,0.06852352142333984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,64,power_law_1.01,0.06744319915771484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,64,power_law_1.01,0.08855680465698243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,64,power_law_1.01,0.10856191635131836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,64,power_law_1.01,0.12108927726745607
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,64,power_law_1.01,0.1488499164581299
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,64,power_law_1.01,0.17241983413696288
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,2,64,balanced,0.021305599212646485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,64,power_law_1.01,0.25892480850219723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,64,power_law_1.01,0.2830271911621094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,64,power_law_1.01,0.4857382583618164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,64,power_law_1.01,0.7086156463623047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,64,power_law_1.01,0.028343040943145752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,64,power_law_1.01,0.9207692718505859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,64,power_law_1.01,0.0288972806930542
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,64,power_law_1.01,0.02846848011016846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,64,power_law_1.01,0.0352128005027771
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,64,power_law_1.01,0.03580800056457519
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,64,power_law_1.01,0.03668479919433594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,64,power_law_1.01,0.039502079486846926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,64,power_law_1.01,0.03740031957626343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,64,power_law_1.01,0.03786112070083618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,64,power_law_1.01,0.038616321086883544
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,64,power_law_1.01,0.04000256061553955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,64,power_law_1.01,0.042007040977478025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,64,power_law_1.01,0.043935999870300294
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,64,power_law_1.01,0.043171839714050295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,64,power_law_1.01,0.04547584056854248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,64,power_law_1.01,0.049584641456604
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,64,power_law_1.01,0.05858176231384278
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,64,power_law_1.01,1.2472972869873047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,64,power_law_1.01,0.06616703987121582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,64,power_law_1.01,0.03872512102127075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,64,power_law_1.01,0.0847219181060791
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,2,64,balanced,0.022726399898529055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,64,power_law_1.01,0.09613823890686035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,64,power_law_1.01,0.1316748809814453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,64,power_law_1.01,0.15942655563354494
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,64,power_law_1.01,0.26511999130249025
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,2,64,power_law_1.01,0.016517119407653807
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,64,power_law_1.01,0.2973734474182129
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,2,64,power_law_1.01,0.016573439836502075
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,2,64,power_law_1.01,0.016622079610824587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,64,power_law_1.01,0.5584332656860351
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,2,64,power_law_1.01,0.016583679914474486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,64,power_law_1.01,0.6255270385742188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,2,64,power_law_1.01,0.016915199756622316
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,2,64,power_law_1.01,0.016558079719543456
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,2,64,power_law_1.01,0.016936960220336913
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,2,64,power_law_1.01,0.016588799953460694
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,2,64,power_law_1.01,0.01733888030052185
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,2,64,power_law_1.01,0.017144320011138917
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,1,128,power_law_1.2,0.23428224563598632
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,2,64,power_law_1.01,0.017374720573425293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,2,64,power_law_1.01,0.0173363196849823
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,2,64,power_law_1.01,0.017349120378494263
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,2,64,power_law_1.01,0.016584960222244264
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,2,64,balanced,0.027243518829345705
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,2,64,balanced,0.03337728023529053
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,2,64,power_law_1.01,0.016929279565811157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,2,64,power_law_1.01,0.020252161026000977
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,2,64,power_law_1.01,0.018516479730606078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,2,64,power_law_1.01,0.01818112015724182
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,2,64,power_law_1.01,0.020625920295715333
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,2,64,power_law_1.01,0.02351871967315674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,64,power_law_1.2,0.027173120975494385
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,64,power_law_1.2,0.026343679428100585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,64,power_law_1.2,0.043196158409118654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,64,power_law_1.2,0.0452620792388916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,64,power_law_1.2,0.049122557640075684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,64,power_law_1.2,0.049237761497497555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,64,power_law_1.2,0.05246592044830323
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,64,power_law_1.2,0.05457920074462891
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,2,64,power_law_1.01,0.025172479152679443
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,64,power_law_1.2,0.05730688095092773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,64,power_law_1.2,0.056134400367736814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,64,power_law_1.2,0.05974143981933594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,64,power_law_1.2,0.06189568042755127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,64,power_law_1.2,0.0645363187789917
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,64,power_law_1.2,0.0747379207611084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,64,power_law_1.2,0.08820992469787597
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,64,power_law_1.2,0.057287678718566895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,64,power_law_1.2,0.09692288398742675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,64,power_law_1.2,0.11580544471740724
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,64,power_law_1.2,0.14322943687438966
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,64,power_law_1.2,0.19090816497802734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,64,power_law_1.2,0.25555967330932616
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,64,power_law_1.2,0.3687795257568359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,64,power_law_1.2,0.4637542343139648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,64,power_law_1.2,0.9873766326904297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,2,64,power_law_1.01,0.03534208059310913
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,64,power_law_1.2,0.71897216796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,64,power_law_1.2,0.02837503910064697
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,64,power_law_1.2,0.02841599941253662
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,64,power_law_1.2,0.035281920433044435
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,64,power_law_1.2,0.03549184083938599
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,64,power_law_1.2,0.03665920019149781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,64,power_law_1.2,0.03788928031921387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,64,power_law_1.2,0.03961215972900391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,64,power_law_1.2,0.040320000648498534
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,64,power_law_1.2,0.04074880123138428
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,64,power_law_1.2,0.03911168098449707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,64,power_law_1.2,0.039933440685272215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,64,power_law_1.2,0.041567997932434084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,64,power_law_1.2,0.04280576229095459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,64,power_law_1.2,0.044025602340698244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,64,power_law_1.2,0.04476287841796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,64,power_law_1.2,1.3794175720214843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,64,power_law_1.2,0.04484479904174805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,64,power_law_1.2,0.051720957756042484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,64,power_law_1.2,0.06425471782684326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,64,power_law_1.2,0.06654079914093017
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,64,power_law_1.2,0.0881663990020752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,64,power_law_1.2,0.12299391746520996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,64,power_law_1.2,0.18919296264648439
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,64,power_law_1.2,0.24846847534179686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,64,power_law_1.2,0.5155814361572266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,64,power_law_1.2,0.3738431930541992
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,2,64,power_law_1.01,0.03765376091003418
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,2,64,power_law_1.2,0.016940799951553346
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,2,64,power_law_1.2,0.016508159637451173
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,2,64,power_law_1.2,0.016583679914474486
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,2,64,power_law_1.2,0.016592639684677123
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,2,64,power_law_1.2,0.016904959678649904
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,2,64,power_law_1.01,0.058616318702697755
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,2,64,power_law_1.2,0.01698559999465942
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,2,64,power_law_1.2,0.016919039487838745
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,2,64,power_law_1.2,0.016588799953460694
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,2,64,power_law_1.2,0.016609280109405516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,2,64,power_law_1.2,0.01706496000289917
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,2,64,power_law_1.01,0.06821375846862793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,2,64,power_law_1.2,0.017007360458374025
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,2,64,power_law_1.2,0.017134079933166506
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,2,64,power_law_1.2,0.017004799842834473
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,2,64,power_law_1.2,0.018549760580062864
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,2,64,power_law_1.01,0.09988224029541015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,2,64,power_law_1.2,0.020954880714416504
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,2,64,power_law_1.01,0.12249088287353516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,64,power_law_1.2,0.7857766723632812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,2,64,power_law_1.2,0.018287359476089476
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,2,64,power_law_1.2,0.01734015941619873
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,2,64,power_law_1.2,0.01897088050842285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,2,64,power_law_1.2,0.021594879627227784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,64,power_law_1.2,0.9460454559326171
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,2,64,power_law_1.2,0.029816319942474367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,64,power_law_1.2,1.904040985107422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,32,balanced,0.022722558975219725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,32,balanced,0.022758400440216063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,32,balanced,0.02231679916381836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,32,balanced,0.02484607934951782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,32,balanced,0.026397440433502194
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,32,balanced,0.04420351982116699
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,32,balanced,0.046045441627502445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,32,balanced,0.047375359535217286
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,32,balanced,0.04863103866577149
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,32,balanced,0.049779200553894044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,32,balanced,0.050621438026428225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,32,balanced,0.05103744029998779
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,32,balanced,0.05266047954559326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,32,balanced,0.054503679275512695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,32,balanced,0.057136640548706055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,32,balanced,0.058434557914733884
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,32,balanced,0.05719935894012451
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,32,balanced,0.06352511882781983
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,32,balanced,0.06964352130889892
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,32,balanced,0.08935168266296387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,32,balanced,0.10493311882019044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,32,balanced,0.12844927787780763
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,32,balanced,0.14882431983947755
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,32,balanced,0.19072511672973633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,2,64,power_law_1.2,0.032565760612487796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,32,balanced,0.2315443229675293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,32,balanced,0.31457279205322264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,32,balanced,0.026757121086120605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,32,balanced,0.39939456939697265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,32,balanced,0.026726400852203368
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,32,balanced,0.02651263952255249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,32,balanced,0.027964160442352292
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,32,balanced,0.02927615880966187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,32,balanced,0.034603519439697264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,32,balanced,0.03460736036300659
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,32,balanced,0.03660928010940552
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,32,balanced,0.03544575929641723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,32,balanced,0.037829120159149174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,32,balanced,0.03660928010940552
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,32,balanced,0.03665791988372803
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,32,balanced,0.03739775896072388
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,32,balanced,0.038675200939178464
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,32,balanced,0.04075007915496826
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,32,balanced,0.04277247905731201
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,32,balanced,0.039705600738525394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,32,balanced,0.04325632095336914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,32,balanced,0.04646912097930908
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,32,balanced,0.04954239845275879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,32,balanced,0.05386623859405517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,32,balanced,0.0721779203414917
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,32,balanced,0.08513407707214356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,32,balanced,0.11289471626281737
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,32,balanced,0.14084863662719727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,32,balanced,0.19392255783081053
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,32,balanced,0.2511206436157226
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,2,64,power_law_1.2,0.045821437835693354
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,2,64,power_law_1.2,0.056755199432373046
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,2,64,power_law_1.2,0.08010368347167969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,4,32,balanced,0.01734272003173828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,4,32,balanced,0.016963839530944824
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,4,32,balanced,0.017825280427932737
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,4,32,balanced,0.017802239656448365
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,4,32,balanced,0.01777024030685425
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,4,32,balanced,0.017779200077056884
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,4,32,balanced,0.017811199426651
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,4,32,balanced,0.01821568012237549
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,4,32,balanced,0.01825152039527893
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,4,32,balanced,0.01780608057975769
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,4,32,balanced,0.018050559759140015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,4,32,balanced,0.017917439937591553
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,4,32,balanced,0.018237440586090087
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,4,32,balanced,0.01814911961555481
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,4,32,balanced,0.01867264032363892
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,4,32,balanced,0.01864575982093811
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,4,32,balanced,0.019151359796524048
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,4,32,balanced,0.01904255986213684
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,2,64,power_law_1.2,0.1167193603515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,2,64,power_law_1.2,0.15156864166259765
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,4,32,balanced,0.01739392042160034
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,4,32,balanced,0.017742079496383668
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,4,32,balanced,0.01783679962158203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,32,power_law_1.01,0.025134079456329346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,32,power_law_1.01,0.02595328092575073
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,32,power_law_1.01,0.025647358894348143
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,32,power_law_1.01,0.04243072032928467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,32,power_law_1.01,0.046083841323852535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,32,power_law_1.01,0.04732160091400146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,32,power_law_1.01,0.04976128101348877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,32,power_law_1.01,0.050146560668945316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,32,power_law_1.01,0.05117184162139893
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,32,power_law_1.01,0.05299456119537353
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,32,power_law_1.01,0.053838081359863274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,32,power_law_1.01,0.05640192031860351
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,32,power_law_1.01,0.06016255855560303
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,32,power_law_1.01,0.06729983806610107
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,32,power_law_1.01,0.07551487922668457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,32,power_law_1.01,0.07460351943969726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,32,power_law_1.01,0.08758399963378907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,32,power_law_1.01,0.10041343688964843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,32,power_law_1.01,0.1129702377319336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,32,power_law_1.01,0.14020223617553712
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,32,power_law_1.01,0.16604543685913087
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,32,power_law_1.01,0.22076160430908204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,32,power_law_1.01,0.30891904830932615
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,32,power_law_1.01,0.4752371215820313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,32,power_law_1.01,0.6513740539550781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,32,power_law_1.01,0.8165964508056641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,32,power_law_1.01,1.1590374755859376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,32,power_law_1.01,0.028499200344085696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,32,power_law_1.01,0.028876800537109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,32,power_law_1.01,0.028545279502868653
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,32,power_law_1.01,0.033775360584259034
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,32,power_law_1.01,0.034542078971862795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,32,power_law_1.01,0.03509504079818725
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,4,32,balanced,0.02187391996383667
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,32,power_law_1.01,0.0358515191078186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,32,power_law_1.01,0.03659775972366333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,32,power_law_1.01,0.03658368110656739
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,32,power_law_1.01,0.037402880191802976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,32,power_law_1.01,0.03917567968368531
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,4,32,balanced,0.028487679958343504
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,4,32,balanced,0.022725119590759277
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,2,64,power_law_1.2,0.20966527938842772
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,32,power_law_1.01,0.04117631912231445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,32,power_law_1.01,0.04364160060882568
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,32,power_law_1.01,0.044035840034484866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,32,power_law_1.01,0.043138561248779295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,32,power_law_1.01,0.04487936019897461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,4,32,balanced,0.025399041175842286
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,32,power_law_1.01,0.05261184215545654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,32,power_law_1.01,0.06028031826019287
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,32,power_law_1.01,0.06461952209472657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,32,power_law_1.01,0.08043392181396485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,32,power_law_1.01,0.10641535758972168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,32,power_law_1.01,0.13618432044982912
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,32,power_law_1.01,0.1965056037902832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,32,power_law_1.01,0.27846656799316405
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,4,32,power_law_1.01,0.017745920419692994
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,32,power_law_1.01,0.3626137542724609
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,4,32,power_law_1.01,0.017786879539489747
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,4,32,power_law_1.01,0.017743359804153442
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,4,32,power_law_1.01,0.017405439615249634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,32,power_law_1.01,0.5383884811401367
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,4,32,power_law_1.01,0.017946879863739013
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,4,32,power_law_1.01,0.01773695945739746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,32,power_law_1.01,0.6843302154541016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,4,32,power_law_1.01,0.017745920419692994
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,4,32,power_law_1.01,0.018236160278320312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,4,32,power_law_1.01,0.018137600421905518
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,4,32,power_law_1.01,0.018240000009536746
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,4,32,balanced,0.03731967926025391
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,4,32,power_law_1.01,0.018216960430145264
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,4,32,power_law_1.01,0.019104000329971314
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,4,32,balanced,0.046732797622680664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,4,32,power_law_1.01,0.018161920309066774
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,4,32,power_law_1.01,0.017656320333480836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,4,32,power_law_1.01,0.022238719463348388
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,4,32,power_law_1.01,0.020687360763549805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,4,32,power_law_1.01,0.019385600090026857
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,4,32,power_law_1.01,0.021493759155273438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,4,32,power_law_1.01,0.01820032000541687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,4,32,power_law_1.01,0.027114241123199462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,32,power_law_1.2,0.026443519592285157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,32,power_law_1.2,0.026379520893096926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,32,power_law_1.2,0.025564160346984867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,32,power_law_1.2,0.04264448165893554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,32,power_law_1.2,0.046452479362487795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,32,power_law_1.2,0.0476697587966919
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,32,power_law_1.2,0.04960512161254883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,32,power_law_1.2,0.05053696155548095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,32,power_law_1.2,0.0523967981338501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,32,power_law_1.2,0.054275841712951654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,32,power_law_1.2,0.05846784114837646
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,32,power_law_1.2,0.06038911819458008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,32,power_law_1.2,0.06511616230010986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,32,power_law_1.2,0.06986112117767333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,32,power_law_1.2,0.07601791858673096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,32,power_law_1.2,0.081342716217041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,32,power_law_1.2,0.09574144363403321
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,32,power_law_1.2,0.11673727989196778
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,32,power_law_1.2,0.12609919548034668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,32,power_law_1.2,0.17045888900756836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,32,power_law_1.2,0.2181644821166992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,32,power_law_1.2,0.31495807647705076
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,32,power_law_1.2,0.3816166305541992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,32,power_law_1.2,0.585346565246582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,32,power_law_1.2,0.7943193817138672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,4,32,power_law_1.01,0.03027456045150757
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,4,32,power_law_1.01,0.05109248161315918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,4,32,power_law_1.01,0.06784512042999268
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,32,power_law_1.2,0.028386559486389164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,32,power_law_1.2,0.028965120315551755
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,32,power_law_1.2,0.029022719860076908
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,4,32,power_law_1.01,0.07698304176330566
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,4,32,power_law_1.01,0.10877696037292481
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,32,power_law_1.2,0.035016961097717285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,4,32,power_law_1.01,0.038679039478302
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,32,power_law_1.2,0.03419264078140259
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,32,power_law_1.2,0.035024640560150144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,32,power_law_1.2,0.03626879930496216
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,32,power_law_1.2,0.03707776069641113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,32,power_law_1.2,0.03790848016738892
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,32,power_law_1.2,0.0399232006072998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,32,power_law_1.2,0.042766079902648926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,32,power_law_1.2,0.0440064001083374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,32,power_law_1.2,0.04353407859802246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,32,power_law_1.2,0.044697599411010744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,32,power_law_1.2,0.04613376140594482
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,32,power_law_1.2,0.05363455772399902
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,32,power_law_1.2,0.0631001615524292
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,4,32,power_law_1.01,0.13815296173095704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,32,power_law_1.2,0.07332352161407471
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,32,power_law_1.2,0.09331711769104004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,32,power_law_1.2,0.13037311553955078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,32,power_law_1.2,0.17331584930419922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,32,power_law_1.2,1.1471257781982422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,32,power_law_1.2,0.24895744323730468
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,4,32,power_law_1.2,0.018177280426025393
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,4,32,power_law_1.2,0.017831679582595825
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,32,power_law_1.2,0.4861900711059571
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,4,32,power_law_1.2,0.018184959888458252
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,4,32,power_law_1.2,0.017825280427932737
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,32,power_law_1.2,0.03821952104568481
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,4,32,power_law_1.2,0.01775871992111206
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,4,32,power_law_1.2,0.018222080469131468
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,4,32,power_law_1.2,0.017889280319213868
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,4,32,power_law_1.2,0.01816447973251343
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,4,32,power_law_1.2,0.017793279886245728
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,4,32,power_law_1.2,0.01811840057373047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,4,32,power_law_1.2,0.018261760473251343
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,4,32,power_law_1.2,0.018220800161361694
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,4,32,power_law_1.2,0.018465280532836914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,32,power_law_1.2,0.34281215667724607
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,4,32,power_law_1.2,0.01979647994041443
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,4,32,power_law_1.2,0.020028159618377686
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,4,32,power_law_1.2,0.02519295930862427
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,32,power_law_1.2,0.7126399993896484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,32,power_law_1.2,1.61649658203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,4,32,power_law_1.2,0.021040639877319335
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,4,32,power_law_1.2,0.022279679775238037
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,4,32,power_law_1.2,0.028884479999542235
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,4,32,power_law_1.2,0.03878783941268921
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,4,32,power_law_1.2,0.0194214403629303
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,8,16,balanced,0.026009600162506103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,8,16,balanced,0.02642047882080078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,8,16,balanced,0.023968000411987305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,8,16,balanced,0.0249727988243103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,8,16,balanced,0.028061439990997316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,8,16,balanced,0.04764416217803955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,8,16,balanced,0.0484991979598999
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,8,16,balanced,0.051326718330383304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,8,16,balanced,0.051454720497131345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,8,16,balanced,0.05250944137573242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,8,16,balanced,0.05462912082672119
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,8,16,balanced,0.05663616180419921
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,8,16,balanced,0.05944191932678222
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,8,16,balanced,0.05627520084381103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,8,16,balanced,0.059573760032653814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,8,16,balanced,0.06290048122406006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,8,16,balanced,0.07019264221191406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,8,16,balanced,0.08637824058532714
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,8,16,balanced,0.10070783615112304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,8,16,balanced,0.1254643154144287
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,8,16,balanced,0.1414310359954834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,8,16,balanced,0.18193920135498048
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,8,16,balanced,0.21874303817749025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,8,16,balanced,0.29511680603027346
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,4,32,power_law_1.2,0.050183677673339845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,8,16,balanced,0.37208190917968753
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,8,16,balanced,0.5252057647705077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,8,16,balanced,0.6753254699707031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,8,16,balanced,0.026730239391326904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,8,16,balanced,0.02857599973678589
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,8,16,balanced,0.02724479913711548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,8,16,balanced,0.028898561000823976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,8,16,balanced,0.029726719856262206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,8,16,balanced,0.03667327880859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,8,16,balanced,0.04029056072235108
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,8,16,balanced,0.039051520824432376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,8,16,balanced,0.039768319129943844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,8,16,balanced,0.03986432075500489
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,8,16,balanced,0.04160639762878418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,8,16,balanced,0.04356991767883301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,8,16,balanced,0.04644864082336426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,8,16,balanced,0.042561278343200684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,8,16,balanced,0.04437888145446777
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,8,16,balanced,0.04521471977233887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,8,16,balanced,0.04848896026611328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,8,16,balanced,0.05002880096435547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,8,16,balanced,0.054791679382324224
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,8,16,balanced,0.07041664123535156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,8,16,balanced,0.08171520233154297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,8,16,balanced,0.11316224098205567
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,8,16,balanced,0.13693183898925781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,8,16,balanced,0.1887615966796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,8,16,balanced,0.23943424224853516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,8,16,balanced,0.3418918228149414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,8,16,balanced,0.44362239837646483
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,32,power_law_1.2,0.8476659393310546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,4,32,power_law_1.2,0.05854080200195313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,4,32,power_law_1.2,0.08639871597290039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,4,32,power_law_1.2,0.11390080451965331
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,4,32,power_law_1.2,0.16368896484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,8,16,balanced,0.017382400035858152
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,8,16,balanced,0.01997439980506897
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,8,16,balanced,0.017779200077056884
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,8,16,balanced,0.019982080459594726
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,8,16,balanced,0.020273919105529788
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,8,16,balanced,0.020660479068756104
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,8,16,balanced,0.02061568021774292
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,8,16,balanced,0.020573439598083495
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,8,16,balanced,0.020682239532470705
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,8,16,balanced,0.02067199945449829
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,8,16,balanced,0.020846080780029298
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,8,16,balanced,0.020622079372406003
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,8,16,balanced,0.020687360763549805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,8,16,balanced,0.021087999343872073
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,8,16,balanced,0.022305281162261964
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,8,16,balanced,0.022517759799957276
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,8,16,balanced,0.019199999570846556
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,4,32,power_law_1.2,0.19443967819213867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,8,16,balanced,0.01777791976928711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,8,16,balanced,0.02020224094390869
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,8,16,balanced,0.019841279983520508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,8,16,power_law_1.01,0.025029120445251463
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,8,16,power_law_1.01,0.026805760860443117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,8,16,power_law_1.01,0.026375679969787597
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,8,16,power_law_1.01,0.027210240364074705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,8,16,power_law_1.01,0.046083841323852535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,8,16,power_law_1.01,0.04934271812438965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,8,16,power_law_1.01,0.051399679183959955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,8,16,power_law_1.01,0.05139711856842041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,8,16,power_law_1.01,0.05275775909423828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,8,16,power_law_1.01,0.05463935852050781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,8,16,balanced,0.019897600412368772
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,8,16,power_law_1.01,0.057184000015258786
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,8,16,power_law_1.01,0.0625382423400879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,8,16,power_law_1.01,0.06082176208496094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,8,16,power_law_1.01,0.0795686388015747
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,8,16,power_law_1.01,0.08352767944335937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,8,16,power_law_1.01,0.08455039978027343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,8,16,power_law_1.01,0.0909823989868164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,8,16,power_law_1.01,0.1126591968536377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,8,16,power_law_1.01,0.12540800094604493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,8,16,power_law_1.01,0.14722175598144532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,8,16,power_law_1.01,0.19204864501953126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,8,16,power_law_1.01,0.24040447235107423
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,8,16,power_law_1.01,0.2979302406311035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,8,16,power_law_1.01,0.44105983734130855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,8,16,power_law_1.01,0.5670143890380859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,8,16,power_law_1.01,0.9608704376220704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,8,16,power_law_1.01,0.028454399108886717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,8,16,power_law_1.01,0.029227519035339357
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,8,16,power_law_1.01,0.029411840438842773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,8,16,power_law_1.01,0.029667840003967282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,8,16,power_law_1.01,0.03664511919021606
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,8,16,balanced,0.02807039976119995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,8,16,power_law_1.01,1.057955856323242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,8,16,power_law_1.01,0.041150717735290526
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,8,16,balanced,0.026439681053161624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,8,16,power_law_1.01,0.038794240951538085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,8,16,power_law_1.01,0.04033664226531982
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,8,16,power_law_1.01,0.04193920135498047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,8,16,power_law_1.01,0.044442877769470215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,8,16,power_law_1.01,0.04684800148010254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,8,16,power_law_1.01,0.0452185583114624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,8,16,power_law_1.01,0.04442240238189697
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,8,16,power_law_1.01,0.04592512130737304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,8,16,power_law_1.01,0.04751232147216797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,8,16,power_law_1.01,0.05334271907806396
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,8,16,balanced,0.04207104206085205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,8,16,power_law_1.01,0.06657792091369628
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,8,16,power_law_1.01,0.03836159944534302
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,8,16,power_law_1.01,0.06756224155426024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,8,16,power_law_1.01,0.09213695526123047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,8,16,power_law_1.01,0.1157913589477539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,8,16,power_law_1.01,0.154967041015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,8,16,power_law_1.01,0.18191104888916015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,8,16,power_law_1.01,0.2787264060974121
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,8,16,power_law_1.01,0.02018048048019409
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,8,16,balanced,0.03502336025238037
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,8,16,power_law_1.01,0.3811443328857422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,8,16,power_law_1.01,0.020230400562286376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,8,16,power_law_1.01,0.019779839515686036
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,8,16,power_law_1.01,0.019481600522994996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,8,16,power_law_1.01,0.02017791986465454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,8,16,power_law_1.01,0.61109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,8,16,power_law_1.01,0.019823360443115234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,8,16,balanced,0.057519359588623045
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,8,16,power_law_1.01,0.020010240077972412
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,8,16,power_law_1.01,0.020087039470672606
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,8,16,power_law_1.01,0.02030080080032349
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,8,16,power_law_1.01,0.02022655963897705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,8,16,power_law_1.01,0.7951538848876953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,8,16,power_law_1.01,0.020368640422821042
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,8,16,power_law_1.01,0.02031359910964966
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,8,16,power_law_1.01,0.020884480476379395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,8,16,power_law_1.01,0.022673919200897216
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,8,16,power_law_1.01,0.020684800148010253
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,8,16,balanced,0.07038464069366454
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,8,16,power_law_1.01,0.01987712025642395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,8,16,power_law_1.01,0.04247039794921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,8,16,power_law_1.01,0.026408960819244386
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,8,16,power_law_1.01,0.02145279884338379
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,8,16,power_law_1.01,0.0245632004737854
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,8,16,power_law_1.01,0.03250943899154663
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,8,16,power_law_1.2,0.025081601142883304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,8,16,power_law_1.2,0.02638335943222046
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,8,16,power_law_1.01,0.052294402122497564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,8,16,power_law_1.2,0.02636415958404541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,8,16,power_law_1.2,0.028038399219512938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,8,16,power_law_1.2,0.04684160232543945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,8,16,power_law_1.2,0.04896128177642822
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,8,16,power_law_1.2,0.051376638412475584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,8,16,power_law_1.2,0.052641282081604
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,8,16,power_law_1.2,0.054214401245117186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,8,16,power_law_1.2,0.055842561721801756
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,8,16,power_law_1.2,0.060775680541992186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,8,16,power_law_1.2,0.0624832010269165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,8,16,power_law_1.2,0.058791680335998533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,8,16,power_law_1.2,0.06795775890350342
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,8,16,power_law_1.2,0.0787724781036377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,8,16,power_law_1.2,0.08126463890075683
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,8,16,power_law_1.2,0.09538816452026368
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,8,16,power_law_1.2,0.1139801597595215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,8,16,power_law_1.2,0.1330239963531494
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,8,16,power_law_1.2,0.1725107192993164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,8,16,power_law_1.2,0.21606016159057617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,8,16,power_law_1.2,0.2752895927429199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,8,16,power_law_1.2,0.49319808959960937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,8,16,power_law_1.2,0.34534912109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,8,16,power_law_1.01,0.06115200042724609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,8,16,power_law_1.2,0.028769280910491947
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,8,16,power_law_1.2,0.02882175922393799
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,8,16,power_law_1.2,0.029137918949127196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,8,16,power_law_1.2,0.02922624111175537
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,8,16,power_law_1.2,0.0354201602935791
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,8,16,power_law_1.2,0.03707776069641113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,8,16,power_law_1.2,0.03862656116485595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,8,16,power_law_1.2,0.03936768054962158
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,8,16,power_law_1.2,0.0406822395324707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,8,16,power_law_1.2,0.04152575969696045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,8,16,power_law_1.2,0.04566143989562989
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,8,16,power_law_1.2,0.6396543884277344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,8,16,power_law_1.2,0.0468723201751709
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,8,16,power_law_1.2,0.044035840034484866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,8,16,power_law_1.2,0.8866124725341796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,8,16,power_law_1.2,0.04773759841918945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,8,16,power_law_1.2,0.04839424133300781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,8,16,power_law_1.2,0.05421055793762207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,8,16,power_law_1.2,0.06499199867248535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,8,16,power_law_1.2,0.07569151878356933
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,8,16,power_law_1.2,0.10486783981323242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,8,16,power_law_1.2,0.12958335876464844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,8,16,power_law_1.2,0.04559232234954834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,8,16,power_law_1.2,0.1790630340576172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,8,16,power_law_1.2,0.2345254325866699
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,8,16,power_law_1.2,0.34777217864990234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,8,16,power_law_1.2,0.48471038818359374
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,8,16,power_law_1.2,0.020171520709991456
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,8,16,power_law_1.2,0.019618560075759888
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,8,16,power_law_1.2,0.019850239753723145
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,8,16,power_law_1.2,0.01868672013282776
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,8,16,power_law_1.2,0.018944000005722043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,8,16,power_law_1.2,0.020135679244995118
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,8,16,power_law_1.2,0.020225279331207276
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,8,16,power_law_1.2,0.020198400020599368
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,8,16,power_law_1.2,0.020165119171142578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,8,16,power_law_1.2,0.020293118953704833
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,8,16,power_law_1.2,0.02067199945449829
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,8,16,power_law_1.2,0.020281600952148437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,8,16,power_law_1.2,0.02059391975402832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,8,16,power_law_1.2,0.020655360221862793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,8,16,power_law_1.01,0.08541824340820312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,8,16,power_law_1.01,0.10471039772033693
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,8,16,power_law_1.2,0.02306816101074219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,8,16,power_law_1.01,0.1442854404449463
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,8,16,power_law_1.01,0.18039680480957032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,8,16,power_law_1.2,0.6697574615478515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,8,16,power_law_1.2,0.9167475128173829
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,8,16,power_law_1.2,0.021422080993652344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,8,16,power_law_1.2,0.022412800788879396
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,8,16,power_law_1.2,0.026007039546966555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,8,16,power_law_1.2,0.03273344039916992
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,8,16,power_law_1.2,0.04201216220855713
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,8,16,power_law_1.2,0.05019008159637452
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,16,8,balanced,0.02596479892730713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,16,8,balanced,0.024714241027832033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,16,8,balanced,0.025976319313049317
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,16,8,balanced,0.028250880241394043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,16,8,balanced,0.031298561096191405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,16,8,balanced,0.05358719825744629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,16,8,balanced,0.05504000186920166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,16,8,balanced,0.05708032131195069
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,16,8,balanced,0.05919360160827637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,16,8,balanced,0.06160128116607666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,16,8,balanced,0.05833216190338135
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,16,8,balanced,0.061975040435791016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,16,8,balanced,0.06412799835205078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,16,8,balanced,0.07231743812561035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,16,8,balanced,0.07829376220703124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,16,8,balanced,0.08628864288330078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,8,16,power_law_1.2,0.06605311870574951
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,16,8,balanced,0.09971327781677246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,16,8,balanced,0.12372351646423341
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,16,8,balanced,0.14201215744018553
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,16,8,balanced,0.18019712448120118
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,16,8,balanced,0.21771135330200195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,16,8,balanced,0.2910694313049317
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,16,8,balanced,0.3628646469116211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,16,8,balanced,0.5093401718139648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,16,8,balanced,0.6518195343017579
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,8,16,power_law_1.2,1.2905690002441408
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,16,8,balanced,0.02645632028579712
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,16,8,balanced,0.027577600479125976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,16,8,balanced,1.2221427154541016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,16,8,balanced,0.028823039531707763
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,16,8,balanced,0.030960640907287597
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,16,8,balanced,0.034556159973144526
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,16,8,balanced,0.9360652923583984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,16,8,balanced,0.042897920608520504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,16,8,balanced,0.044270081520080565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,16,8,balanced,0.04608767986297607
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,16,8,balanced,0.04739071846008301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,16,8,balanced,0.04899199962615967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,16,8,balanced,0.0453004789352417
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,16,8,balanced,0.048089599609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,16,8,balanced,0.04903552055358887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,16,8,balanced,0.05100800037384033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,16,8,balanced,0.05369215965270996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,16,8,balanced,0.05273471832275391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,16,8,balanced,0.05933184146881103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,16,8,balanced,0.07278079986572265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,16,8,balanced,0.08482943534851074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,16,8,balanced,0.10930432319641112
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,16,8,balanced,0.13150464057922365
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,16,8,balanced,0.18779903411865234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,16,8,balanced,0.23192960739135743
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,16,8,balanced,0.3317184066772461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,16,8,balanced,0.4276774215698242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,16,8,balanced,0.6229056167602539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,16,8,balanced,0.8144306945800782
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,8,16,power_law_1.2,0.07535999774932861
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,8,16,power_law_1.2,0.11053695678710937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,8,16,power_law_1.2,0.14591487884521484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,8,16,power_law_1.2,0.21475967407226562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,16,8,balanced,0.018625279664993284
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,16,8,balanced,0.025053439140319822
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,16,8,balanced,0.02476799964904785
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,16,8,balanced,0.0247705602645874
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,16,8,balanced,0.025118720531463624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,16,8,balanced,0.025103359222412108
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,16,8,balanced,0.025866239070892333
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,16,8,balanced,0.018978559970855714
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,16,8,balanced,0.02521600008010864
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,16,8,balanced,0.024391679763793944
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,16,8,balanced,0.02518656015396118
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,16,8,balanced,0.020285439491271973
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,16,8,balanced,0.024766719341278075
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,16,8,balanced,0.027096319198608398
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,16,8,balanced,0.027182080745697022
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,16,8,balanced,0.028487679958343504
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,16,8,balanced,0.030956799983978274
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,8,16,power_law_1.2,0.2645670318603516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,16,8,balanced,0.02179327964782715
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,16,8,balanced,0.024798719882965087
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,16,8,balanced,0.02480128049850464
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,16,8,power_law_1.01,0.02683903932571411
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,16,8,balanced,0.024384000301361085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,16,8,power_law_1.01,0.027631359100341795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,16,8,power_law_1.01,0.028432641029357907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,16,8,power_law_1.01,0.03223423957824707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,16,8,power_law_1.01,0.051457281112670894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,16,8,power_law_1.01,0.02750080108642578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,16,8,power_law_1.01,0.05386623859405517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,16,8,power_law_1.01,0.05674111843109131
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,16,8,power_law_1.01,0.057140479087829585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,16,8,power_law_1.01,0.059253759384155266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,16,8,power_law_1.01,0.0628492784500122
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,16,8,power_law_1.01,0.06450687885284424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,16,8,power_law_1.01,0.06320127964019776
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,16,8,power_law_1.01,0.0724351978302002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,16,8,power_law_1.01,0.0832307243347168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,16,8,power_law_1.01,0.07725823879241943
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,16,8,power_law_1.01,0.09720959663391113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,16,8,power_law_1.01,0.11763968467712402
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,16,8,power_law_1.01,0.13823488235473633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,16,8,power_law_1.01,0.16114816665649415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,16,8,power_law_1.01,0.20466688156127927
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,16,8,power_law_1.01,0.27050367355346683
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,16,8,power_law_1.01,0.325351676940918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,16,8,power_law_1.01,0.4227507019042969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,16,8,power_law_1.01,0.5651852798461914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,16,8,power_law_1.01,0.9328383636474609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,16,8,power_law_1.01,0.028440320491790773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,16,8,power_law_1.01,0.030520319938659668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,16,8,power_law_1.01,1.2376908874511718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,16,8,power_law_1.01,0.0318336009979248
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,16,8,balanced,0.061820158958435054
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,16,8,power_law_1.01,0.03220479965209961
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,16,8,balanced,0.0723199987411499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,16,8,power_law_1.01,0.03460992097854614
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,16,8,power_law_1.01,0.041831679344177246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,16,8,power_law_1.01,0.04277632236480713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,16,8,power_law_1.01,0.043888640403747556
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,16,8,balanced,0.041751041412353515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,16,8,power_law_1.01,0.04762239933013916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,16,8,power_law_1.01,0.04973824024200439
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,16,8,balanced,0.0452569580078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,16,8,power_law_1.01,0.04727168083190918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,16,8,power_law_1.01,0.04927872180938721
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,16,8,power_law_1.01,0.04948224067687988
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,16,8,power_law_1.01,0.05536255836486816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,16,8,power_law_1.01,0.0594432020187378
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,16,8,power_law_1.01,0.07106560230255127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,16,8,power_law_1.01,0.08291071891784668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,16,8,power_law_1.01,0.10973183631896972
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,16,8,power_law_1.01,0.13201663970947267
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,16,8,power_law_1.01,0.1803481674194336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,16,8,power_law_1.01,0.2133580780029297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,16,8,power_law_1.01,0.0448576021194458
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,16,8,power_law_1.01,0.3226995086669922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,16,8,power_law_1.01,0.42383998870849615
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,16,8,power_law_1.01,0.02150399923324585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,16,8,power_law_1.01,0.6119551849365235
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,16,8,balanced,0.09932928085327149
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,16,8,power_law_1.01,0.023978240489959717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,16,8,power_law_1.01,0.7967961883544922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,16,8,power_law_1.01,0.02428031921386719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,16,8,power_law_1.01,0.02431488037109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,16,8,power_law_1.01,0.02426624059677124
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,16,8,power_law_1.01,0.02459903955459595
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,16,8,power_law_1.01,0.02433919906616211
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,16,8,power_law_1.01,0.023927040100097656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,16,8,power_law_1.01,0.02475519895553589
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,16,8,power_law_1.01,0.02472575902938843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,16,8,power_law_1.01,0.04716288089752198
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,16,8,power_law_1.01,0.02516864061355591
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,16,8,power_law_1.01,0.02521600008010864
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,16,8,power_law_1.01,0.025884160995483397
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,16,8,power_law_1.01,0.027358720302581786
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,16,8,power_law_1.01,0.02231679916381836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,16,8,balanced,0.12299776077270508
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,16,8,power_law_1.01,0.021273601055145263
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,16,8,power_law_1.01,0.035987200736999514
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,16,8,power_law_1.01,0.0289574408531189
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,16,8,power_law_1.01,0.040757761001586915
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,16,8,power_law_1.01,0.11000576019287109
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,16,8,power_law_1.01,0.08273152351379395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,16,8,power_law_1.2,0.02477184057235718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,16,8,power_law_1.2,0.02557823896408081
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,16,8,power_law_1.01,0.057322239875793456
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,16,8,power_law_1.2,0.028444159030914306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,16,8,power_law_1.2,0.0293503999710083
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,16,8,power_law_1.2,0.03176320075988769
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,16,8,power_law_1.2,0.05664768218994141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,16,8,power_law_1.2,0.0562879991531372
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,16,8,power_law_1.2,0.058169598579406734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,16,8,power_law_1.2,0.06000895977020264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,16,8,power_law_1.2,0.06328576087951661
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,16,8,power_law_1.2,0.06201727867126465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,16,8,power_law_1.2,0.06453760147094727
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,16,8,power_law_1.01,0.04893311977386475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,16,8,power_law_1.2,0.0764031982421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,16,8,power_law_1.2,0.08342911720275878
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,16,8,power_law_1.2,0.09062911987304687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,16,8,power_law_1.2,0.10079872131347656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,16,8,power_law_1.2,0.12210047721862793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,16,8,power_law_1.2,0.054656000137329096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,16,8,power_law_1.2,0.13980287551879883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,16,8,power_law_1.2,0.1785548782348633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,16,8,power_law_1.2,0.20739456176757814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,16,8,power_law_1.2,0.286874885559082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,16,8,power_law_1.2,0.35986305236816407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,16,8,power_law_1.2,0.5158335876464843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,16,8,power_law_1.2,0.030115840435028074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,16,8,power_law_1.2,0.029653759002685548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,16,8,power_law_1.2,0.03169663906097412
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,16,8,power_law_1.2,0.03216000080108643
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,16,8,power_law_1.2,0.03307647943496704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,16,8,power_law_1.2,0.04200575828552246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,16,8,power_law_1.2,0.6600717163085937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,16,8,power_law_1.2,0.04244351863861084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,16,8,power_law_1.2,0.044372482299804686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,16,8,power_law_1.2,0.045670399665832515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,16,8,power_law_1.2,0.04648191928863525
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,16,8,power_law_1.2,0.049697279930114746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,16,8,power_law_1.2,0.04811391830444336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,16,8,power_law_1.2,0.04833920001983642
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,16,8,power_law_1.2,0.05015039920806885
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,16,8,power_law_1.2,0.05096320152282715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,16,8,power_law_1.2,0.053786878585815434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,16,8,power_law_1.2,0.05681024074554444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,16,8,power_law_1.2,0.07252736091613769
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,16,8,power_law_1.2,0.08378623962402344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,16,8,power_law_1.2,0.11156479835510254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,16,8,power_law_1.2,0.13259136199951171
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,16,8,power_law_1.2,0.18232959747314453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,16,8,power_law_1.2,0.24353151321411132
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,16,8,power_law_1.2,0.4772787094116211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,16,8,power_law_1.2,0.3512358474731445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,16,8,power_law_1.2,0.6754994964599609
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,16,8,power_law_1.2,0.021862399578094483
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,16,8,power_law_1.2,0.9570790100097657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,16,8,power_law_1.2,0.021098239421844484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,16,8,power_law_1.2,0.02272128105163574
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,16,8,power_law_1.2,0.022663679122924805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,16,8,power_law_1.2,0.023121919631958008
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,16,8,power_law_1.2,0.023879680633544922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,16,8,power_law_1.2,0.02431999921798706
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,16,8,power_law_1.2,0.02429824113845825
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,16,8,power_law_1.2,0.024357120990753173
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,16,8,power_law_1.2,0.02466048002243042
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,16,8,power_law_1.2,0.02554879903793335
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,16,8,power_law_1.2,0.025712640285491944
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,16,8,power_law_1.01,0.1315392017364502
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,16,8,power_law_1.2,0.025785601139068602
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,16,8,power_law_1.01,0.16942975997924806
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,16,8,power_law_1.01,0.3090163230895996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,16,8,power_law_1.2,1.2633881378173828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,16,8,power_law_1.2,0.04654079914093018
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,16,8,power_law_1.2,0.8948966217041014
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,16,8,power_law_1.2,0.03054336071014404
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,16,8,power_law_1.2,0.028012800216674804
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,16,8,power_law_1.2,0.04083583831787109
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,16,8,power_law_1.2,0.032289280891418456
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,16,8,power_law_1.2,0.06248832225799561
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,16,8,power_law_1.2,0.07393280029296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,16,8,power_law_1.2,0.10333056449890136
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,32,4,balanced,0.024070401191711426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,32,4,balanced,0.025217280387878415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,32,4,balanced,0.027573759555816653
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,32,4,balanced,0.0306329607963562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,32,4,balanced,0.035024640560150144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,32,4,balanced,0.042000641822814946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,32,4,balanced,0.062051839828491204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,32,4,balanced,0.0641267204284668
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,16,8,power_law_1.2,0.13847295761108397
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,32,4,balanced,0.0653004789352417
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,32,4,balanced,0.06854400157928467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,32,4,balanced,0.06635136127471923
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,32,4,balanced,0.07132031917572021
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,32,4,balanced,0.08082431793212891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,32,4,balanced,0.08807935714721679
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,32,4,balanced,0.09476608276367188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,32,4,balanced,0.10702336311340332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,32,4,balanced,0.06409855842590331
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,32,4,balanced,0.12905344009399414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,32,4,balanced,0.1464857578277588
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,32,4,balanced,0.1855219268798828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,32,4,balanced,0.22220415115356446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,32,4,balanced,0.29602943420410155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,32,4,balanced,0.36860416412353514
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,32,4,balanced,0.515676155090332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,32,4,balanced,0.6569023895263671
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,32,4,balanced,0.943031005859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,32,4,balanced,0.027594239711761476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,32,4,balanced,0.02850303888320923
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,32,4,balanced,0.03300607919692993
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,32,4,balanced,1.2541977691650392
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,32,4,balanced,0.03625600099563599
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,32,4,balanced,0.044381442070007326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,32,4,balanced,0.04536064147949219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,32,4,balanced,0.05363071918487548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,32,4,balanced,0.05503359794616699
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,32,4,balanced,0.056765441894531254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,32,4,balanced,0.058873600959777825
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,32,4,balanced,0.053835520744323725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,32,4,balanced,0.05677696228027344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,32,4,balanced,0.05602560043334961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,32,4,balanced,0.0595904016494751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,32,4,balanced,0.05947775840759277
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,32,4,balanced,0.06112383842468262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,32,4,balanced,0.06739071846008302
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,32,4,balanced,0.08054143905639648
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,16,8,power_law_1.2,0.2625459289550781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,32,4,balanced,0.0916915225982666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,32,4,balanced,0.1168934440612793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,32,4,balanced,0.13751935958862305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,32,4,balanced,0.20190719604492186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,32,4,balanced,0.24260351181030276
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,32,4,balanced,0.34691200256347654
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,16,8,power_law_1.2,0.1796313667297363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,32,4,balanced,0.449854736328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,32,4,balanced,0.6645606231689454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,32,4,power_law_1.01,0.02764928102493286
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,32,4,power_law_1.01,0.025208320617675785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,32,4,power_law_1.01,0.028971519470214847
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,32,4,power_law_1.01,0.03092351913452148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,32,4,power_law_1.01,0.033369600772857666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,32,4,balanced,0.8695053100585938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,32,4,power_law_1.01,0.04034687995910645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,32,4,power_law_1.01,0.06368000030517577
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,32,4,power_law_1.01,0.06529535770416259
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,32,4,power_law_1.01,0.07045631885528565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,32,4,power_law_1.01,0.0674124813079834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,32,4,power_law_1.01,0.06974592208862304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,32,4,power_law_1.01,0.07829247951507569
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,32,4,power_law_1.01,0.08562432289123535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,32,4,power_law_1.01,0.09347071647644042
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,32,4,power_law_1.01,0.10648320198059083
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,32,4,power_law_1.01,0.1278553581237793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,32,4,power_law_1.01,0.14582015991210937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,32,4,power_law_1.01,0.18136703491210937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,32,4,power_law_1.01,0.0669491195678711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,32,4,power_law_1.01,0.22165248870849608
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,32,4,power_law_1.01,0.2939046478271484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,32,4,power_law_1.01,0.37107967376708983
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,32,4,power_law_1.01,0.5252428817749023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,32,4,power_law_1.01,0.062001919746398924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,32,4,power_law_1.01,0.6759129333496093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,32,4,power_law_1.01,0.0288371205329895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,32,4,power_law_1.01,0.03507200002670288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,32,4,power_law_1.01,0.03306623935699463
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,32,4,power_law_1.01,0.9938764953613282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,32,4,power_law_1.01,0.03736191987991333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,32,4,power_law_1.01,0.03994751930236816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,32,4,power_law_1.01,0.04401279926300049
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,32,4,power_law_1.01,0.051508479118347175
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,32,4,power_law_1.01,1.2996902465820312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,32,4,power_law_1.01,0.053635840415954594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,32,4,power_law_1.01,0.05589888095855713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,32,4,power_law_1.01,0.05755136013031006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,32,4,power_law_1.01,0.05759359836578369
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,32,4,power_law_1.01,0.05502848148345947
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,32,4,power_law_1.01,0.056318721771240234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,32,4,power_law_1.01,0.059311361312866205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,32,4,power_law_1.01,0.060737280845642096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,32,4,power_law_1.01,0.0647872018814087
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,32,4,power_law_1.01,0.06910592079162597
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,32,4,power_law_1.01,0.08137344360351563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,32,4,power_law_1.01,0.09636992454528809
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,32,4,power_law_1.01,0.1238912010192871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,32,4,power_law_1.01,0.15384448051452637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,32,4,power_law_1.01,0.20541568756103518
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,32,4,power_law_1.01,0.26219648361206055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,32,4,power_law_1.01,0.36488449096679687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,32,4,power_law_1.01,0.4889932632446289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,32,4,power_law_1.2,0.02564095973968506
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,32,4,power_law_1.2,0.028811519145965574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,32,4,power_law_1.01,0.6918707275390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,32,4,power_law_1.2,0.03045759916305542
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,32,4,power_law_1.2,0.025950720310211183
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,32,4,power_law_1.2,0.04050943851470947
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,32,4,power_law_1.2,0.03392640113830566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,32,4,power_law_1.01,0.9131378936767579
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,32,4,power_law_1.2,0.06363135814666748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,32,4,power_law_1.2,0.0665011215209961
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,16,8,power_law_1.2,0.464977912902832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,32,4,power_law_1.2,0.06731904029846192
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,32,4,power_law_1.2,0.07098239898681641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,32,4,power_law_1.2,0.06739712238311768
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,32,4,power_law_1.2,0.07752575874328613
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,32,4,power_law_1.2,0.08916095733642579
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,32,4,power_law_1.2,0.09463935852050781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,32,4,power_law_1.2,0.10895615577697755
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,32,4,power_law_1.2,0.12954112052917482
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,32,4,power_law_1.2,0.14804736137390137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,32,4,power_law_1.2,0.06187903881072998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,32,4,power_law_1.2,0.186694393157959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,32,4,power_law_1.2,0.21898752212524414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,32,4,power_law_1.2,0.29781503677368165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,32,4,power_law_1.2,0.367564811706543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,32,4,power_law_1.2,0.07052800178527832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,16,8,power_law_1.2,0.3554240036010742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,32,4,power_law_1.2,0.029693439006805423
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,32,4,power_law_1.2,0.032538878917694095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,32,4,power_law_1.2,0.5275008010864257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,32,4,power_law_1.2,0.035535359382629396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,32,4,power_law_1.2,0.036602880954742435
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,32,4,power_law_1.2,0.04771967887878418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,32,4,power_law_1.2,0.03806976079940796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,32,4,power_law_1.2,0.05182208061218262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,32,4,power_law_1.2,0.05429632186889648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,32,4,power_law_1.2,0.05625728130340576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,32,4,power_law_1.2,0.05705344200134277
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,32,4,power_law_1.2,0.05963776111602783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,32,4,power_law_1.2,0.05544320106506347
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,32,4,power_law_1.2,0.05660160064697266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,32,4,power_law_1.2,0.0595686388015747
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,32,4,power_law_1.2,0.06084991931915283
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,32,4,power_law_1.2,0.06370944023132324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,32,4,power_law_1.2,0.06882944107055664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,32,4,power_law_1.2,0.08386943817138673
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,32,4,power_law_1.2,0.09848447799682616
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,32,4,power_law_1.2,0.12516608238220214
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,32,4,power_law_1.2,0.15549311637878419
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,32,4,power_law_1.2,0.6754137420654297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,32,4,power_law_1.2,0.26428159713745114
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,32,4,power_law_1.2,0.21577215194702148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,32,4,power_law_1.2,0.38263038635253904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,32,4,power_law_1.2,0.4965465545654297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,4,32,balanced,0.030849280357360842
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,4,32,balanced,0.03326080083847046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,4,32,balanced,0.03210239887237549
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,32,4,power_law_1.2,0.7092467498779297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,4,32,balanced,0.05643263816833496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,4,32,balanced,0.11041919708251953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,4,32,balanced,0.1096947193145752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,4,32,balanced,0.11437439918518066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,4,32,balanced,0.11042559623718262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,4,32,balanced,0.11682304382324218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,4,32,balanced,0.1153536033630371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,4,32,balanced,0.1157363224029541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,4,32,balanced,0.1196339225769043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,4,32,balanced,0.12483967781066893
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,4,32,balanced,0.03131648063659668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,4,32,balanced,0.1252019214630127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,4,32,balanced,0.12967040061950683
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,4,32,balanced,0.12517888069152833
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,4,32,balanced,0.13472895622253417
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,4,32,balanced,0.1393395233154297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,4,32,balanced,0.15524352073669434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,4,32,balanced,0.17070079803466798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,4,32,balanced,0.20156160354614258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,4,32,balanced,0.23410688400268556
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,4,32,balanced,0.2930137634277344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,32,4,power_law_1.2,1.0156237030029298
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,4,32,balanced,0.3584870529174805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,4,32,balanced,0.032170240879058835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,4,32,balanced,0.5104857635498047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,4,32,balanced,0.03207551956176758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,4,32,balanced,0.03292160034179688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,4,32,balanced,0.032975358963012694
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,4,32,balanced,0.675230712890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,4,32,balanced,0.03581056118011475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,4,32,balanced,0.043302397727966305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,4,32,balanced,0.04937215805053711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,4,32,balanced,0.04609536170959473
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,4,32,balanced,0.05179135799407959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,4,32,balanced,0.046452479362487795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,4,32,balanced,0.04686207771301269
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,4,32,balanced,0.04871424198150635
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,4,32,balanced,0.05182591915130615
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,4,32,balanced,0.05275775909423828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,4,32,balanced,0.05608575820922852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,4,32,balanced,0.05412864208221435
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,4,32,balanced,0.06039040088653565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,4,32,balanced,0.06312575817108154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,4,32,balanced,0.07235455989837647
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,4,32,balanced,0.08288384437561035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,4,32,balanced,0.10445568084716797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,4,32,balanced,0.04408063888549805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,4,32,balanced,0.12279680252075195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,4,32,balanced,0.16418304443359374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,4,32,balanced,0.20361984252929685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,4,32,balanced,0.2812544059753418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,4,32,balanced,0.3766515350341797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,32,4,power_law_1.2,0.9568370819091797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,32,4,power_law_1.2,1.3206668090820313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,4,32,balanced,0.019408639669418335
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,4,32,balanced,0.01940608024597168
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,4,32,balanced,0.019037439823150634
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,4,32,balanced,0.02078336000442505
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,4,32,balanced,0.019444479942321777
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,4,32,balanced,0.021112320423126222
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,4,32,balanced,0.021125121116638185
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,4,32,balanced,0.021076478958129884
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,4,32,balanced,0.02108288049697876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,4,32,balanced,0.02185472011566162
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,4,32,balanced,0.02152064085006714
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,4,32,balanced,0.021529600620269776
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,4,32,balanced,0.021488640308380127
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,4,32,balanced,0.021864960193634035
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,4,32,balanced,0.02183039903640747
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,4,32,balanced,0.02196991920471191
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,4,32,balanced,0.021994240283966064
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,4,32,balanced,0.02236288070678711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,4,32,balanced,0.023879680633544922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,4,32,balanced,0.02066431999206543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,4,32,balanced,0.020696320533752442
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,4,32,power_law_1.01,0.06051583766937256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,4,32,power_law_1.01,0.06050559997558593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,4,32,power_law_1.01,0.06109951972961426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,4,32,power_law_1.01,0.10411135673522949
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,4,32,power_law_1.01,0.10519295692443849
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,4,32,power_law_1.01,0.10923775672912597
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,4,32,power_law_1.01,0.11292032241821288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,4,32,power_law_1.01,0.11264127731323241
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,4,32,power_law_1.01,0.11533056259155275
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,4,32,power_law_1.01,0.12123135566711425
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,4,32,power_law_1.01,0.12330368041992187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,4,32,power_law_1.01,0.13041024208068847
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,4,32,power_law_1.01,0.12660223960876466
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,4,32,power_law_1.01,0.1306713581085205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,4,32,power_law_1.01,0.13806079864501952
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,4,32,power_law_1.01,0.1424678421020508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,4,32,power_law_1.01,0.16308607101440428
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,4,32,power_law_1.01,0.1820800018310547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,4,32,power_law_1.01,0.22477184295654298
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,4,32,power_law_1.01,0.25651968002319336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,4,32,power_law_1.01,0.3908019256591797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,4,32,power_law_1.01,0.5136243057250977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,4,32,power_law_1.01,0.650297622680664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,4,32,power_law_1.01,0.9764236450195313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,4,32,power_law_1.01,1.3221363830566406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,4,32,power_law_1.01,1.8880152893066406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,4,32,power_law_1.01,0.03462271928787232
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,4,32,power_law_1.01,0.034579200744628905
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,4,32,balanced,0.030744318962097165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,4,32,power_law_1.01,0.03536256074905396
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,4,32,balanced,0.04371071815490722
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,4,32,balanced,0.05292416095733643
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,4,32,power_law_1.01,0.03540607929229737
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,4,32,power_law_1.01,0.04271232128143311
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,4,32,power_law_1.01,0.0435865592956543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,4,32,power_law_1.01,0.05018112182617187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,4,32,power_law_1.01,0.050417919158935544
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,4,32,balanced,0.0286028790473938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,4,32,power_law_1.01,0.04724736213684082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,4,32,power_law_1.01,0.0474073600769043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,4,32,power_law_1.01,0.04980607986450195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,4,32,power_law_1.01,0.05345791816711426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,4,32,power_law_1.01,0.05389567852020264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,4,32,power_law_1.01,0.06081408023834228
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,4,32,power_law_1.01,0.062287359237670904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,4,32,power_law_1.01,0.0670681619644165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,4,32,power_law_1.01,0.0736191987991333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,4,32,power_law_1.01,0.09422592163085938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,4,32,power_law_1.01,0.11365504264831543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,4,32,power_law_1.01,0.1511513614654541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,4,32,power_law_1.01,0.21333248138427735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,4,32,power_law_1.01,0.267193603515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,4,32,power_law_1.01,0.05867519855499268
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,4,32,power_law_1.01,0.43569793701171877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,4,32,power_law_1.01,0.4612057495117187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,4,32,power_law_1.01,0.020719358921051027
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,4,32,power_law_1.01,0.021022720336914064
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,4,32,power_law_1.01,0.020702719688415527
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,4,32,power_law_1.01,0.7019635009765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,4,32,power_law_1.01,0.020730879306793213
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,4,32,power_law_1.01,0.02071552038192749
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,4,32,power_law_1.01,0.020700159072875975
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,4,32,power_law_1.01,0.02171008110046387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,4,32,power_law_1.01,0.9070976257324219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,4,32,balanced,0.07851776123046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,4,32,power_law_1.01,0.021786880493164063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,4,32,power_law_1.01,0.021502718925476075
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,4,32,power_law_1.01,0.021523199081420898
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,4,32,power_law_1.01,0.021731839179992676
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,4,32,power_law_1.01,0.021867520809173584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,4,32,power_law_1.01,0.02156287908554077
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,4,32,power_law_1.01,0.02191103935241699
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,4,32,power_law_1.01,0.021597440242767336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,4,32,power_law_1.01,0.026371839046478274
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,4,32,power_law_1.01,0.04807295799255371
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,4,32,balanced,0.061273598670959474
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,4,32,power_law_1.01,0.11005311965942384
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,4,32,power_law_1.01,0.024140799045562746
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,4,32,power_law_1.01,0.03078783988952637
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,4,32,power_law_1.01,0.022655999660491942
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,4,32,power_law_1.2,0.06051583766937256
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,4,32,power_law_1.01,0.03952896118164063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,4,32,power_law_1.2,0.06065663814544677
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,4,32,power_law_1.2,0.06077439785003662
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,4,32,power_law_1.2,0.10477567672729493
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,4,32,power_law_1.01,0.06086783885955811
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,4,32,power_law_1.2,0.105863676071167
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,4,32,power_law_1.2,0.11109503746032715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,4,32,power_law_1.2,0.11196672439575195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,4,32,power_law_1.2,0.11363967895507812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,4,32,power_law_1.01,0.07509376049041748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,4,32,power_law_1.2,0.11491583824157714
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,4,32,power_law_1.2,0.1156223964691162
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,4,32,power_law_1.2,0.12051199913024901
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,4,32,power_law_1.2,0.12830592155456544
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,4,32,power_law_1.2,0.131714563369751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,4,32,power_law_1.2,0.12973567962646484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,4,32,power_law_1.2,0.1374054431915283
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,4,32,power_law_1.2,0.14066176414489745
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,4,32,power_law_1.2,0.1538163185119629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,4,32,power_law_1.2,0.18147071838378906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,4,32,power_law_1.2,0.19659008026123045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,4,32,power_law_1.2,0.2544921684265137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,4,32,power_law_1.2,0.2954995155334473
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,4,32,power_law_1.01,0.10584832191467286
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,4,32,power_law_1.2,0.44397697448730467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,4,32,power_law_1.2,0.604089584350586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,4,32,power_law_1.2,0.034801919460296635
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,4,32,power_law_1.2,0.7964454650878906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,4,32,power_law_1.2,0.034919679164886475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,4,32,power_law_1.2,0.03548160076141357
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,4,32,power_law_1.2,0.04260096073150635
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,4,32,power_law_1.2,0.043955202102661135
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,4,32,power_law_1.2,0.0436620807647705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,4,32,power_law_1.2,0.04814720153808594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,4,32,power_law_1.2,0.04650752067565918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,4,32,power_law_1.2,1.155187225341797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,4,32,power_law_1.2,0.04776703834533692
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,4,32,power_law_1.2,0.04974207878112793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,4,32,power_law_1.2,0.05074687957763672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,4,32,power_law_1.2,0.05299327850341797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,4,32,power_law_1.2,0.056815361976623534
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,4,32,power_law_1.2,0.05764863967895508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,4,32,power_law_1.2,0.06361472129821777
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,4,32,power_law_1.2,0.06680191993713379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,4,32,power_law_1.2,0.07399295806884766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,4,32,power_law_1.2,0.09068032264709472
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,4,32,power_law_1.2,0.10536064147949217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,4,32,power_law_1.2,0.14249088287353515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,4,32,power_law_1.2,0.16272127151489257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,4,32,power_law_1.2,0.22796800613403317
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,4,32,power_law_1.2,0.35559806823730467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,4,32,power_law_1.2,0.46631809234619137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,4,32,power_law_1.2,0.7017164611816407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,4,32,power_law_1.2,0.02060800075531006
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,4,32,power_law_1.2,0.02070528030395508
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,4,32,power_law_1.2,0.020622079372406003
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,4,32,power_law_1.2,1.892194519042969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,4,32,power_law_1.2,0.020599040985107422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,4,32,power_law_1.2,0.021049599647521972
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,4,32,power_law_1.2,0.021446399688720703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,4,32,power_law_1.2,0.021442561149597167
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,4,32,power_law_1.2,1.2585177612304688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,4,32,power_law_1.2,0.02145024061203003
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,4,32,power_law_1.2,1.011868133544922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,4,32,power_law_1.2,0.021689600944519043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,4,32,power_law_1.2,0.02191744089126587
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,4,32,power_law_1.01,0.13656448364257812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,4,32,power_law_1.2,0.021497600078582764
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,4,32,power_law_1.2,0.02150144100189209
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,4,32,power_law_1.2,0.023653120994567872
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,4,32,power_law_1.2,2.3248704528808597
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,4,32,power_law_1.01,0.17922815322875976
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,4,32,power_law_1.01,0.27142656326293946
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,4,32,power_law_1.2,0.02066431999206543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,4,32,power_law_1.2,0.021410560607910155
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,4,32,power_law_1.2,0.02499072074890137
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,4,32,power_law_1.2,0.023165440559387206
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,4,32,power_law_1.2,0.030943999290466307
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,4,32,power_law_1.2,0.033729279041290285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,4,32,power_law_1.2,0.047662081718444826
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,8,16,balanced,0.030051839351654053
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,4,32,power_law_1.2,0.05609983921051025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,8,16,balanced,0.030027520656585694
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,8,16,balanced,0.031331839561462405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,8,16,balanced,0.03297408103942871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,8,16,balanced,0.05796224117279053
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,8,16,balanced,0.10905216217041017
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,8,16,balanced,0.11600000381469726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,8,16,balanced,0.1129190444946289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,8,16,balanced,0.11336959838867186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,8,16,balanced,0.11672960281372072
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,8,16,balanced,0.12034943580627441
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,8,16,balanced,0.12443008422851562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,8,16,balanced,0.12951168060302734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,8,16,balanced,0.12279040336608886
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,8,16,balanced,0.1265727996826172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,8,16,balanced,0.1282534408569336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,8,16,balanced,0.133306884765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,8,16,balanced,0.147390718460083
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,8,16,balanced,0.1612403106689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,8,16,balanced,0.1895359992980957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,8,16,balanced,0.21638784408569336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,8,16,balanced,0.27093120574951174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,8,16,balanced,0.3269388961791992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,8,16,balanced,0.43453311920166016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,8,16,balanced,0.5462886428833007
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,8,16,balanced,0.7951961517333984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,8,16,balanced,0.03167999982833862
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,8,16,balanced,0.03173120021820068
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,8,16,balanced,1.050603485107422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,8,16,balanced,0.03214976072311401
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,8,16,balanced,0.0341209602355957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,8,16,balanced,0.03788928031921387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,8,16,balanced,0.05224832057952881
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,8,16,balanced,0.05168384075164795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,8,16,balanced,0.05051263809204102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,8,16,balanced,0.05134592056274414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,8,16,balanced,0.05313536167144776
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,8,16,balanced,0.05500671863555908
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,8,16,balanced,0.05555840015411377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,8,16,balanced,0.059988479614257816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,8,16,balanced,0.05900415897369384
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,8,16,balanced,0.0614028787612915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,8,16,balanced,0.06408448219299316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,8,16,balanced,0.06653439998626709
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,8,16,balanced,0.07367680072784424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,8,16,balanced,0.08453120231628418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,8,16,balanced,0.09957504272460938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,8,16,balanced,0.11831168174743652
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,8,16,balanced,0.15874176025390624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,8,16,balanced,0.1920115280151367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,8,16,balanced,0.26831104278564455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,8,16,balanced,0.3425791931152344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,8,16,balanced,0.4960255813598633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,8,16,balanced,0.6405554962158203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,4,32,power_law_1.2,0.07963520050048828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,4,32,power_law_1.2,0.10498815536499025
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,4,32,power_law_1.2,0.14132736206054688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,4,32,power_law_1.2,0.16856704711914064
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,4,32,power_law_1.2,0.26420991897583007
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,4,32,power_law_1.2,0.3386431884765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,8,16,balanced,0.019389439821243287
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,8,16,balanced,0.01982208013534546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,8,16,balanced,0.0242905592918396
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,8,16,balanced,0.024024319648742676
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,8,16,balanced,0.024060161113739015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,8,16,balanced,0.024372479915618896
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,8,16,balanced,0.024761600494384767
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,8,16,balanced,0.024313600063323976
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,8,16,balanced,0.019461120367050173
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,8,16,balanced,0.024669439792633058
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,8,16,balanced,0.025126399993896487
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,8,16,balanced,0.025148160457611084
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,8,16,balanced,0.026012160778045655
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,8,16,balanced,0.02663935899734497
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,8,16,balanced,0.028049919605255123
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,8,16,balanced,0.029600000381469725
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,8,16,balanced,0.020770559310913088
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,8,16,balanced,0.022301440238952634
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,8,16,balanced,0.023816959857940672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,8,16,balanced,0.024092159271240234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,8,16,power_law_1.01,0.056514558792114256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,8,16,power_law_1.01,0.061246719360351566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,8,16,power_law_1.01,0.06077695846557617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,8,16,power_law_1.01,0.06032767772674561
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,8,16,power_law_1.01,0.10512127876281738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,8,16,power_law_1.01,0.10943615913391114
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,8,16,power_law_1.01,0.11044095993041991
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,8,16,power_law_1.01,0.11224703788757324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,8,16,power_law_1.01,0.11685119628906251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,8,16,power_law_1.01,0.11936767578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,8,16,balanced,0.02394495964050293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,8,16,power_law_1.01,0.12076288223266603
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,8,16,power_law_1.01,0.1273408031463623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,8,16,power_law_1.01,0.13051008224487304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,8,16,power_law_1.01,0.1272383975982666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,8,16,power_law_1.01,0.13324288368225096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,8,16,power_law_1.01,0.13551487922668456
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,8,16,power_law_1.01,0.1499289608001709
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,8,16,power_law_1.01,0.16714624404907225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,8,16,power_law_1.01,0.18913408279418947
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,8,16,power_law_1.01,0.2171558380126953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,8,16,power_law_1.01,0.26210432052612304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,8,16,power_law_1.01,0.37429630279541015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,8,16,power_law_1.01,0.43656574249267577
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,8,16,power_law_1.01,0.7151808166503907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,8,16,power_law_1.01,0.8610137939453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,8,16,power_law_1.01,1.2302758026123048
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,8,16,power_law_1.01,0.039961600303649904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,8,16,power_law_1.01,0.03904639959335327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,8,16,power_law_1.01,0.038675200939178464
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,8,16,power_law_1.01,1.7579110717773438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,8,16,power_law_1.01,0.03947135925292969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,8,16,power_law_1.01,0.047274241447448725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,8,16,power_law_1.01,0.04896383762359619
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,8,16,power_law_1.01,0.052101120948791504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,8,16,power_law_1.01,0.0510041618347168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,8,16,power_law_1.01,0.0539353609085083
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,8,16,power_law_1.01,0.05266304016113281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,8,16,power_law_1.01,0.05658751964569092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,8,16,power_law_1.01,0.06184192180633545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,8,16,power_law_1.01,0.06648704051971435
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,8,16,power_law_1.01,0.06353024005889893
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,8,16,power_law_1.01,0.06614016056060791
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,8,16,power_law_1.01,0.0678707218170166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,8,16,power_law_1.01,0.07140096187591552
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,8,16,power_law_1.01,0.08498432159423827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,8,16,power_law_1.01,0.09963007926940917
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,8,16,balanced,0.07181695938110352
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,8,16,balanced,0.0409446382522583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,8,16,power_law_1.01,0.13100543975830078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,8,16,power_law_1.01,0.1634611129760742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,8,16,balanced,0.04558720111846924
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,8,16,balanced,0.060890879631042484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,8,16,power_law_1.01,0.3400038528442383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,8,16,power_law_1.01,0.25341184616088863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,8,16,power_law_1.01,0.5057011032104493
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,8,16,power_law_1.01,0.022579200267791748
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,8,16,power_law_1.01,0.02367232084274292
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,8,16,power_law_1.01,0.023590400218963622
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,8,16,power_law_1.01,0.023649280071258542
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,8,16,power_law_1.01,0.02358016014099121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,8,16,power_law_1.01,0.7578009796142579
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,8,16,power_law_1.01,0.02365567922592163
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,8,16,power_law_1.01,0.023989760875701906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,8,16,power_law_1.01,0.023897600173950196
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,8,16,power_law_1.01,0.024311039447784424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,8,16,power_law_1.01,0.9883225250244141
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,8,16,power_law_1.01,0.02440063953399658
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,8,16,power_law_1.01,0.024769279956817627
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,8,16,power_law_1.01,0.02588799953460693
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,8,16,power_law_1.01,0.02592895984649658
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,8,16,power_law_1.01,0.027228159904479982
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,8,16,power_law_1.01,0.022385919094085695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,8,16,power_law_1.01,0.024718079566955566
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,8,16,balanced,0.08796544075012207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,8,16,power_law_1.01,0.20255104064941407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,8,16,balanced,0.11421055793762207
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,8,16,power_law_1.01,0.029119999408721925
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,8,16,power_law_1.01,0.038794240951538085
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,8,16,power_law_1.01,0.04494336128234863
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,8,16,power_law_1.01,0.03479552030563354
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,8,16,power_law_1.2,0.055258879661560054
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,8,16,power_law_1.2,0.060362238883972165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,8,16,power_law_1.2,0.061882882118225103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,8,16,power_law_1.2,0.05869823932647705
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,8,16,power_law_1.01,0.08227968215942383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,8,16,power_law_1.2,0.10588159561157226
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,8,16,power_law_1.01,0.05754752159118652
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,8,16,power_law_1.01,0.09192319869995116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,8,16,power_law_1.2,0.1107801628112793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,8,16,power_law_1.2,0.11087103843688965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,8,16,power_law_1.2,0.11418111801147461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,8,16,power_law_1.2,0.11689087867736817
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,8,16,power_law_1.2,0.12365183830261231
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,8,16,power_law_1.2,0.1277184009552002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,8,16,power_law_1.2,0.1276313591003418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,8,16,power_law_1.2,0.12863871574401856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,8,16,power_law_1.2,0.13403648376464844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,8,16,power_law_1.2,0.1376460838317871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,8,16,power_law_1.2,0.14911871910095215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,8,16,power_law_1.2,0.17521791458129882
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,8,16,power_law_1.2,0.19642240524291993
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,8,16,power_law_1.2,0.2530790328979492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,8,16,power_law_1.2,0.11671168327331542
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,8,16,power_law_1.2,0.3037708854675293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,8,16,power_law_1.2,0.4085580825805664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,8,16,power_law_1.2,0.5442841720581055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,8,16,power_law_1.2,0.79499267578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,8,16,power_law_1.2,0.9581772613525391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,8,16,power_law_1.2,0.037962241172790526
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,8,16,power_law_1.2,0.039047679901123046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,8,16,power_law_1.2,0.03974656105041504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,8,16,power_law_1.2,0.03954560041427612
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,8,16,power_law_1.01,0.13259391784667968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,8,16,power_law_1.2,0.04967679977416992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,8,16,power_law_1.2,0.049724159240722654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,8,16,power_law_1.2,0.05123839855194092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,8,16,power_law_1.2,0.05422080039978028
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,8,16,power_law_1.2,0.052276477813720704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,8,16,power_law_1.2,0.05419904232025147
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,8,16,power_law_1.2,0.061679358482360835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,8,16,power_law_1.2,0.06370304107666017
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,8,16,power_law_1.2,0.05936895847320557
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,8,16,power_law_1.2,0.061643519401550294
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,8,16,power_law_1.2,0.06621568202972412
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,8,16,power_law_1.2,0.07016191959381105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,8,16,power_law_1.2,0.07583360195159912
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,8,16,power_law_1.2,0.0925055980682373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,8,16,power_law_1.2,0.11709823608398437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,8,16,power_law_1.2,0.14547967910766602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,8,16,power_law_1.2,0.17857023239135744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,8,16,power_law_1.2,0.24811008453369138
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,8,16,power_law_1.2,0.34323585510253907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,8,16,power_law_1.2,0.46101505279541016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,8,16,power_law_1.2,0.6532364654541015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,8,16,power_law_1.2,1.494978485107422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,8,16,power_law_1.2,0.022650880813598635
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,8,16,power_law_1.2,0.023493120670318605
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,8,16,power_law_1.2,0.023884799480438233
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,8,16,power_law_1.2,0.023595519065856933
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,8,16,power_law_1.2,0.02280319929122925
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,8,16,power_law_1.2,0.023582720756530763
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,8,16,power_law_1.2,1.1503616333007813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,8,16,power_law_1.2,0.023491840362548828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,8,16,power_law_1.2,0.024048640727996826
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,8,16,power_law_1.2,0.9160409545898437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,8,16,power_law_1.2,0.023481600284576416
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,8,16,power_law_1.2,0.02394495964050293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,8,16,power_law_1.2,0.024550399780273437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,8,16,power_law_1.2,0.025553920269012452
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,8,16,power_law_1.2,0.025213439464569092
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,8,16,power_law_1.2,0.026360321044921874
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,8,16,power_law_1.2,0.026019840240478515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,8,16,power_law_1.2,2.0896922302246095
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,8,16,power_law_1.01,0.14994303703308104
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,8,16,power_law_1.01,0.23008127212524415
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,8,16,power_law_1.01,0.2919283294677734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,8,16,power_law_1.2,0.038178560733795164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,8,16,power_law_1.2,0.044279041290283205
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,8,16,power_law_1.2,0.02852735996246338
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,8,16,power_law_1.2,0.03157887935638427
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,8,16,power_law_1.2,0.05484543800354004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,16,8,balanced,0.030840320587158205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,16,8,balanced,0.0301145601272583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,16,8,balanced,0.03212928056716919
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,16,8,balanced,0.035802879333496095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,16,8,balanced,0.062494721412658695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,16,8,balanced,0.11666048049926758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,16,8,balanced,0.1196774387359619
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,8,16,power_law_1.2,0.1093900775909424
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,8,16,power_law_1.2,0.07859456062316894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,16,8,balanced,0.12366463661193847
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,16,8,balanced,0.13058560371398925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,16,8,balanced,0.12898816108703612
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,16,8,balanced,0.12475775718688964
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,16,8,balanced,0.12798975944519042
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,16,8,balanced,0.1325708770751953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,16,8,balanced,0.13829888343811034
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,16,8,balanced,0.14068479537963868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,16,8,balanced,0.14672256469726563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,16,8,balanced,0.15992064476013185
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,16,8,balanced,0.18484607696533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,16,8,balanced,0.2118387222290039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,16,8,balanced,0.2660902404785156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,16,8,balanced,0.3156697654724121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,16,8,balanced,0.42077182769775395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,16,8,balanced,0.5242022323608398
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,8,16,power_law_1.2,0.08825087547302246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,16,8,balanced,0.7247462463378906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,16,8,balanced,0.03175807952880859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,16,8,balanced,0.9340608215332031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,16,8,balanced,1.3708390808105468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,16,8,balanced,0.03948415994644165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,16,8,balanced,0.0451532793045044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,16,8,balanced,0.06159615993499756
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,16,8,balanced,0.03388159990310669
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,16,8,balanced,1.8238105773925781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,16,8,balanced,0.06158976078033447
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,16,8,balanced,0.06311552047729492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,16,8,balanced,0.06686848163604736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,16,8,balanced,0.06705152034759522
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,16,8,balanced,0.06510591983795166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,16,8,balanced,0.06779391765594482
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,16,8,balanced,0.07178880214691162
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,16,8,balanced,0.07687808036804199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,16,8,balanced,0.07692416191101074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,16,8,balanced,0.08040960311889649
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,16,8,balanced,0.0898521614074707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,16,8,balanced,0.10530048370361329
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,16,8,balanced,0.12225279808044434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,16,8,balanced,0.15494784355163574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,16,8,balanced,0.18781824111938478
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,16,8,balanced,0.2657241630554199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,16,8,balanced,0.32950271606445314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,16,8,balanced,0.4695718383789062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,16,8,balanced,0.6229516983032226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,16,8,balanced,0.8864524841308594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,16,8,balanced,1.156332778930664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,8,16,power_law_1.2,0.18377727508544922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,8,16,power_law_1.2,0.2229055976867676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,16,8,balanced,0.036871678829193115
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,8,16,power_law_1.2,0.3394918441772461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,8,16,power_law_1.2,0.4587071990966797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,16,8,balanced,0.0210534405708313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,16,8,balanced,0.0316595196723938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,16,8,balanced,0.03248768091201783
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,16,8,balanced,0.03258879899978638
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,16,8,balanced,0.03239808082580566
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,16,8,balanced,0.03253632068634033
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,16,8,balanced,0.03257728099822998
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,16,8,balanced,0.0229913592338562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,16,8,balanced,0.021429760456085207
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,16,8,balanced,0.03305088043212891
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,16,8,balanced,0.03474816083908081
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,16,8,balanced,0.035825920104980466
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,16,8,balanced,0.03920383930206299
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,16,8,balanced,0.04119167804718017
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,16,8,balanced,0.046648321151733396
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,16,8,balanced,0.03568896055221558
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,16,8,balanced,0.02398848056793213
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,16,8,balanced,0.03141504049301148
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,16,8,balanced,0.03109119892120361
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,16,8,balanced,0.06387584209442139
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,16,8,power_law_1.01,0.0362611198425293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,16,8,power_law_1.01,0.056332798004150385
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,16,8,balanced,0.09491840362548828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,16,8,power_law_1.01,0.0562713623046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,16,8,power_law_1.01,0.060697598457336424
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,16,8,balanced,0.11437055587768555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,16,8,power_law_1.01,0.06477952003479004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,16,8,power_law_1.01,0.1153587245941162
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,16,8,power_law_1.01,0.11610112190246583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,16,8,power_law_1.01,0.11898112297058105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,16,8,power_law_1.01,0.12492799758911133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,16,8,power_law_1.01,0.12567808151245116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,16,8,power_law_1.01,0.13240447998046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,16,8,power_law_1.01,0.13025919914245604
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,16,8,power_law_1.01,0.12807552337646483
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,16,8,power_law_1.01,0.1341875171661377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,16,8,power_law_1.01,0.1377894401550293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,16,8,power_law_1.01,0.14224767684936523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,16,8,power_law_1.01,0.15919360160827636
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,16,8,power_law_1.01,0.17626367568969728
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,16,8,power_law_1.01,0.20675712585449219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,16,8,power_law_1.01,0.23477119445800781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,16,8,power_law_1.01,0.2853593635559082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,16,8,balanced,0.15340415954589842
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,16,8,power_law_1.01,0.3824576187133789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,16,8,power_law_1.01,0.44855808258056645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,16,8,power_law_1.01,0.6684851074218751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,16,8,power_law_1.01,0.037067520618438723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,16,8,power_law_1.01,0.8193958282470704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,16,8,power_law_1.01,0.04188543796539307
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,16,8,power_law_1.01,0.04135551929473877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,16,8,power_law_1.01,1.3083815002441406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,16,8,power_law_1.01,0.0444377613067627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,16,8,power_law_1.01,0.06043647766113282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,16,8,power_law_1.01,0.060762882232666016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,16,8,power_law_1.01,0.06221824169158936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,16,8,power_law_1.01,1.7445082092285156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,16,8,power_law_1.01,0.06576767921447754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,16,8,power_law_1.01,0.06378367900848389
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,16,8,power_law_1.01,0.06965888023376465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,16,8,power_law_1.01,0.07352831840515137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,16,8,power_law_1.01,0.06861055850982666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,16,8,power_law_1.01,0.07339392185211183
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,16,8,power_law_1.01,0.0842304039001465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,16,8,power_law_1.01,0.07812992095947266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,16,8,power_law_1.01,0.04552703857421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,16,8,power_law_1.01,0.10319104194641113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,16,8,power_law_1.01,0.11717375755310058
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,16,8,power_law_1.01,0.15870335578918457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,16,8,power_law_1.01,0.18515199661254883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,16,8,power_law_1.01,0.2523391914367676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,16,8,power_law_1.01,0.3258086395263672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,16,8,power_law_1.01,0.07633791923522949
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,16,8,power_law_1.01,0.4655142211914063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,16,8,power_law_1.01,0.5953267288208008
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,16,8,power_law_1.01,0.024211199283599855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,16,8,power_law_1.01,0.9149311828613282
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,16,8,power_law_1.01,0.02759167909622192
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,16,8,power_law_1.01,0.026759679317474368
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,16,8,power_law_1.01,0.029012479782104493
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,16,8,balanced,0.1961996841430664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,16,8,power_law_1.01,1.0882406616210938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,16,8,power_law_1.01,0.03116544008255005
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,16,8,power_law_1.01,0.029648640155792237
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,16,8,power_law_1.01,0.03223680019378662
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,16,8,power_law_1.01,0.03130880117416382
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,16,8,power_law_1.01,0.03215359926223755
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,16,8,power_law_1.01,0.032395520210266114
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,16,8,power_law_1.01,0.034214398860931396
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,16,8,power_law_1.01,0.034542078971862795
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,16,8,power_law_1.01,0.036469759941101076
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,16,8,power_law_1.01,0.02886784076690674
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,16,8,power_law_1.01,0.03314687967300415
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,16,8,power_law_1.01,0.056492800712585445
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,16,8,power_law_1.01,0.07362559795379639
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,16,8,power_law_1.01,0.09016703605651856
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,16,8,power_law_1.01,0.04285056114196777
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,16,8,power_law_1.01,0.12153216361999512
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,16,8,power_law_1.01,0.13742207527160644
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,16,8,power_law_1.01,0.17119232177734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,16,8,power_law_1.01,0.21940607070922852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,16,8,power_law_1.2,0.034124801158905024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,16,8,power_law_1.2,0.05153920173645019
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,16,8,power_law_1.2,0.05616511821746826
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,16,8,power_law_1.2,0.05820543766021728
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,16,8,power_law_1.2,0.0644543981552124
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,16,8,power_law_1.01,0.39435775756835934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,16,8,power_law_1.2,0.11445119857788086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,16,8,power_law_1.2,0.11731072425842286
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,16,8,power_law_1.2,0.1197555160522461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,16,8,power_law_1.2,0.12308095932006835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,16,8,power_law_1.2,0.12665727615356445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,16,8,power_law_1.2,0.12928895950317382
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,16,8,power_law_1.2,0.13003520011901853
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,16,8,power_law_1.2,0.136627197265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,16,8,power_law_1.2,0.13928704261779784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,16,8,power_law_1.2,0.14279040336608886
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,16,8,power_law_1.2,0.15732480049133302
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,16,8,power_law_1.2,0.18126848220825195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,16,8,power_law_1.2,0.19966720581054687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,16,8,power_law_1.2,0.24114048004150393
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,16,8,power_law_1.2,0.2955379295349121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,16,8,power_law_1.2,0.41260929107666017
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,16,8,power_law_1.2,0.5417049789428712
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,16,8,power_law_1.2,0.7666636657714844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,16,8,power_law_1.2,1.03870849609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,16,8,power_law_1.2,1.488019256591797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,16,8,power_law_1.2,0.13137151718139647
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,16,8,power_law_1.2,0.036988160610198974
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,16,8,power_law_1.2,0.042516479492187495
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,16,8,power_law_1.2,0.044304637908935546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,16,8,power_law_1.2,0.04711552143096924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,16,8,power_law_1.2,0.059613437652587885
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,16,8,power_law_1.2,0.05982336044311524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,16,8,power_law_1.2,0.06237184047698975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,16,8,power_law_1.2,0.06578048229217529
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,16,8,power_law_1.2,0.06625152111053467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,16,8,power_law_1.2,0.06722815990447997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,16,8,power_law_1.2,0.06934144020080565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,16,8,power_law_1.2,0.06949888229370117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,16,8,power_law_1.2,0.07428351879119874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,16,8,power_law_1.2,0.07600384235382081
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,16,8,power_law_1.2,0.07926911830902099
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,16,8,power_law_1.2,0.08742015838623048
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,16,8,power_law_1.2,0.10471296310424805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,16,8,power_law_1.2,0.12430591583251953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,16,8,power_law_1.2,0.16363391876220704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,16,8,power_law_1.2,0.1879987144470215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,16,8,power_law_1.2,0.2791321563720703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,16,8,power_law_1.2,0.03856127977371216
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,16,8,power_law_1.2,0.3594611358642578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,16,8,power_law_1.2,0.4878860855102539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,16,8,power_law_1.2,0.6768013000488281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,16,8,power_law_1.2,1.9845619201660156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,16,8,power_law_1.2,0.02436352014541626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,16,8,power_law_1.2,0.025374720096588133
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,16,8,power_law_1.2,0.027060480117797853
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,16,8,power_law_1.2,0.028811519145965574
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,16,8,power_law_1.2,0.031783680915832516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,16,8,power_law_1.2,0.9845747375488282
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,16,8,power_law_1.2,0.03192831993103028
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,16,8,power_law_1.2,0.03339904069900513
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,16,8,power_law_1.2,0.033077759742736815
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,16,8,power_law_1.2,0.03420543909072876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,16,8,power_law_1.2,0.03394047975540161
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,16,8,power_law_1.2,0.0369267201423645
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,16,8,power_law_1.2,0.039441919326782225
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,16,8,power_law_1.2,0.04130815982818604
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,16,8,power_law_1.2,0.04606336116790771
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,16,8,power_law_1.2,1.2763417816162108
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,16,8,power_law_1.2,0.05643136024475097
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,16,8,power_law_1.2,0.07264639854431151
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,16,8,power_law_1.2,0.03172863960266113
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,16,8,power_law_1.2,0.02971519947052002
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,16,8,power_law_1.2,0.026696960926055908
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,16,8,power_law_1.2,0.18277503967285155
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,16,8,power_law_1.2,0.10840576171875001
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,16,8,power_law_1.2,0.6322854232788087
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,16,8,power_law_1.2,0.2660121536254883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,32,4,balanced,0.03259135961532593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,32,4,balanced,0.04736767768859863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,32,4,balanced,0.06873472213745117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,32,4,balanced,0.08119680404663085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,32,4,balanced,0.13054976463317872
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,16,8,power_law_1.2,0.3784435272216797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,32,4,balanced,0.13459839820861816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,32,4,balanced,0.13495424270629883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,32,4,balanced,0.14014975547790526
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,32,4,balanced,0.13415295600891114
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,32,4,balanced,0.138024959564209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,32,4,balanced,0.1431488037109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,32,4,balanced,0.14833919525146483
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,32,4,balanced,0.15089792251586914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,32,4,balanced,0.15753727912902832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,32,4,balanced,0.16969600677490235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,32,4,balanced,0.1957004737854004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,32,4,balanced,0.22123136520385742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,32,4,balanced,0.2766323280334473
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,32,4,balanced,0.3274585723876953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,32,4,balanced,0.43012737274169927
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,32,4,balanced,0.5367283248901368
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,32,4,balanced,0.7334604644775391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,32,4,balanced,0.9696857452392578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,32,4,balanced,1.4237338256835936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,32,4,balanced,0.03304831981658936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,32,4,balanced,0.036231679916381834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,32,4,balanced,1.8430540466308591
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,32,4,balanced,0.04235775947570801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,32,4,balanced,0.04849535942077637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,32,4,balanced,0.057145600318908696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,32,4,balanced,0.06409088134765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,32,4,balanced,0.07636608123779297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,32,4,balanced,0.03360512018203735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,32,4,balanced,0.08004223823547363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,32,4,balanced,0.07798912048339843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,32,4,balanced,0.07751167774200439
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,32,4,balanced,0.0818227195739746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,32,4,balanced,0.08140288352966309
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,32,4,balanced,0.0837183952331543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,32,4,balanced,0.08975104331970216
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,32,4,balanced,0.09070079803466798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,32,4,balanced,0.09497856140136719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,32,4,balanced,0.10269824028015137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,32,4,balanced,0.11988479614257812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,32,4,balanced,0.13410816192626954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,32,4,balanced,0.1644108772277832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,32,4,balanced,0.19613056182861327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,32,4,balanced,0.034600958824157715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,32,4,balanced,0.2847462463378906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,32,4,balanced,0.34583168029785155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,32,4,balanced,0.48941184997558596
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,32,4,balanced,0.6550656127929687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,32,4,balanced,0.9177433776855469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,32,4,balanced,1.2134655761718751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,32,4,power_law_1.01,0.05673088073730469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,32,4,power_law_1.01,0.03717504024505615
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,32,4,power_law_1.01,0.06208255767822266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,32,4,power_law_1.01,0.0682431983947754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,32,4,power_law_1.01,0.12712575912475585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,32,4,power_law_1.01,0.07894911766052246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,32,4,power_law_1.01,0.13137408256530764
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,32,4,power_law_1.01,0.03631103992462158
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,32,4,power_law_1.01,0.1344806385040283
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,32,4,power_law_1.01,0.14289407730102538
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,32,4,power_law_1.01,0.14698880195617675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,32,4,power_law_1.01,0.1395456027984619
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,32,4,power_law_1.01,0.1490892791748047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,32,4,power_law_1.01,0.15692031860351563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,32,4,power_law_1.01,0.17125375747680666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,32,4,power_law_1.01,0.19656831741333008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,32,4,power_law_1.01,0.22161407470703126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,32,4,power_law_1.01,0.2749785614013672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,32,4,power_law_1.01,0.3189760017395019
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,32,4,power_law_1.01,0.43011711120605467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,32,4,power_law_1.01,0.5214835357666016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,32,4,power_law_1.01,0.13560319900512696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,32,4,power_law_1.01,0.745327377319336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,32,4,power_law_1.01,0.9767027282714844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,32,4,power_law_1.01,1.470055694580078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,32,4,power_law_1.01,0.03796479940414429
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,32,4,power_law_1.01,0.13733887672424316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,32,4,power_law_1.01,1.8956710815429687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,32,4,power_law_1.01,0.041521921157836914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,32,4,power_law_1.01,0.04567552089691162
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,32,4,power_law_1.01,0.061916160583496097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,32,4,power_law_1.01,0.05180287837982178
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,32,4,power_law_1.01,0.05697535991668701
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,32,4,power_law_1.01,0.0748799991607666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,32,4,power_law_1.01,0.07669760227203369
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,32,4,power_law_1.01,0.07734272003173828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,32,4,power_law_1.01,0.08077695846557617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,32,4,power_law_1.01,0.08170623779296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,32,4,power_law_1.01,0.08900351524353027
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,32,4,power_law_1.01,0.09378944396972656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,32,4,power_law_1.01,0.08311807632446289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,32,4,power_law_1.01,0.09028223991394044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,32,4,power_law_1.01,0.10426239967346192
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,32,4,power_law_1.01,0.14143872261047363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,32,4,power_law_1.01,0.17609088897705077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,32,4,power_law_1.01,0.08098943710327149
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,32,4,power_law_1.01,0.21035135269165037
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,32,4,power_law_1.01,0.2923456001281738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,32,4,power_law_1.01,0.12307968139648437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,32,4,power_law_1.01,0.3479923248291016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,32,4,power_law_1.01,0.5173273468017578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,32,4,power_law_1.01,0.6576742553710938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,32,4,power_law_1.2,0.036614398956298824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,32,4,power_law_1.01,0.9701798248291016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,32,4,power_law_1.01,1.2434368133544922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,32,4,power_law_1.2,0.05556863784790039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,32,4,power_law_1.2,0.04106624126434326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,32,4,power_law_1.2,0.06209663867950439
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,32,4,power_law_1.2,0.06574975967407226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,32,4,power_law_1.2,0.13060223579406738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,32,4,power_law_1.2,0.13454719543457033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,32,4,power_law_1.2,0.12678784370422364
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,32,4,power_law_1.2,0.07806464195251464
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,32,4,power_law_1.2,0.1364787197113037
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,32,4,power_law_1.2,0.14099072456359862
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,32,4,power_law_1.2,0.138155517578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,32,4,power_law_1.2,0.14088704109191894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,32,4,power_law_1.2,0.14717056274414062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,32,4,power_law_1.2,0.15034239768981933
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,32,4,power_law_1.2,0.15963647842407228
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,32,4,power_law_1.2,0.19906816482543946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,32,4,power_law_1.2,0.17272832870483398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,32,4,power_law_1.2,0.22402816772460937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,32,4,power_law_1.2,0.28042112350463866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,32,4,power_law_1.2,0.33049217224121097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,32,4,power_law_1.2,0.4314675140380859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,32,4,power_law_1.2,0.5546905517578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,32,4,power_law_1.2,0.7773811340332031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,32,4,power_law_1.2,0.03699968099594116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,32,4,power_law_1.2,0.03988735914230347
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,32,4,power_law_1.2,1.0156979370117187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,32,4,power_law_1.2,0.04566527843475342
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,32,4,power_law_1.2,0.05094912052154541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,32,4,power_law_1.2,0.05408127784729004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,32,4,power_law_1.2,0.07352191925048827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,32,4,power_law_1.2,0.07639039993286133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,32,4,power_law_1.2,0.07834879875183105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,32,4,power_law_1.2,0.07977344036102295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,32,4,power_law_1.2,0.08206720352172851
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,32,4,power_law_1.2,0.08039679527282714
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,32,4,power_law_1.2,0.08464768409729004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,32,4,power_law_1.2,0.09036160469055175
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,32,4,power_law_1.2,0.09100288391113281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,32,4,power_law_1.2,0.09590527534484863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,32,4,power_law_1.2,1.4837478637695312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,32,4,power_law_1.2,0.10683775901794434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,32,4,power_law_1.2,0.12544896125793456
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,32,4,power_law_1.2,0.14364928245544434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,32,4,power_law_1.2,0.21689983367919924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,32,4,power_law_1.2,0.17939840316772462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,32,4,power_law_1.2,0.061514239311218265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,32,4,power_law_1.2,0.2990656089782715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,32,4,power_law_1.2,0.3682009506225586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,32,4,power_law_1.2,0.5380467224121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,32,4,power_law_1.2,1.0020582580566406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,32,4,power_law_1.2,0.6826764678955077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,128,balanced,0.07094655990600586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,128,balanced,0.06741119861602783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,128,balanced,0.07105792045593262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,128,balanced,0.0674892807006836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,128,balanced,0.07076352119445802
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,128,balanced,0.07786880016326905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,128,balanced,0.08739711761474608
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,128,balanced,0.14411264419555664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,128,balanced,0.1436633586883545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,128,balanced,0.1441100788116455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,128,balanced,0.1459776020050049
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,128,balanced,0.1463372802734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,128,balanced,0.14661503791809083
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,128,balanced,0.14951552391052245
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,128,balanced,0.15027071952819823
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,128,balanced,0.15082624435424805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,128,balanced,0.15431808471679687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,128,balanced,0.1598579216003418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,128,balanced,0.1646156883239746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,128,balanced,0.17334655761718748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,128,balanced,0.1747724723815918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,128,balanced,0.18097280502319335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,128,balanced,0.19773567199707032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,128,balanced,0.21399551391601562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,32,4,power_law_1.2,1.3091813659667968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,128,balanced,0.25464448928833006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,128,balanced,0.32528766632080075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,128,balanced,0.03834239959716797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,128,balanced,0.04116991996765137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,128,balanced,0.44371585845947265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,128,balanced,0.04124671936035156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,32,4,power_law_1.2,1.9683187866210936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,128,balanced,0.041141757965087886
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,128,balanced,0.054145278930664065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,128,balanced,0.05924992084503174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,128,balanced,0.05971968173980713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,128,balanced,0.03752448081970215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,128,balanced,0.05940735816955567
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,128,balanced,0.0603110408782959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,128,balanced,0.04536448001861572
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,128,balanced,0.060824317932128905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,128,balanced,0.06958720207214356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,128,balanced,0.06357247829437256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,128,balanced,0.07213952064514159
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,128,balanced,0.060697598457336424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,128,balanced,0.06861440181732179
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,128,balanced,0.0713804817199707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,128,balanced,0.08013440132141113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,128,balanced,0.07435391902923584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,128,balanced,0.08131711959838868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,128,balanced,0.08580351829528808
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,128,balanced,0.09960703849792481
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,128,balanced,0.11686528205871582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,128,balanced,0.14576383590698244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,128,balanced,0.06596352100372314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,128,balanced,0.17669631958007812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,1,128,balanced,0.020276479721069336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,1,128,balanced,0.020285439491271973
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,1,128,balanced,0.021456639766693115
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,1,128,balanced,0.025555200576782226
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,1,128,balanced,0.02543872117996216
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,1,128,balanced,0.025703680515289307
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,1,128,balanced,0.02592639923095703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,1,128,balanced,0.020590078830718995
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,1,128,balanced,0.025946879386901857
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,1,128,balanced,0.025794560909271243
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,1,128,balanced,0.026027519702911377
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,1,128,balanced,0.025941760540008546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,1,128,balanced,0.02634239912033081
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,1,128,balanced,0.026338560581207277
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,1,128,balanced,0.02678015947341919
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,1,128,balanced,0.02681472063064575
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,1,128,balanced,0.025921280384063723
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,1,128,balanced,0.028314878940582277
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,1,128,balanced,0.02807039976119995
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,1,128,balanced,0.029621760845184326
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,1,128,balanced,0.025597438812255856
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,1,128,balanced,0.020673279762268064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,128,power_law_1.01,0.08982272148132324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,128,power_law_1.01,0.08706175804138183
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,128,power_law_1.01,0.08826239585876465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,128,power_law_1.01,0.08836352348327636
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,128,power_law_1.01,0.14179967880249023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,128,power_law_1.01,0.14259327888488768
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,128,power_law_1.01,0.1467520046234131
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,128,power_law_1.01,0.14725119590759278
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,128,power_law_1.01,0.14825471878051757
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,128,power_law_1.01,0.1487705612182617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,128,power_law_1.01,0.15039360046386718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,128,power_law_1.01,0.15218815803527833
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,128,power_law_1.01,0.1533414363861084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,128,power_law_1.01,0.1621683120727539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,128,power_law_1.01,0.16344959259033204
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,1,128,balanced,0.019856640100479127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,128,power_law_1.01,0.16873472213745117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,128,power_law_1.01,0.1854924774169922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,128,power_law_1.01,0.19249792098999025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,128,power_law_1.01,0.24637695312500002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,128,power_law_1.01,0.27824768066406247
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,128,power_law_1.01,0.34754558563232424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,128,power_law_1.01,0.4620159912109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,128,power_law_1.01,0.6530585479736328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,128,power_law_1.01,0.7443583679199219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,128,power_law_1.01,1.1976358032226562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,128,power_law_1.01,0.17341567993164062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,128,power_law_1.01,1.5606591796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,128,power_law_1.01,0.05392640113830567
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,128,power_law_1.01,0.05368959903717041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,128,power_law_1.01,0.05162240028381347
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,128,power_law_1.01,0.05399807929992676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,128,power_law_1.01,0.05868800163269043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,128,power_law_1.01,0.06021376132965088
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,128,power_law_1.01,0.06096767902374267
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,128,power_law_1.01,0.058709759712219235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,128,power_law_1.01,0.06138239860534668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,128,power_law_1.01,0.06288512229919434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,128,power_law_1.01,0.06301824092864991
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,128,power_law_1.01,0.06648320198059082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,128,power_law_1.01,0.06612736225128174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,128,power_law_1.01,0.06950528144836425
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,128,power_law_1.01,0.0729472017288208
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,128,power_law_1.01,0.0776255989074707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,128,power_law_1.01,0.08231679916381836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,128,power_law_1.01,0.08114560127258301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,128,power_law_1.01,0.08824064254760741
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,128,power_law_1.01,0.10210687637329101
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,128,power_law_1.01,0.12277248382568359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,128,power_law_1.01,0.16583423614501952
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,128,power_law_1.01,0.21044607162475587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,128,power_law_1.01,0.24970367431640622
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,128,power_law_1.01,0.3411801528930664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,128,power_law_1.01,0.5488063812255859
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,1,128,power_law_1.01,0.02531327962875366
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,1,128,power_law_1.01,0.02549247980117798
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,1,128,power_law_1.01,0.025651199817657473
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,1,128,power_law_1.01,0.02351360082626343
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,1,128,power_law_1.01,0.024350719451904295
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,1,128,power_law_1.01,0.025382399559020996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,128,power_law_1.01,0.7052352142333984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,1,128,power_law_1.01,0.025560319423675537
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,1,128,power_law_1.01,0.025282559394836424
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,1,128,power_law_1.01,0.02519808053970337
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,1,128,power_law_1.01,0.025612800121307372
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,1,128,power_law_1.01,0.025601279735565186
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,1,128,power_law_1.01,0.025952000617980957
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,1,128,power_law_1.01,0.025477120876312254
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,1,128,power_law_1.01,0.026314239501953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,1,128,power_law_1.01,0.026017279624938966
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,1,128,power_law_1.01,0.026682879924774174
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,1,128,power_law_1.01,0.026869759559631345
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,1,128,balanced,0.045987839698791506
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,1,128,balanced,0.052645120620727535
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,1,128,balanced,0.0442303991317749
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,1,128,balanced,0.02522239923477173
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,1,128,power_law_1.01,0.032838399410247805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,1,128,power_law_1.01,0.040586237907409665
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,1,128,power_law_1.01,0.043726081848144534
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,1,128,power_law_1.01,0.024024319648742676
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,1,128,power_law_1.01,0.054493441581726074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,128,power_law_1.2,0.08687744140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,128,power_law_1.2,0.08678527832031249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,128,power_law_1.2,0.08757247924804687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,128,power_law_1.2,0.14017919540405271
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,128,power_law_1.2,0.14655360221862793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,128,power_law_1.2,0.14223487854003908
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,1,128,power_law_1.01,0.0665228796005249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,128,power_law_1.2,0.15123968124389647
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,128,power_law_1.2,0.14720255851745606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,128,power_law_1.2,0.15060607910156248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,128,power_law_1.2,0.16145792007446289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,128,power_law_1.2,0.15179903984069826
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,128,power_law_1.2,0.15898367881774902
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,128,power_law_1.2,0.16620288848876955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,128,power_law_1.2,0.17201919555664064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,128,power_law_1.2,0.18430335998535158
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,128,power_law_1.2,0.17668224334716798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,128,power_law_1.2,0.19118976593017578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,128,power_law_1.2,0.211592960357666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,128,power_law_1.2,0.27253887176513675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,128,power_law_1.2,0.32783233642578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,128,power_law_1.2,0.4839155197143555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,128,power_law_1.2,0.6494169616699219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,128,power_law_1.2,0.9077196502685547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,128,power_law_1.2,1.2034623718261719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,128,power_law_1.2,0.053324799537658694
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,128,power_law_1.2,0.054097919464111334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,128,power_law_1.2,0.08565504074096679
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,128,power_law_1.2,1.921201934814453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,128,power_law_1.2,0.0555072021484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,1,128,power_law_1.01,0.08198016166687012
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,128,power_law_1.2,0.05477375984191894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,128,power_law_1.2,0.0586240005493164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,128,power_law_1.2,0.06036736011505127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,128,power_law_1.2,0.06756608009338379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,128,power_law_1.2,0.06408959865570069
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,128,power_law_1.2,0.07171328067779541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,128,power_law_1.2,2.8180096435546877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,128,power_law_1.2,0.052067837715148925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,128,power_law_1.2,0.06916224002838135
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,128,power_law_1.2,0.07208191871643066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,128,power_law_1.2,0.0774515199661255
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,128,power_law_1.2,0.07795584201812744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,128,power_law_1.2,0.07924736022949219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,128,power_law_1.2,0.08665472030639648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,128,power_law_1.2,0.0971008014678955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,128,power_law_1.2,0.11966591835021974
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,128,power_law_1.2,0.1379520034790039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,128,power_law_1.2,0.19614208221435547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,128,power_law_1.2,0.06701056003570557
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,128,power_law_1.2,0.24306560516357423
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,128,power_law_1.2,0.3660108947753906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,1,128,power_law_1.2,0.025367040634155274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,128,power_law_1.2,0.5685478210449219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,1,128,power_law_1.2,0.02563584089279175
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,128,power_law_1.2,0.8333567810058593
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,1,128,power_law_1.2,0.02573440074920654
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,1,128,power_law_1.2,0.025420799255371097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,128,power_law_1.2,1.2010598754882813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,1,128,power_law_1.2,0.02397439956665039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,1,128,power_law_1.2,0.025487360954284666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,1,128,power_law_1.2,0.024358398914337158
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,1,128,power_law_1.2,0.025539839267730714
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,1,128,power_law_1.2,0.02601471900939941
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,1,128,power_law_1.2,0.025026559829711914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,1,128,power_law_1.2,0.026357760429382326
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,1,128,power_law_1.2,0.025539839267730714
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,1,128,power_law_1.2,0.026654720306396484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,1,128,power_law_1.2,0.026841599941253663
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,1,128,power_law_1.2,0.026356480121612548
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,1,128,power_law_1.2,0.033788158893585205
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,1,128,power_law_1.2,0.033784320354461675
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,1,128,power_law_1.2,0.02597120046615601
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,1,128,power_law_1.2,0.027471361160278318
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,1,128,power_law_1.01,0.10229120254516602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,128,power_law_1.2,0.058847999572753905
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,1,128,power_law_1.01,0.14994048118591308
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,1,128,power_law_1.01,0.1872742462158203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,1,128,power_law_1.2,0.04936063766479492
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,1,128,power_law_1.2,0.07212160110473634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,64,balanced,0.03756288051605224
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,64,balanced,0.03789056062698364
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,1,128,power_law_1.2,0.04526847839355469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,64,balanced,0.03790719985961914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,64,balanced,0.035443201065063476
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,1,128,power_law_1.2,0.08036992073059082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,64,balanced,0.07468031883239747
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,64,balanced,0.087575044631958
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,64,balanced,0.13843968391418457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,64,balanced,0.13575551986694337
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,64,balanced,0.1428492832183838
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,64,balanced,0.14116352081298827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,64,balanced,0.1477132797241211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,64,balanced,0.1426585578918457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,64,balanced,0.1456704044342041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,64,balanced,0.14507264137268067
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,64,balanced,0.06157440185546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,64,balanced,0.15981568336486815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,64,balanced,0.1576460838317871
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,1,128,power_law_1.2,0.10530816078186037
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,64,balanced,0.16645631790161133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,64,balanced,0.1769331169128418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,64,balanced,0.19045375823974608
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,64,balanced,0.21924480438232422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,64,balanced,0.25357183456420895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,64,balanced,0.29567615509033207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,64,balanced,0.3740121459960938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,64,balanced,0.1517311954498291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,64,balanced,0.03543168067932129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,64,balanced,0.036536319255828856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,64,balanced,0.03532799959182739
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,64,balanced,0.035397119522094726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,64,balanced,0.03754879951477051
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,64,balanced,0.04558847904205322
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,64,balanced,0.05632256031036377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,64,balanced,0.06457344055175782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,64,balanced,0.0678656005859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,64,balanced,0.06414591789245605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,64,balanced,0.06643712043762207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,64,balanced,0.07423359870910644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,64,balanced,0.06657279968261719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,64,balanced,0.06819712162017823
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,64,balanced,0.07166079998016357
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,64,balanced,0.07376768112182616
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,64,balanced,0.07771520137786865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,64,balanced,0.5291763305664062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,64,balanced,0.07362559795379639
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,64,balanced,0.07794303894042968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,64,balanced,0.08148863792419433
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,64,balanced,0.09309439659118653
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,64,balanced,0.1029747200012207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,64,balanced,0.1258687973022461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,64,balanced,0.15818495750427247
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,64,balanced,0.19698944091796874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,64,balanced,0.25227392196655274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,64,balanced,0.0693171215057373
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,1,128,power_law_1.2,0.15159040451049804
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,1,128,power_law_1.2,0.23708288192749025
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,1,128,power_law_1.2,0.27184768676757814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,64,balanced,0.13403519630432129
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,2,64,balanced,0.02039936065673828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,2,64,balanced,0.02025984048843384
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,2,64,balanced,0.02022144079208374
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,2,64,balanced,0.02642047882080078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,2,64,balanced,0.026894080638885497
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,2,64,balanced,0.026789119243621828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,2,64,balanced,0.027384319305419923
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,2,64,balanced,0.027107839584350583
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,2,64,balanced,0.026918399333953857
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,2,64,balanced,0.026757121086120605
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,2,64,balanced,0.027178239822387696
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,2,64,balanced,0.027229440212249757
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,2,64,balanced,0.028039679527282712
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,2,64,balanced,0.02021503925323486
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,2,64,balanced,0.02840447902679443
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,2,64,balanced,0.026421759128570553
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,2,64,balanced,0.03004928112030029
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,2,64,balanced,0.030924799442291262
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,2,64,balanced,0.03336319923400879
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,2,64,balanced,0.0279040002822876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,64,power_law_1.01,0.08463487625122071
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,64,power_law_1.01,0.08477439880371093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,64,power_law_1.01,0.08560768127441407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,64,power_law_1.01,0.08768768310546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,64,power_law_1.01,0.1329075241088867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,2,64,balanced,0.020490241050720216
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,64,power_law_1.01,0.13140224456787108
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,64,power_law_1.01,0.1356428813934326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,64,power_law_1.01,0.1347532844543457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,64,power_law_1.01,0.13779071807861326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,64,power_law_1.01,0.13903743743896485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,64,power_law_1.01,0.14180095672607423
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,64,power_law_1.01,0.14463104248046874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,64,power_law_1.01,0.14695296287536622
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,64,power_law_1.01,0.15298432350158692
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,64,power_law_1.01,0.15843071937561035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,64,power_law_1.01,0.1632601547241211
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,2,64,balanced,0.022705919742584228
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,64,power_law_1.01,0.161079044342041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,64,power_law_1.01,0.1777574348449707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,64,power_law_1.01,0.19338367462158204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,64,power_law_1.01,0.22473472595214844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,64,power_law_1.01,0.25961599349975584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,2,64,balanced,0.02697472095489502
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,64,power_law_1.01,0.2802022361755371
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,2,64,balanced,0.027148799896240233
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,64,power_law_1.01,0.4100799942016602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,64,power_law_1.01,0.5088703918457032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,64,power_law_1.01,0.05581056118011475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,64,power_law_1.01,0.7991398620605469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,64,power_law_1.01,0.9493875122070312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,64,power_law_1.01,0.05510015964508057
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,64,power_law_1.01,0.05695871829986572
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,64,power_law_1.01,0.05670911788940429
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,64,power_law_1.01,0.06291071891784668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,64,power_law_1.01,1.3039949035644531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,64,power_law_1.01,0.05908991813659668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,64,power_law_1.01,0.05639808177947998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,64,power_law_1.01,0.06448768138885498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,64,power_law_1.01,0.06507904052734374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,64,power_law_1.01,0.06543615818023682
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,64,power_law_1.01,0.06940671920776367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,64,power_law_1.01,0.06928768157958984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,64,power_law_1.01,0.07360127925872803
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,64,power_law_1.01,0.07842688083648681
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,64,power_law_1.01,0.07793407917022706
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,64,power_law_1.01,0.07390207767486572
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,64,power_law_1.01,0.08206463813781738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,64,power_law_1.01,0.08955648422241211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,64,power_law_1.01,0.10645376205444337
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,64,power_law_1.01,0.1681727981567383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,64,power_law_1.01,0.12862719535827635
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,64,power_law_1.01,0.21386240005493162
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,64,power_law_1.01,0.06518015861511231
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,64,power_law_1.01,0.27656959533691405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,64,power_law_1.01,0.33659263610839846
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,2,64,power_law_1.01,0.026726400852203368
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,2,64,power_law_1.01,0.026403839588165286
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,64,power_law_1.01,0.5397900772094727
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,2,64,power_law_1.01,0.026353919506073
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,2,64,power_law_1.01,0.02649087905883789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,2,64,power_law_1.01,0.02493311882019043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,2,64,power_law_1.01,0.02584575891494751
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,2,64,power_law_1.01,0.02641535997390747
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,2,64,power_law_1.01,0.02637696027755737
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,2,64,power_law_1.01,0.026356480121612548
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,2,64,power_law_1.01,0.02667135953903198
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,2,64,power_law_1.01,0.025969920158386228
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,2,64,power_law_1.01,0.025657598972320554
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,2,64,power_law_1.01,0.02758143901824951
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,2,64,power_law_1.01,0.026085119247436523
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,2,64,power_law_1.01,0.02646912097930908
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,2,64,power_law_1.01,0.02758271932601929
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,2,64,power_law_1.01,0.02901760101318359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,64,power_law_1.01,0.6517823791503906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,2,64,power_law_1.01,0.0253657603263855
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,2,64,balanced,0.05170048236846923
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,2,64,balanced,0.06125823974609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,2,64,balanced,0.04889599800109863
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,2,64,power_law_1.01,0.03293567895889282
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,2,64,power_law_1.01,0.03746815919876099
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,2,64,power_law_1.01,0.04473855972290039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,64,power_law_1.2,0.08386816024780273
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,2,64,power_law_1.01,0.05435391902923584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,2,64,power_law_1.01,0.06195839881896973
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,64,power_law_1.2,0.08366975784301758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,64,power_law_1.2,0.08470656394958495
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,64,power_law_1.2,0.08668800354003905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,64,power_law_1.2,0.13267711639404295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,64,power_law_1.2,0.13090047836303711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,64,power_law_1.2,0.13373951911926268
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,64,power_law_1.2,0.13436032295227052
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,64,power_law_1.2,0.13890048027038576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,64,power_law_1.2,0.13685759544372558
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,64,power_law_1.2,0.14149375915527346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,64,power_law_1.2,0.14330623626708985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,64,power_law_1.2,0.1495961570739746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,64,power_law_1.2,0.15521280288696288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,64,power_law_1.2,0.15929216384887696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,64,power_law_1.2,0.16369279861450195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,64,power_law_1.2,0.1688640022277832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,64,power_law_1.2,0.18369024276733398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,64,power_law_1.2,0.1975641632080078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,64,power_law_1.2,0.24067712783813477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,64,power_law_1.2,0.3128870391845703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,64,power_law_1.2,0.4107993698120117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,64,power_law_1.2,0.4866265487670899
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,64,power_law_1.2,0.7135104370117188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,64,power_law_1.2,0.05608320236206055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,64,power_law_1.2,0.989364471435547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,64,power_law_1.2,0.05629695892333984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,64,power_law_1.2,1.5012786865234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,64,power_law_1.2,0.05678847789764404
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,64,power_law_1.2,0.05661695957183838
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,64,power_law_1.2,0.05890048027038575
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,64,power_law_1.2,0.053840641975402836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,64,power_law_1.2,2.136907501220703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,64,power_law_1.2,0.054369277954101555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,64,power_law_1.2,0.058508801460266116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,64,power_law_1.2,0.06412415981292725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,64,power_law_1.2,0.06444287776947022
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,64,power_law_1.2,0.0660364818572998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,64,power_law_1.2,0.0684761619567871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,64,power_law_1.2,0.0713753604888916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,64,power_law_1.2,0.07513728141784667
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,64,power_law_1.2,0.08178303718566896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,64,power_law_1.2,0.075829758644104
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,64,power_law_1.2,0.07860479831695558
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,64,power_law_1.2,0.08476544380187988
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,64,power_law_1.2,0.09805695533752441
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,64,power_law_1.2,0.12044032096862793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,64,power_law_1.2,0.14295552253723146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,64,power_law_1.2,0.1986969566345215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,64,power_law_1.2,0.22771711349487306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,64,power_law_1.2,0.3945011138916016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,64,power_law_1.2,0.49169662475585935
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,2,64,power_law_1.2,0.027002880573272704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,64,power_law_1.2,0.6900915527343751
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,2,64,power_law_1.2,0.026389760971069338
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,2,64,power_law_1.2,0.027256319522857664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,2,64,power_law_1.2,0.02715775966644287
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,2,64,power_law_1.2,0.024686079025268554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,64,power_law_1.2,0.9641190338134766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,2,64,power_law_1.2,0.02551039934158325
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,2,64,power_law_1.2,0.026333439350128173
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,2,64,power_law_1.2,0.026472959518432614
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,2,64,power_law_1.2,0.02709120035171509
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,2,64,power_law_1.2,0.026968319416046143
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,2,64,power_law_1.2,0.026077439785003664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,2,64,power_law_1.2,0.026753280162811276
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,2,64,power_law_1.2,0.02806272029876709
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,2,64,power_law_1.2,0.02838016033172607
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,2,64,power_law_1.2,0.027831039428710937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,2,64,power_law_1.2,0.03273216009140015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,2,64,power_law_1.2,0.025386240482330326
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,2,64,power_law_1.2,0.026566400527954104
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,2,64,power_law_1.01,0.07726975917816162
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,2,64,power_law_1.01,0.09334400177001953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,2,64,power_law_1.01,0.13604736328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,2,64,power_law_1.01,0.15498496055603028
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,2,64,power_law_1.2,0.04540031909942627
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,2,64,power_law_1.2,0.03572479963302612
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,2,64,power_law_1.2,0.05496064186096191
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,4,32,balanced,0.03209088087081909
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,32,balanced,0.03298943996429443
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,32,balanced,0.033011200428009035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,32,balanced,0.0354585599899292
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,32,balanced,0.06085631847381592
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,2,64,power_law_1.2,0.06366591930389405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,32,balanced,0.07572864055633545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,32,balanced,0.08619392395019532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,32,balanced,0.1392460823059082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,32,balanced,0.1369279956817627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,32,balanced,0.13651455879211424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,32,balanced,0.13959808349609376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,32,balanced,0.1438169574737549
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,32,balanced,0.14383487701416015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,32,balanced,0.15164928436279296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,32,balanced,0.15433216094970703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,32,balanced,0.1554854393005371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,32,balanced,0.15263999938964842
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,32,balanced,0.16009344100952147
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,32,balanced,0.16680063247680663
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,32,balanced,0.18110336303710936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,32,balanced,0.19852544784545897
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,32,balanced,0.23017343521118164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,32,balanced,0.2684416007995606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,32,balanced,0.3249728012084961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,32,balanced,0.3947443389892578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,32,balanced,0.034167039394378665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,32,balanced,0.527696647644043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,32,balanced,0.033807361125946046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,32,balanced,0.035423998832702634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,32,balanced,0.7360934448242188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,32,balanced,0.03585024118423462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,32,balanced,0.03854464054107666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,32,balanced,0.04771711826324463
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,32,balanced,0.06174079895019531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,32,balanced,0.07139840126037597
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,32,balanced,0.06896639823913574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,32,balanced,0.0692083215713501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,32,balanced,0.0716876792907715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,32,balanced,0.07341440200805664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,32,balanced,0.07360896110534668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,32,balanced,0.07950592041015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,32,balanced,0.08062848091125488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,32,balanced,0.08220159530639648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,32,balanced,0.07640063762664795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,32,balanced,0.08179200172424317
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,32,balanced,0.08374784469604493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,32,balanced,0.09410816192626953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,32,balanced,0.10355711936950684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,32,balanced,0.12250240325927735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,32,balanced,0.14257535934448243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,32,balanced,0.1789695930480957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,32,balanced,0.23206399917602538
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,32,balanced,0.3061631965637207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,32,balanced,0.3955046463012695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,2,64,power_law_1.2,0.07890431880950928
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,2,64,power_law_1.2,0.11883392333984374
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,2,64,power_law_1.2,0.12996095657348633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,2,64,power_law_1.2,0.20555648803710938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,2,64,power_law_1.2,0.23774848937988283
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,4,32,balanced,0.020737280845642088
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,4,32,balanced,0.032152318954467775
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,4,32,balanced,0.03222527980804443
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,4,32,balanced,0.02028287887573242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,4,32,balanced,0.03254400014877319
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,4,32,balanced,0.03386879920959472
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,4,32,balanced,0.020222721099853517
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,4,32,balanced,0.03437311887741089
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,4,32,balanced,0.03398528099060059
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,4,32,balanced,0.03411328077316284
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,4,32,balanced,0.03389568090438842
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,4,32,balanced,0.03385727882385254
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,4,32,balanced,0.034928638935089115
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,4,32,balanced,0.03578239917755127
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,4,32,balanced,0.03637759923934937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,4,32,balanced,0.039377920627593994
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,4,32,balanced,0.041011199951171876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,4,32,balanced,0.04387968063354492
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,4,32,balanced,0.03301503896713257
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,4,32,balanced,0.021148159503936767
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,4,32,power_law_1.01,0.07689472198486327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,32,power_law_1.01,0.08447999954223633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,32,power_law_1.01,0.08426495552062988
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,4,32,balanced,0.022301440238952634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,32,power_law_1.01,0.0852940845489502
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,32,power_law_1.01,0.08796671867370606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,32,power_law_1.01,0.13222399711608887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,32,power_law_1.01,0.13421952247619628
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,32,power_law_1.01,0.13390591621398926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,32,power_law_1.01,0.13785344123840332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,32,power_law_1.01,0.13521792411804198
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,32,power_law_1.01,0.14040831565856932
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,32,power_law_1.01,0.1449945640563965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,32,power_law_1.01,0.1480281639099121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,32,power_law_1.01,0.15734399795532225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,32,power_law_1.01,0.15605119705200193
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,32,power_law_1.01,0.15470975875854492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,32,power_law_1.01,0.16015743255615233
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,32,power_law_1.01,0.1731328010559082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,4,32,balanced,0.024940800666809083
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,32,power_law_1.01,0.18814336776733398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,32,power_law_1.01,0.21703296661376953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,32,power_law_1.01,0.24702463150024415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,32,power_law_1.01,0.31625343322753907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,32,power_law_1.01,0.36643966674804684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,32,power_law_1.01,0.5599488067626953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,32,power_law_1.01,0.7301888275146484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,32,power_law_1.01,0.04359424114227295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,32,power_law_1.01,1.004343032836914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,32,power_law_1.01,0.0587059211730957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,32,power_law_1.01,0.058853120803833
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,32,power_law_1.01,0.059608321189880374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,32,power_law_1.01,1.2392140960693359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,32,power_law_1.01,0.06147456169128418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,32,power_law_1.01,0.0640768003463745
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,32,power_law_1.01,0.061146879196166994
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,32,power_law_1.01,0.06382336139678954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,32,power_law_1.01,0.06540287971496582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,32,power_law_1.01,0.06524415969848632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,32,power_law_1.01,0.0688972806930542
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,32,power_law_1.01,0.07252352237701416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,32,power_law_1.01,0.07530752182006836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,32,power_law_1.01,0.08071935653686524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,32,power_law_1.01,0.08211839675903321
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,32,power_law_1.01,0.0789734411239624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,32,power_law_1.01,0.08247551918029786
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,32,power_law_1.01,0.0977625560760498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,32,power_law_1.01,0.11732480049133301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,32,power_law_1.01,0.14078080177307128
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,4,32,balanced,0.03125760078430176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,32,power_law_1.01,0.18122751235961915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,32,power_law_1.01,0.23726079940795897
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,32,power_law_1.01,0.3338393783569336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,32,power_law_1.01,0.08568448066711426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,32,power_law_1.01,0.3857830429077148
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,4,32,power_law_1.01,0.02591104030609131
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,4,32,power_law_1.01,0.03153408050537109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,32,power_law_1.01,0.5742233657836915
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,4,32,power_law_1.01,0.03247744083404541
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,4,32,power_law_1.01,0.0320793604850769
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,4,32,power_law_1.01,0.03277568101882934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,32,power_law_1.01,0.7536345672607422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,4,32,power_law_1.01,0.030074880123138425
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,4,32,power_law_1.01,0.028323841094970704
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,4,32,power_law_1.01,0.03048703908920288
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,4,32,power_law_1.01,0.028887040615081787
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,4,32,power_law_1.01,0.030000638961791993
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,4,32,power_law_1.01,0.0298304009437561
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,4,32,power_law_1.01,0.03000447988510132
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,4,32,power_law_1.01,0.03036288022994995
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,4,32,power_law_1.01,0.030563840866088866
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,4,32,power_law_1.01,0.03167232036590576
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,4,32,power_law_1.01,0.03312000036239624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,4,32,power_law_1.01,0.032951040267944334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,4,32,power_law_1.01,0.035270400047302246
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,4,32,balanced,0.03289599895477295
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,4,32,balanced,0.06149888038635254
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,4,32,balanced,0.06730879783630371
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,4,32,balanced,0.07994495868682862
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,4,32,power_law_1.01,0.03781120061874389
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,4,32,power_law_1.01,0.042673921585083006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,4,32,power_law_1.2,0.07679743766784668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,32,power_law_1.2,0.08409728050231934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,32,power_law_1.2,0.08460160255432128
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,32,power_law_1.2,0.08455039978027343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,32,power_law_1.2,0.085797119140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,4,32,power_law_1.01,0.06292223930358887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,32,power_law_1.2,0.1291801643371582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,32,power_law_1.2,0.1330022430419922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,32,power_law_1.2,0.13466496467590333
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,4,32,power_law_1.01,0.04941696166992188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,32,power_law_1.2,0.13737728118896483
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,32,power_law_1.2,0.13554688453674318
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,32,power_law_1.2,0.14555904388427734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,32,power_law_1.2,0.15128576278686523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,32,power_law_1.2,0.15418880462646484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,32,power_law_1.2,0.1522713565826416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,32,power_law_1.2,0.1572211170196533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,32,power_law_1.2,0.1617932891845703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,32,power_law_1.2,0.17730560302734374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,32,power_law_1.2,0.19025279998779296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,32,power_law_1.2,0.14308992385864258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,32,power_law_1.2,0.2277939224243164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,32,power_law_1.2,0.27121280670166015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,32,power_law_1.2,0.37067264556884766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,32,power_law_1.2,0.485978889465332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,32,power_law_1.2,0.6395289611816406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,32,power_law_1.2,0.8296601867675781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,32,power_law_1.2,0.04542848110198975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,32,power_law_1.2,0.05904384136199951
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,32,power_law_1.2,1.5268006896972657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,32,power_law_1.2,0.05833087921142578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,4,32,power_law_1.01,0.07702655792236328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,32,power_law_1.2,0.060555520057678225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,32,power_law_1.2,1.6185650634765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,32,power_law_1.2,0.061743359565734866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,32,power_law_1.2,0.0662604808807373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,32,power_law_1.2,0.06425856113433838
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,32,power_law_1.2,0.06573311805725097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,32,power_law_1.2,0.06332672119140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,32,power_law_1.2,0.0635481595993042
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,32,power_law_1.2,0.06802815914154052
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,32,power_law_1.2,0.07195263862609863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,32,power_law_1.2,0.07466239929199218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,32,power_law_1.2,0.07807616233825684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,32,power_law_1.2,0.07868671894073487
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,32,power_law_1.2,0.07783808231353759
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,32,power_law_1.2,0.08293120384216308
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,32,power_law_1.2,0.09285375595092774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,32,power_law_1.2,0.11102335929870606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,32,power_law_1.2,0.12391936302185058
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,32,power_law_1.2,0.1529152011871338
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,32,power_law_1.2,0.20421888351440431
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,32,power_law_1.2,0.26696319580078126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,32,power_law_1.2,0.41021568298339844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,32,power_law_1.2,0.48035327911376957
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,4,32,power_law_1.2,0.02598016023635864
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,4,32,power_law_1.2,0.03264640092849731
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,32,power_law_1.2,0.7585523223876953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,4,32,power_law_1.2,0.03212287902832031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,4,32,power_law_1.2,0.032865281105041506
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,4,32,power_law_1.2,0.03211008071899414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,32,power_law_1.2,1.0925580596923827
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,4,32,power_law_1.2,0.029525759220123293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,4,32,power_law_1.2,0.030716159343719483
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,4,32,power_law_1.2,0.028787200450897214
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,4,32,power_law_1.2,0.029615359306335447
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,4,32,power_law_1.2,0.029278719425201417
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,4,32,power_law_1.2,0.02942591905593872
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,4,32,power_law_1.2,0.02941440105438232
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,4,32,power_law_1.2,0.030412800312042236
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,4,32,power_law_1.2,0.03100928068161011
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,4,32,power_law_1.2,0.03199104070663452
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,4,32,power_law_1.2,0.033720319271087644
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,4,32,power_law_1.2,0.028838400840759275
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,4,32,power_law_1.01,0.0947276782989502
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,4,32,power_law_1.01,0.11881216049194336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,4,32,power_law_1.01,0.15167872428894041
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,4,32,power_law_1.01,0.20627328872680667
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,4,32,power_law_1.2,0.038293759822845455
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,4,32,power_law_1.2,0.04937856197357178
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,4,32,power_law_1.2,0.056390399932861324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,8,16,balanced,0.03750911951065063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,8,16,balanced,0.03708159923553467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,8,16,balanced,0.0338483190536499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,8,16,balanced,0.036538879871368404
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,4,32,power_law_1.2,0.041427202224731445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,8,16,balanced,0.06291071891784668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,8,16,balanced,0.07826303958892822
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,8,16,balanced,0.0923788833618164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,8,16,balanced,0.14572928428649903
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,4,32,power_law_1.2,0.08768511772155761
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,4,32,power_law_1.2,0.07309951782226562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,8,16,balanced,0.14734463691711425
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,4,32,power_law_1.2,0.13623552322387694
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,8,16,balanced,0.15238528251647948
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,8,16,balanced,0.15858688354492187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,8,16,balanced,0.15985535621643066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,8,16,balanced,0.1528115177154541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,8,16,balanced,0.15883392333984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,8,16,balanced,0.16061952590942383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,8,16,balanced,0.16576255798339842
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,8,16,balanced,0.18098688125610352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,8,16,balanced,0.19597312927246094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,8,16,balanced,0.22221696853637693
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,8,16,balanced,0.254335994720459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,8,16,balanced,0.3131046485900879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,8,16,balanced,0.14781311988830564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,8,16,balanced,0.3768819046020508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,8,16,balanced,0.4923097610473633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,8,16,balanced,0.03755136013031006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,8,16,balanced,0.034584319591522215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,8,16,balanced,0.03749504089355469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,8,16,balanced,0.04197760105133057
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,8,16,balanced,0.044843521118164066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,8,16,balanced,0.8561459350585938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,8,16,balanced,0.05705088138580322
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,8,16,balanced,0.07218175888061523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,8,16,balanced,0.08180992126464844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,8,16,balanced,0.08444800376892089
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,8,16,balanced,1.1741632080078124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,8,16,balanced,0.0891596794128418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,8,16,balanced,0.09217663764953613
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,8,16,balanced,0.0931545639038086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,8,16,balanced,0.0867903995513916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,8,16,balanced,0.09008640289306641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,8,16,balanced,0.09168767929077148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,8,16,balanced,0.09350399971008301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,8,16,balanced,0.6141836929321289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,8,16,balanced,0.102042875289917
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,8,16,balanced,0.11127936363220214
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,8,16,balanced,0.1296985626220703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,8,16,balanced,0.14897536277770995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,8,16,balanced,0.18519935607910157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,8,16,balanced,0.22165504455566407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,8,16,balanced,0.29254655838012694
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,8,16,balanced,0.08289664268493653
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,8,16,balanced,0.38655872344970704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,8,16,balanced,0.5224908828735352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,8,16,balanced,0.6843071746826171
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,4,32,power_law_1.2,0.17838464736938478
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,4,32,power_law_1.2,0.2327116775512695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,4,32,power_law_1.2,0.3610879898071289
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,8,16,balanced,0.02273024082183838
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,8,16,balanced,0.022076160907745362
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,8,16,balanced,0.021908481121063233
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,8,16,balanced,0.04175487995147705
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,8,16,balanced,0.04184703826904297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,8,16,balanced,0.04232319831848145
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,8,16,balanced,0.04293759822845459
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,8,16,balanced,0.04282368183135986
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,8,16,balanced,0.04272895812988281
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,8,16,balanced,0.04318592071533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,8,16,balanced,0.04290175914764404
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,8,16,balanced,0.025368320941925048
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,8,16,balanced,0.04337791919708252
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,8,16,balanced,0.04420735836029053
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,8,16,balanced,0.04540927886962891
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,8,16,balanced,0.04705023765563965
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,8,16,balanced,0.04861440181732178
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,8,16,balanced,0.05188223838806152
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,8,16,balanced,0.0555724811553955
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,8,16,balanced,0.06146687984466552
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,8,16,power_law_1.01,0.05970943927764892
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,8,16,balanced,0.02388223886489868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,8,16,power_law_1.01,0.07411456108093262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,8,16,power_law_1.01,0.08570879936218262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,8,16,power_law_1.01,0.08676095962524413
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,8,16,power_law_1.01,0.08930432319641113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,8,16,power_law_1.01,0.08538368225097656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,8,16,power_law_1.01,0.1377459239959717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,8,16,power_law_1.01,0.14071295738220216
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,8,16,power_law_1.01,0.1443443202972412
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,8,16,power_law_1.01,0.14663295745849608
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,8,16,power_law_1.01,0.15268351554870604
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,8,16,power_law_1.01,0.15517056465148926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,8,16,power_law_1.01,0.1553318405151367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,8,16,power_law_1.01,0.15242112159729004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,8,16,power_law_1.01,0.1577868843078613
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,8,16,power_law_1.01,0.16029184341430663
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,8,16,power_law_1.01,0.16695552825927734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,8,16,power_law_1.01,0.18308479309082032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,8,16,power_law_1.01,0.20273151397705078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,8,16,power_law_1.01,0.23872640609741208
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,8,16,power_law_1.01,0.2759820747375488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,8,16,power_law_1.01,0.36081024169921877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,8,16,power_law_1.01,0.45744895935058594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,8,16,power_law_1.01,0.5704089736938476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,8,16,power_law_1.01,0.7882380676269531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,8,16,power_law_1.01,1.2477005004882813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,8,16,balanced,0.03190016031265259
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,8,16,power_law_1.01,0.04125696182250976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,8,16,power_law_1.01,0.052712960243225096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,8,16,power_law_1.01,0.06628479957580566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,8,16,power_law_1.01,0.0662720012664795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,8,16,power_law_1.01,1.4930816650390626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,8,16,power_law_1.01,0.07023871898651123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,8,16,power_law_1.01,0.06661503791809083
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,8,16,power_law_1.01,0.07309440135955811
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,8,16,power_law_1.01,0.07635072231292725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,8,16,power_law_1.01,0.07689727783203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,8,16,power_law_1.01,0.0767091178894043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,8,16,power_law_1.01,0.08271103858947754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,8,16,power_law_1.01,0.0844825553894043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,8,16,power_law_1.01,0.09031552314758301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,8,16,power_law_1.01,0.0842086410522461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,8,16,power_law_1.01,0.08609791755676269
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,8,16,power_law_1.01,0.08790143966674804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,8,16,power_law_1.01,0.09242752075195312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,8,16,power_law_1.01,0.10549247741699219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,8,16,power_law_1.01,0.11251584053039551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,8,16,power_law_1.01,0.1455743980407715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,8,16,power_law_1.01,0.17740415573120116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,8,16,power_law_1.01,0.21010047912597657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,8,16,power_law_1.01,0.2727014350891113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,8,16,power_law_1.01,0.3793471908569336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,8,16,power_law_1.01,0.5030656051635742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,8,16,power_law_1.01,0.02558464050292969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,8,16,power_law_1.01,0.031769599914550785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,8,16,power_law_1.01,0.8914291381835937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,8,16,power_law_1.01,0.041795840263366704
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,8,16,power_law_1.01,0.040352001190185546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,8,16,power_law_1.01,0.04096896171569824
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,8,16,power_law_1.01,0.03938047885894776
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,8,16,power_law_1.01,0.03378560066223145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,8,16,power_law_1.01,0.76932861328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,8,16,power_law_1.01,0.03682048082351684
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,8,16,power_law_1.01,0.037281279563903806
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,8,16,power_law_1.01,0.03720704078674316
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,8,16,power_law_1.01,0.03741055965423584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,8,16,power_law_1.01,0.038602240085601806
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,8,16,power_law_1.01,0.038767359256744384
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,8,16,power_law_1.01,0.0390118408203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,8,16,power_law_1.01,0.04101888179779053
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,8,16,power_law_1.01,0.04137856006622315
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,8,16,power_law_1.01,0.043438081741333005
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,8,16,power_law_1.01,0.04588160037994385
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,8,16,balanced,0.04175615787506103
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,8,16,balanced,0.042247681617736815
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,8,16,balanced,0.08514304161071777
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,8,16,balanced,0.09698559761047362
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,8,16,balanced,0.12187264442443849
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,8,16,power_law_1.01,0.051272959709167475
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,8,16,power_law_1.01,0.062172160148620606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,8,16,power_law_1.2,0.059550719261169435
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,8,16,power_law_1.2,0.0746060800552368
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,8,16,power_law_1.2,0.08358271598815918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,8,16,power_law_1.2,0.08700287818908692
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,8,16,power_law_1.01,0.07098239898681641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,8,16,power_law_1.2,0.08922880172729493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,8,16,power_law_1.2,0.14024191856384277
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,8,16,power_law_1.2,0.14007167816162108
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,8,16,power_law_1.2,0.13939583778381348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,8,16,power_law_1.2,0.14223103523254393
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,8,16,power_law_1.2,0.14751999855041503
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,8,16,power_law_1.2,0.15261568069458006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,8,16,power_law_1.2,0.15742079734802245
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,8,16,power_law_1.2,0.15673343658447264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,8,16,power_law_1.2,0.1547980785369873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,8,16,power_law_1.2,0.1583615970611572
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,8,16,power_law_1.01,0.08146047592163086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,8,16,power_law_1.2,0.16338176727294923
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,8,16,power_law_1.01,0.10831104278564455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,8,16,power_law_1.2,0.16761600494384765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,8,16,power_law_1.2,0.19077119827270508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,8,16,power_law_1.2,0.2115238380432129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,8,16,power_law_1.2,0.25899648666381836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,8,16,power_law_1.2,0.3023859214782715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,8,16,power_law_1.2,0.38501888275146484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,8,16,power_law_1.2,0.5138304138183594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,8,16,power_law_1.2,0.7078540802001954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,8,16,power_law_1.2,0.040349440574646
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,8,16,power_law_1.2,0.928106231689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,8,16,power_law_1.2,0.04891007900238037
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,8,16,power_law_1.2,0.06355584144592286
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,8,16,power_law_1.2,1.4293069458007812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,8,16,power_law_1.2,0.06656640052795411
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,8,16,power_law_1.2,0.06918528079986572
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,8,16,power_law_1.2,0.07609087944030761
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,8,16,power_law_1.2,1.6900697326660157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,8,16,power_law_1.2,0.07541247844696045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,8,16,power_law_1.2,0.07410560131072999
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,8,16,power_law_1.2,0.08121983528137207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,8,16,power_law_1.2,0.08747776031494141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,8,16,power_law_1.2,0.08903679847717286
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,8,16,power_law_1.2,0.0826905632019043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,8,16,power_law_1.2,0.08683520317077636
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,8,16,power_law_1.2,0.07529600143432617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,8,16,power_law_1.2,0.09104767799377442
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,8,16,power_law_1.2,0.09125760078430176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,8,16,power_law_1.2,0.08036479949951172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,8,16,power_law_1.2,0.10798336029052735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,8,16,power_law_1.2,0.12369664192199707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,8,16,power_law_1.2,0.1497766399383545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,8,16,power_law_1.2,0.18220544815063477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,8,16,power_law_1.2,0.24111999511718749
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,8,16,power_law_1.2,0.3105536079406738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,8,16,power_law_1.2,0.4527628707885742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,8,16,power_law_1.2,0.6111219024658203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,8,16,power_law_1.2,0.025447680950164798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,8,16,power_law_1.2,0.8020480346679687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,8,16,power_law_1.2,0.03157119989395142
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,8,16,power_law_1.01,0.13164928436279297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,8,16,power_law_1.2,0.03990528106689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,8,16,power_law_1.2,1.178229751586914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,8,16,power_law_1.01,0.22565376281738284
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,8,16,power_law_1.2,0.04149375915527344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,8,16,power_law_1.2,0.04108928203582764
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,8,16,power_law_1.2,0.03572992086410522
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,8,16,power_law_1.2,0.03844480037689209
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,8,16,power_law_1.2,0.034900479316711426
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,8,16,power_law_1.2,0.037391359806060794
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,8,16,power_law_1.2,0.038433279991149906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,8,16,power_law_1.2,0.03782016038894653
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,8,16,power_law_1.2,0.03921535968780517
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,8,16,power_law_1.2,0.039756801128387455
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,8,16,power_law_1.2,0.039946238994598385
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,8,16,power_law_1.2,0.04241919994354248
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,8,16,power_law_1.2,0.04599552154541016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,8,16,power_law_1.2,0.053532161712646485
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,8,16,power_law_1.2,0.04124671936035156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,8,16,power_law_1.01,0.15089664459228516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,8,16,power_law_1.01,0.29905279159545894
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,8,16,power_law_1.2,0.05794559955596924
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,8,16,power_law_1.2,0.06944384098052978
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,8,16,power_law_1.2,0.12911231994628908
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,8,16,power_law_1.2,0.10439423561096191
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,16,8,balanced,0.03548543930053711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,16,8,balanced,0.034635519981384276
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,16,8,balanced,0.03743360042572021
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,8,16,power_law_1.2,0.24116607666015627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,16,8,balanced,0.04961535930633545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,16,8,balanced,0.06926591873168945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,16,8,balanced,0.08963968276977538
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,16,8,balanced,0.10128640174865722
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,16,8,balanced,0.1633203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,16,8,balanced,0.16737407684326172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,16,8,balanced,0.17073280334472657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,16,8,balanced,0.1654694366455078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,16,8,balanced,0.17175552368164063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,16,8,balanced,0.17429376602172852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,16,8,balanced,0.18200960159301757
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,16,8,balanced,0.18559999465942384
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,16,8,balanced,0.19435647964477537
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,16,8,balanced,0.20618368148803712
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,8,16,power_law_1.2,0.08097920417785645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,16,8,balanced,0.23475839614868166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,16,8,balanced,0.2637900733947754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,16,8,balanced,0.3202342224121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,16,8,balanced,0.3777612686157227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,16,8,balanced,0.49145729064941407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,16,8,balanced,0.6104627227783203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,16,8,balanced,0.03536256074905396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,16,8,balanced,0.8363136291503906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,16,8,balanced,0.038260478973388676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,16,8,balanced,0.04525824069976807
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,8,16,power_law_1.2,0.18779775619506836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,16,8,balanced,1.0670783996582032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,16,8,balanced,0.060244479179382325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,16,8,balanced,0.09402624130249024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,16,8,balanced,0.07770751953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,16,8,balanced,0.10568063735961915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,16,8,balanced,0.10839551925659179
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,16,8,balanced,0.11188480377197266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,16,8,balanced,1.52893310546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,16,8,balanced,0.10469504356384278
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,16,8,balanced,0.10894207954406739
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,16,8,balanced,0.11198719978332519
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,16,8,balanced,0.11716352462768556
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,16,8,balanced,0.11675135612487793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,16,8,balanced,0.12232704162597656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,16,8,balanced,0.13089023590087892
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,16,8,balanced,0.14684160232543947
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,16,8,balanced,0.16542720794677734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,16,8,balanced,0.052985601425170906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,16,8,balanced,2.0837542724609373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,16,8,balanced,0.20044799804687502
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,16,8,balanced,0.23530879974365235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,16,8,balanced,0.3068876838684082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,16,8,balanced,0.3729459381103516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,16,8,balanced,0.5096857452392578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,16,8,balanced,0.6794380950927734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,16,8,balanced,0.9561023712158203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,16,8,balanced,1.2578253173828124
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,8,16,power_law_1.2,0.3612416076660156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,8,16,power_law_1.2,0.49921279907226557
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,16,8,balanced,0.026854400634765623
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,16,8,balanced,0.05820672035217285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,16,8,balanced,0.02726016044616699
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,16,8,balanced,0.05837056159973144
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,16,8,balanced,0.028714239597320557
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,16,8,balanced,0.0584665584564209
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,16,8,balanced,0.058860797882080075
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,16,8,balanced,0.03045504093170166
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,16,8,balanced,0.058894081115722655
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,16,8,balanced,0.05893760204315186
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,16,8,balanced,0.05970943927764892
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,16,8,balanced,0.05993728160858154
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,16,8,balanced,0.061329917907714845
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,16,8,balanced,0.06291967868804932
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,16,8,balanced,0.06413311958312988
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,16,8,balanced,0.06680448055267334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,16,8,balanced,0.03504640102386475
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,16,8,balanced,0.07547391891479492
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,16,8,balanced,0.0821286392211914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,16,8,balanced,0.09298432350158692
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,16,8,power_law_1.01,0.03707648038864136
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,16,8,balanced,0.06968959808349609
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,16,8,balanced,0.04825471878051758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,16,8,power_law_1.01,0.06230400085449219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,16,8,power_law_1.01,0.07921792030334472
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,16,8,power_law_1.01,0.07962240219116211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,16,8,power_law_1.01,0.08714112281799316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,16,8,power_law_1.01,0.0876095962524414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,16,8,power_law_1.01,0.08925056457519531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,16,8,power_law_1.01,0.15820672035217284
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,16,8,power_law_1.01,0.16074623107910155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,16,8,power_law_1.01,0.16234111785888672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,16,8,power_law_1.01,0.16713855743408204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,16,8,power_law_1.01,0.17001983642578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,16,8,power_law_1.01,0.16595327377319335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,16,8,power_law_1.01,0.17206655502319337
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,16,8,power_law_1.01,0.1766758346557617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,16,8,power_law_1.01,0.18011903762817383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,16,8,power_law_1.01,0.1939507293701172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,16,8,power_law_1.01,0.21621120452880857
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,16,8,power_law_1.01,0.2405849647521973
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,16,8,power_law_1.01,0.27819648742675784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,16,8,power_law_1.01,0.35020416259765624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,16,8,power_law_1.01,0.44607105255126955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,16,8,power_law_1.01,0.5381043243408203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,16,8,power_law_1.01,0.7112102508544922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,16,8,power_law_1.01,0.926202850341797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,16,8,power_law_1.01,0.040737280845642085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,16,8,power_law_1.01,0.04811264038085937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,16,8,power_law_1.01,1.427914276123047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,16,8,power_law_1.01,0.06253568172454835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,16,8,power_law_1.01,0.06852096080780029
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,16,8,power_law_1.01,1.7978495788574218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,16,8,power_law_1.01,0.08361087799072266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,16,8,power_law_1.01,0.07737343788146973
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,16,8,power_law_1.01,0.09700991630554198
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,16,8,power_law_1.01,0.09798015594482422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,16,8,power_law_1.01,0.10168448448181153
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,16,8,power_law_1.01,0.07785088062286377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,16,8,power_law_1.01,0.09803008079528809
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,16,8,power_law_1.01,0.09615872383117677
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,16,8,power_law_1.01,0.10315008163452148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,16,8,power_law_1.01,0.10664959907531739
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,16,8,power_law_1.01,0.09530752182006837
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,16,8,power_law_1.01,0.10740863800048828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,16,8,balanced,0.05801727771759033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,16,8,power_law_1.01,0.1122316837310791
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,16,8,power_law_1.01,0.13074560165405275
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,16,8,power_law_1.01,0.14221952438354493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,16,8,power_law_1.01,0.1746188735961914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,16,8,power_law_1.01,0.21010303497314453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,16,8,power_law_1.01,0.2789964866638184
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,16,8,power_law_1.01,0.3442623901367187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,16,8,power_law_1.01,0.4496025466918946
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,16,8,power_law_1.01,0.030088961124420166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,16,8,power_law_1.01,0.624664306640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,16,8,power_law_1.01,0.034265599250793456
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,16,8,power_law_1.01,0.8417958068847657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,16,8,power_law_1.01,0.04851840019226074
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,16,8,power_law_1.01,0.04654335975646973
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,16,8,power_law_1.01,0.05139584064483642
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,16,8,power_law_1.01,1.122003173828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,16,8,power_law_1.01,0.0570956802368164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,16,8,power_law_1.01,0.057390079498291016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,16,8,power_law_1.01,0.05793536186218262
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,16,8,power_law_1.01,0.05196032047271728
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,16,8,power_law_1.01,0.05530752182006836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,16,8,power_law_1.01,0.05317247867584228
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,16,8,power_law_1.01,0.055580158233642575
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,16,8,power_law_1.01,0.054864640235900874
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,16,8,power_law_1.01,0.05759488105773926
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,16,8,power_law_1.01,0.0592460823059082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,16,8,power_law_1.01,0.059911680221557614
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,16,8,power_law_1.01,0.06263423919677734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,16,8,power_law_1.01,0.06504191875457763
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,16,8,balanced,0.058325757980346685
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,16,8,balanced,0.20894847869873048
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,16,8,balanced,0.1354483222961426
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,16,8,balanced,0.1606617546081543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,16,8,power_law_1.01,0.0785318422317505
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,16,8,power_law_1.01,0.08899840354919433
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,16,8,power_law_1.01,0.10396032333374024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,16,8,power_law_1.2,0.03780224084854126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,16,8,power_law_1.2,0.061249279975891115
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,16,8,power_law_1.2,0.07446271896362304
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,16,8,power_law_1.01,0.13555583953857422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,16,8,power_law_1.2,0.07849088191986084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,16,8,power_law_1.2,0.08298368453979492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,16,8,power_law_1.2,0.0972646427154541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,16,8,power_law_1.2,0.0891097640991211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,16,8,power_law_1.2,0.1571072006225586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,16,8,power_law_1.2,0.15986047744750978
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,16,8,power_law_1.2,0.16186111450195312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,16,8,power_law_1.2,0.16732416152954102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,16,8,power_law_1.2,0.1669580841064453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,16,8,power_law_1.2,0.1715007972717285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,16,8,power_law_1.2,0.17222272872924804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,16,8,power_law_1.2,0.17763711929321288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,16,8,power_law_1.2,0.18245248794555663
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,16,8,power_law_1.2,0.19498111724853515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,16,8,power_law_1.2,0.22113536834716796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,16,8,power_law_1.2,0.24498687744140626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,16,8,power_law_1.2,0.30038272857666015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,16,8,power_law_1.2,0.34937984466552735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,16,8,power_law_1.2,0.4407295989990234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,16,8,power_law_1.2,0.5825689697265626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,16,8,power_law_1.2,0.7886732482910157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,16,8,power_law_1.2,1.022097930908203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,16,8,power_law_1.2,0.0412556791305542
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,16,8,power_law_1.2,0.04770431995391845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,16,8,power_law_1.2,1.543891143798828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,16,8,power_law_1.2,0.058269438743591306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,16,8,power_law_1.2,0.06718463897705078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,16,8,power_law_1.2,0.07687295913696289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,16,8,power_law_1.2,1.8084249877929686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,16,8,power_law_1.2,0.08380543708801269
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,16,8,power_law_1.2,0.08890624046325683
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,16,8,power_law_1.2,0.0942527961730957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,16,8,power_law_1.2,0.09534591674804688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,16,8,power_law_1.2,0.09498368263244629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,16,8,power_law_1.2,0.09972864151000976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,16,8,power_law_1.2,0.09788288116455078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,16,8,power_law_1.2,0.09751423835754394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,16,8,power_law_1.2,0.10236543655395507
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,16,8,power_law_1.2,0.10614527702331542
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,16,8,power_law_1.2,0.10993535995483399
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,16,8,power_law_1.2,0.11297151565551758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,16,8,power_law_1.2,0.1369600009918213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,16,8,power_law_1.2,0.14926591873168946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,16,8,power_law_1.2,0.187457275390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,16,8,power_law_1.2,0.21893760681152347
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,16,8,power_law_1.2,0.2882124710083008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,16,8,power_law_1.2,0.36858497619628905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,16,8,power_law_1.2,0.5073612976074219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,16,8,power_law_1.2,0.6724569702148437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,16,8,power_law_1.2,0.030109438896179196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,16,8,power_law_1.2,0.9974809265136718
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,16,8,power_law_1.2,0.033459200859069824
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,16,8,power_law_1.2,0.042109441757202146
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,16,8,power_law_1.2,0.045401601791381835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,16,8,power_law_1.2,1.2563455963134766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,16,8,power_law_1.2,0.051210241317749025
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,16,8,power_law_1.2,0.05547647953033448
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,16,8,power_law_1.2,0.057541117668151856
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,16,8,power_law_1.2,0.05555712223052979
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,16,8,power_law_1.2,0.057287678718566895
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,16,8,power_law_1.2,0.055302400588989255
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,16,8,power_law_1.2,0.05278975963592529
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,16,8,power_law_1.2,0.05425663948059082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,16,8,power_law_1.2,0.05543551921844483
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,16,8,power_law_1.2,0.05867648124694824
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,16,8,power_law_1.2,0.059877119064331054
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,16,8,power_law_1.2,0.062048001289367674
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,16,8,power_law_1.2,0.06612864017486572
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,16,8,power_law_1.01,0.15490176200866698
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,16,8,power_law_1.2,0.07704832077026366
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,16,8,power_law_1.01,0.2323776054382324
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,16,8,power_law_1.01,0.31067136764526365
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,16,8,power_law_1.2,0.2313420867919922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,16,8,power_law_1.01,0.4302732849121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,16,8,power_law_1.01,0.5532121658325195
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,16,8,power_law_1.2,0.08530943870544434
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,16,8,power_law_1.2,0.12584320068359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,32,4,balanced,0.032499198913574216
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,32,4,balanced,0.03740799903869629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,32,4,balanced,0.034617600440979
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,32,4,balanced,0.05246463775634765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,32,4,balanced,0.0764249610900879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,32,4,balanced,0.09703167915344238
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,32,4,balanced,0.11116031646728515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,32,4,balanced,0.11289728164672852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,32,4,balanced,0.1139673614501953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,32,4,balanced,0.11526656150817871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,32,4,balanced,0.19736831665039062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,16,8,power_law_1.2,0.1055014419555664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,32,4,balanced,0.20039167404174804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,32,4,balanced,0.20622720718383789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,32,4,balanced,0.21130752563476562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,32,4,balanced,0.21558399200439454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,32,4,balanced,0.22349824905395507
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,32,4,balanced,0.23711360931396483
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,16,8,power_law_1.2,0.36968193054199217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,32,4,balanced,0.26422271728515623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,32,4,balanced,0.2923737525939941
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,32,4,balanced,0.3470758438110352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,32,4,balanced,0.40308990478515627
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,16,8,power_law_1.2,0.4614361572265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,16,8,power_law_1.2,0.17954816818237304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,32,4,balanced,0.5193702316284179
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,32,4,balanced,0.6373555374145508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,32,4,balanced,0.03420799970626831
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,32,4,balanced,0.8563008117675782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,32,4,balanced,0.03780992031097412
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,32,4,balanced,0.04524543762207031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,32,4,balanced,0.06115839958190918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,16,8,power_law_1.2,0.6611148834228515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,32,4,balanced,0.07679999828338623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,32,4,balanced,0.11238911628723144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,32,4,balanced,0.12423168182373048
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,32,4,balanced,0.1259712028503418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,32,4,balanced,1.5462924194335936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,32,4,balanced,0.12796031951904296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,32,4,balanced,0.12844415664672854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,32,4,balanced,0.14977408409118653
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,32,4,balanced,0.1534988784790039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,32,4,balanced,0.1563468837738037
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,32,4,balanced,0.16178688049316406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,32,4,balanced,0.1610700798034668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,32,4,balanced,0.16672256469726562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,32,4,balanced,0.17467775344848632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,32,4,balanced,0.19147136688232422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,32,4,balanced,0.20960512161254882
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,32,4,balanced,0.2445337677001953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,32,4,balanced,2.12606201171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,32,4,balanced,1.0907225799560547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,32,4,balanced,0.2763007926940918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,32,4,balanced,0.34608001708984376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,32,4,balanced,0.4107302474975586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,32,4,balanced,0.5447846221923828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,32,4,balanced,0.7697714996337891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,32,4,balanced,1.0350643157958985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,32,4,balanced,1.3758399963378907
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,16,8,power_law_1.2,0.9726041412353515
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,32,4,balanced,0.03824512004852295
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,32,4,balanced,0.0374015998840332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,32,4,balanced,0.03961855888366699
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,32,4,balanced,0.08903039932250977
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,32,4,balanced,0.04383615970611572
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,32,4,balanced,0.08938240051269532
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,32,4,balanced,0.08908032417297364
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,32,4,balanced,0.08999168395996093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,32,4,balanced,0.05472383975982666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,32,4,balanced,0.09126655578613281
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,32,4,balanced,0.09249407768249511
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,32,4,balanced,0.09197567939758301
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,32,4,balanced,0.09471360206604004
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,32,4,balanced,0.09675392150878906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,32,4,balanced,0.10059264183044433
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,32,4,balanced,0.10573439598083496
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,32,4,balanced,0.1111564826965332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,32,4,balanced,0.12100735664367676
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,32,4,balanced,0.13341055870056154
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,32,4,balanced,0.08781951904296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,32,4,balanced,0.1557094383239746
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,32,4,balanced,0.08913663864135743
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,32,4,power_law_1.01,0.035502080917358396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,32,4,power_law_1.01,0.03879551887512207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,32,4,power_law_1.01,0.07303423881530761
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,32,4,power_law_1.01,0.08452863693237304
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,32,4,balanced,0.07268991947174072
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,32,4,power_law_1.01,0.09280384063720704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,32,4,power_law_1.01,0.0931276798248291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,32,4,power_law_1.01,0.10590463638305665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,32,4,power_law_1.01,0.10161408424377441
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,32,4,power_law_1.01,0.1037273597717285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,32,4,power_law_1.01,0.1912499237060547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,32,4,power_law_1.01,0.19539327621459962
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,32,4,power_law_1.01,0.19951744079589845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,32,4,power_law_1.01,0.06393727779388428
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,32,4,power_law_1.01,0.2080281639099121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,32,4,power_law_1.01,0.21006847381591798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,32,4,power_law_1.01,0.22027008056640626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,32,4,power_law_1.01,0.2336844825744629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,32,4,power_law_1.01,0.2601958465576172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,32,4,power_law_1.01,0.2860902404785156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,32,4,power_law_1.01,0.34731006622314453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,32,4,power_law_1.01,0.3939161682128906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,32,4,power_law_1.01,0.5115520095825195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,32,4,power_law_1.01,0.6051161575317383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,32,4,power_law_1.01,0.8430016326904297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,32,4,power_law_1.01,1.0571174621582031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,32,4,power_law_1.01,0.03956223964691162
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,32,4,power_law_1.01,0.04619008064270019
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,32,4,power_law_1.01,0.05746816158294678
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,32,4,power_law_1.01,0.07172607898712158
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,32,4,power_law_1.01,0.09150336265563966
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,32,4,power_law_1.01,0.10012543678283692
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,32,4,power_law_1.01,0.10292608261108398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,32,4,power_law_1.01,2.0286015319824218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,32,4,power_law_1.01,0.11238783836364745
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,32,4,power_law_1.01,0.1128115177154541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,32,4,power_law_1.01,0.11143936157226561
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,32,4,power_law_1.01,1.5442303466796876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,32,4,power_law_1.01,0.12777983665466308
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,32,4,power_law_1.01,0.1321548843383789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,32,4,power_law_1.01,0.13293312072753907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,32,4,power_law_1.01,0.14175359725952147
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,32,4,power_law_1.01,0.14366463661193846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,32,4,power_law_1.01,0.1461068820953369
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,32,4,power_law_1.01,0.15976575851440428
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,32,4,power_law_1.01,0.17827072143554687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,32,4,power_law_1.01,0.19502208709716798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,32,4,power_law_1.01,0.23505023956298826
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,32,4,power_law_1.01,0.27325952529907227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,32,4,power_law_1.01,0.34947326660156247
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,32,4,power_law_1.01,0.4242086410522461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,32,4,power_law_1.01,0.5914329528808594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,32,4,power_law_1.01,0.7427891540527344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,32,4,balanced,0.08843392372131348
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,32,4,power_law_1.01,0.040755200386047366
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,32,4,power_law_1.01,0.044440321922302246
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,32,4,power_law_1.01,0.05613056182861328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,32,4,power_law_1.01,0.05444479942321777
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,32,4,power_law_1.01,1.0711615753173827
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,32,4,power_law_1.01,0.062255358695983885
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,32,4,power_law_1.01,0.07071104049682617
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,32,4,power_law_1.01,0.07149824142456054
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,32,4,power_law_1.01,0.07890175819396973
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,32,4,power_law_1.01,0.08136192321777344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,32,4,power_law_1.01,0.08121472358703613
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,32,4,power_law_1.01,0.0838425636291504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,32,4,power_law_1.01,1.3669223022460937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,32,4,power_law_1.01,0.08164223670959472
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,32,4,power_law_1.01,0.07997439861297607
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,32,4,power_law_1.01,0.08307456016540528
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,32,4,power_law_1.01,0.08602239608764647
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,32,4,power_law_1.01,0.0886899185180664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,32,4,power_law_1.01,0.09634559631347657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,32,4,power_law_1.01,0.1047372817993164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,32,4,balanced,0.2332377624511719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,32,4,balanced,0.3752268981933594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,32,4,balanced,0.2815795135498047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,32,4,power_law_1.01,0.14507648468017578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,32,4,power_law_1.01,0.12605952262878417
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,32,4,power_law_1.01,0.18461311340332032
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,32,4,power_law_1.01,0.22637952804565428
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,32,4,power_law_1.2,0.03462784051895142
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,32,4,power_law_1.2,0.03913088083267212
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,32,4,power_law_1.2,0.06248064041137695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,32,4,power_law_1.2,0.07202176094055177
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,32,4,power_law_1.2,0.08291839599609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,32,4,power_law_1.2,0.09445247650146485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,32,4,power_law_1.2,0.0911616039276123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,32,4,power_law_1.2,0.10197376251220702
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,32,4,power_law_1.2,0.10337535858154297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,32,4,power_law_1.2,0.10157312393188475
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,32,4,power_law_1.01,0.29905792236328127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,32,4,power_law_1.2,0.1900223922729492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,32,4,power_law_1.2,0.19565439224243164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,32,4,power_law_1.2,0.19918975830078126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,32,4,power_law_1.2,0.20729343414306642
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,32,4,power_law_1.2,0.2107814407348633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,32,4,power_law_1.2,0.21982463836669922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,32,4,power_law_1.2,0.2322995185852051
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,32,4,power_law_1.2,0.2623526382446289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,32,4,power_law_1.2,0.29051391601562504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,32,4,power_law_1.2,0.3529024124145508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,32,4,power_law_1.2,0.39244544982910157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,32,4,power_law_1.2,0.5114278411865234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,32,4,power_law_1.2,0.6258867263793946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,32,4,power_law_1.2,0.854540786743164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,32,4,power_law_1.2,0.03910399913787842
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,32,4,power_law_1.2,1.115947494506836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,32,4,power_law_1.2,0.046069760322570805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,32,4,power_law_1.2,0.053820161819458014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,32,4,power_law_1.2,1.6226765441894533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,32,4,power_law_1.2,0.06969727993011474
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,32,4,power_law_1.2,0.08987135887145996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,32,4,power_law_1.2,0.10199935913085938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,32,4,power_law_1.2,0.09874943733215333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,32,4,power_law_1.2,0.10788352012634277
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,32,4,power_law_1.2,2.1102272033691407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,32,4,power_law_1.2,0.11144448280334474
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,32,4,power_law_1.2,0.11532032012939453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,32,4,power_law_1.2,0.1342745590209961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,32,4,power_law_1.2,0.12921600341796874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,32,4,power_law_1.2,0.13375743865966797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,32,4,power_law_1.2,0.14161791801452636
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,32,4,power_law_1.2,0.14263039588928222
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,32,4,power_law_1.2,0.1790822410583496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,32,4,power_law_1.2,0.16024959564208985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,32,4,power_law_1.2,0.20094720840454103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,32,4,power_law_1.2,0.2446361541748047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,32,4,power_law_1.2,0.2835136032104492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,32,4,power_law_1.2,0.3644313430786133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,32,4,power_law_1.2,0.45913726806640626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,32,4,power_law_1.2,0.1463270378112793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,32,4,power_law_1.2,0.6271398544311524
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,32,4,power_law_1.2,0.0410752010345459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,32,4,power_law_1.2,0.7854656219482422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,32,4,power_law_1.2,0.04400767803192139
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,32,4,power_law_1.2,0.05057407855987549
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,32,4,power_law_1.2,1.1199014282226563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,32,4,power_law_1.2,0.056314878463745124
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,32,4,power_law_1.2,0.06219136238098144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,32,4,power_law_1.2,1.465301818847656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,32,4,power_law_1.2,0.06761216163635254
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,32,4,power_law_1.2,0.06994431972503662
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,32,4,power_law_1.2,0.07580031871795653
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,32,4,power_law_1.2,0.08487039566040039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,32,4,power_law_1.2,0.0825113582611084
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,32,4,power_law_1.2,0.08530559539794921
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,32,4,power_law_1.2,0.08628607749938964
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,32,4,power_law_1.2,0.08277503967285156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,32,4,power_law_1.2,0.08346495628356934
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,32,4,power_law_1.2,0.08912128448486328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,32,4,power_law_1.2,0.09295743942260742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,32,4,power_law_1.2,0.09841792106628418
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,32,4,power_law_1.2,0.12431615829467772
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,32,4,power_law_1.01,0.404956169128418
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,32,4,power_law_1.01,0.5009369659423828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,32,4,power_law_1.01,0.7653990173339844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,32,4,power_law_1.01,0.9341862487792969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,32,4,power_law_1.2,0.14635007858276367
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,32,4,power_law_1.2,0.22107263565063479
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,128,balanced,0.027228159904479982
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,32,4,power_law_1.2,0.32877822875976564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,128,balanced,0.025989120006561277
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,128,balanced,0.026397440433502194
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,128,balanced,0.026828799247741696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,128,balanced,0.02629120111465454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,128,balanced,0.044513278007507324
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,32,4,power_law_1.2,0.1754867172241211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,128,balanced,0.04515711784362793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,128,balanced,0.02440063953399658
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,128,balanced,0.04562943935394287
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,128,balanced,0.04547455787658691
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,128,balanced,0.047397122383117676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,128,balanced,0.047261438369750976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,128,balanced,0.047320318222045896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,128,balanced,0.048139519691467285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,128,balanced,0.04933504104614258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,128,balanced,0.05049088001251221
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,128,balanced,0.053183999061584476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,128,balanced,0.05787007808685303
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,128,balanced,0.06412415981292725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,128,balanced,0.06284416198730469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,128,balanced,0.04565375804901123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,128,balanced,0.07924352169036866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,128,balanced,0.08876799583435059
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,128,balanced,0.10224639892578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,128,balanced,0.11604864120483398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,128,balanced,0.14136575698852538
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,128,balanced,0.02845567941665649
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,128,balanced,0.1715852737426758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,128,balanced,0.028925440311431884
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,128,balanced,0.028843519687652586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,128,balanced,0.028016641139984134
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,128,balanced,0.02883968114852905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,128,balanced,0.02891520023345947
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,128,balanced,0.03499520063400269
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,128,balanced,0.03524607896804809
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,128,balanced,0.03541376113891602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,128,balanced,0.03581056118011475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,128,balanced,0.03541887998580932
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,128,balanced,0.03547264099121093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,128,balanced,0.0354150390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,128,balanced,0.036215040683746334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,128,balanced,0.03660671949386597
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,128,balanced,0.036652801036834715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,128,balanced,0.037849600315094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,128,balanced,0.038254079818725584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,128,balanced,0.03956608057022094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,128,balanced,0.04204671859741211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,128,balanced,0.03880831956863403
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,128,balanced,0.04520832061767578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,128,balanced,0.04845823764801026
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,128,balanced,0.05560192108154297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,128,balanced,0.06081664085388183
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,128,balanced,0.07916160106658936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,128,balanced,0.09820672035217284
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,32,4,power_law_1.2,0.3685619354248047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,32,4,power_law_1.2,0.6094540786743164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,32,4,power_law_1.2,0.7542848205566406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,32,4,power_law_1.2,1.3196914672851563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,1,128,balanced,0.01618175983428955
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,1,128,balanced,0.016565760374069212
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,1,128,balanced,0.01650048017501831
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,1,128,balanced,0.016567039489746093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,1,128,balanced,0.016390399932861326
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,1,128,balanced,0.015886080265045167
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,1,128,balanced,0.016549119949340822
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,1,128,balanced,0.016527359485626218
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,1,128,balanced,0.016599040031433105
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,1,128,balanced,0.016195839643478392
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,1,128,balanced,0.016254719495773316
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,1,128,balanced,0.01658112049102783
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,1,128,balanced,0.016606719493865968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,1,128,balanced,0.016661759614944458
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,1,128,balanced,0.016602879762649535
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,1,128,balanced,0.01624575972557068
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,1,128,balanced,0.016933120489120483
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,1,128,balanced,0.0170035195350647
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,1,128,balanced,0.01698304057121277
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,1,128,balanced,0.01665791988372803
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,1,128,balanced,0.016163840293884277
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,32,4,power_law_1.2,1.6143128967285159
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,1,128,balanced,0.01662976026535034
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,128,power_law_1.01,0.027243518829345705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,128,power_law_1.01,0.027679998874664307
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,128,power_law_1.01,0.027756800651550294
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,128,power_law_1.01,0.045207037925720214
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,128,power_law_1.01,0.027964160442352292
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,128,power_law_1.01,0.046419200897216795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,128,power_law_1.01,0.047237119674682616
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,128,power_law_1.01,0.05289216041564941
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,128,power_law_1.01,0.05515903949737548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,128,power_law_1.01,0.04662399768829346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,128,power_law_1.01,0.0569817590713501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,128,power_law_1.01,0.05867008209228516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,128,power_law_1.01,0.06369791984558106
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,128,power_law_1.01,0.04872704029083252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,128,power_law_1.01,0.07141248226165772
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,128,power_law_1.01,0.07784319877624511
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,128,power_law_1.01,0.07720320224761963
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,128,power_law_1.01,0.09222784042358398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,128,power_law_1.01,0.09815679550170898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,128,power_law_1.01,0.115033597946167
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,128,power_law_1.01,0.13706368446350098
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,128,power_law_1.01,0.17828479766845703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,128,power_law_1.01,0.2498419189453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,128,power_law_1.01,0.3293030548095703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,128,power_law_1.01,0.4924531173706055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,128,power_law_1.01,0.6448409271240234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,128,power_law_1.01,0.02881407976150513
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,128,power_law_1.01,0.9832998657226563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,128,power_law_1.01,0.02839423894882202
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,128,power_law_1.01,0.0285478401184082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,128,power_law_1.01,0.027648000717163085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,128,power_law_1.01,0.03584768056869507
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,128,power_law_1.01,0.035403521060943605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,128,power_law_1.01,0.036081919670104975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,128,power_law_1.01,0.03619199991226196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,128,power_law_1.01,0.03739264011383057
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,128,power_law_1.01,0.037039361000061034
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,128,power_law_1.01,0.03788288116455078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,128,power_law_1.01,0.03492991924285889
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,128,power_law_1.01,0.037843201160430905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,128,power_law_1.01,0.03957376003265381
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,128,power_law_1.01,0.04111231803894043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,128,power_law_1.01,0.042813439369201664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,128,power_law_1.01,0.04240128040313721
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,128,power_law_1.01,0.048313598632812496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,128,power_law_1.01,0.05246463775634765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,128,power_law_1.01,0.06269567966461181
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,128,power_law_1.01,0.06923136234283447
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,128,power_law_1.01,0.09613568305969238
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,128,power_law_1.01,0.112806396484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,128,power_law_1.01,0.16032384872436523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,128,power_law_1.01,0.24394880294799806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,128,power_law_1.01,0.3203865432739258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,1,128,power_law_1.01,0.01615231990814209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,128,power_law_1.01,0.48191360473632816
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,1,128,power_law_1.01,0.01617535948753357
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,1,128,power_law_1.01,0.016172800064086914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,1,128,power_law_1.01,0.01618175983428955
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,1,128,power_law_1.01,0.01658112049102783
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,1,128,power_law_1.01,0.016518399715423585
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,1,128,power_law_1.01,0.01650048017501831
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,1,128,power_law_1.01,0.016606719493865968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,1,128,power_law_1.01,0.0165120005607605
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,1,128,power_law_1.01,0.016491520404815673
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,1,128,power_law_1.01,0.016550400257110597
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,1,128,power_law_1.01,0.016919039487838745
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,1,128,power_law_1.01,0.01659775972366333
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,1,128,power_law_1.01,0.016538879871368407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,1,128,balanced,0.01637120008468628
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,1,128,power_law_1.01,0.016933120489120483
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,1,128,power_law_1.01,0.016917760372161868
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,1,128,balanced,0.019027199745178223
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,1,128,balanced,0.021300480365753174
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,1,128,balanced,0.02619263887405395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,1,128,balanced,0.018666239976882933
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,1,128,power_law_1.01,0.016564480066299438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,1,128,power_law_1.01,0.017765120267868043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,1,128,power_law_1.01,0.018250240087509157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,1,128,power_law_1.01,0.020207359790802
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,128,power_law_1.2,0.02841599941253662
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,128,power_law_1.2,0.02681727886199951
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,128,power_law_1.2,0.02648832082748413
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,128,power_law_1.2,0.043589119911193845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,128,power_law_1.2,0.04451456069946289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,128,power_law_1.2,0.04828544139862061
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,128,power_law_1.2,0.049870080947875976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,128,power_law_1.2,0.04937727928161621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,128,power_law_1.2,0.05392384052276611
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,128,power_law_1.2,0.05819263935089112
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,128,power_law_1.2,0.06583936214447021
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,128,power_law_1.2,0.05769728183746338
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,128,power_law_1.2,0.06507391929626465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,128,power_law_1.2,0.06918399810791016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,128,power_law_1.2,0.07735936164855957
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,1,128,power_law_1.01,0.021534719467163087
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,128,power_law_1.2,0.08194815635681152
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,1,128,power_law_1.01,0.02726912021636963
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,128,power_law_1.2,0.08181632041931153
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,1,128,power_law_1.01,0.029648640155792237
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,128,power_law_1.2,0.09562751770019531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,128,power_law_1.2,0.11094528198242189
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,128,power_law_1.2,0.13868160247802735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,128,power_law_1.2,0.16509439468383788
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,128,power_law_1.2,0.26152448654174804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,128,power_law_1.2,0.334543342590332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,128,power_law_1.2,0.4934630584716797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,128,power_law_1.2,0.029666559696197508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,128,power_law_1.2,0.7259584045410156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,128,power_law_1.2,0.029016320705413816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,128,power_law_1.2,0.02882688045501709
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,128,power_law_1.2,1.0034086608886719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,128,power_law_1.2,0.03500672101974488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,128,power_law_1.2,0.03545216083526611
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,128,power_law_1.2,0.03619456052780151
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,128,power_law_1.2,0.035825920104980466
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,128,power_law_1.2,0.03787904024124146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,128,power_law_1.2,0.03619839906692505
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,128,power_law_1.2,1.575536651611328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,128,power_law_1.2,0.0370419192314148
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,1,128,power_law_1.01,0.03982079982757568
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,128,power_law_1.2,0.0375167989730835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,128,power_law_1.2,0.03953023910522461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,128,power_law_1.2,0.04064256191253662
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,128,power_law_1.2,0.04253952026367187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,128,power_law_1.2,0.04112127780914306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,128,power_law_1.2,0.044421119689941405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,128,power_law_1.2,0.04882815837860107
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,128,power_law_1.2,0.0575820779800415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,128,power_law_1.2,0.07179264068603516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,128,power_law_1.2,0.08228480339050293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,128,power_law_1.2,0.13088255882263183
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,128,power_law_1.2,0.17703039169311524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,128,power_law_1.2,0.28665344238281254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,128,power_law_1.2,0.3055500793457031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,128,power_law_1.2,0.039468801021575926
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,1,128,power_law_1.2,0.01617792010307312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,1,128,power_law_1.2,0.01656191945075989
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,1,128,power_law_1.2,0.01618559956550598
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,128,power_law_1.2,0.6229158401489258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,1,128,power_law_1.2,0.01649664044380188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,1,128,power_law_1.2,0.01648640036582947
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,1,128,power_law_1.2,0.016582399606704712
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,1,128,power_law_1.2,0.016590080261230468
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,1,128,power_law_1.2,0.016344319581985473
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,128,power_law_1.2,0.812099838256836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,1,128,power_law_1.2,0.01664639949798584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,1,128,power_law_1.2,0.01651584029197693
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,1,128,power_law_1.2,0.016989439725875854
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,1,128,power_law_1.2,0.017004799842834473
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,1,128,power_law_1.2,0.016910079717636108
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,1,128,power_law_1.2,0.017007360458374025
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,1,128,power_law_1.2,0.016537599563598633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,1,128,power_law_1.2,0.016989439725875854
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,1,128,power_law_1.01,0.052389121055603026
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,1,128,power_law_1.01,0.08279680252075196
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,1,128,power_law_1.01,0.06793727874755859
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,1,128,power_law_1.2,0.06435967922210693
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,1,128,power_law_1.2,0.018612480163574217
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,1,128,power_law_1.2,0.026677761077880856
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,1,128,power_law_1.2,0.01755136013031006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,2,64,balanced,0.02414463996887207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,2,64,balanced,0.02267904043197632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,2,64,balanced,0.023850240707397462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,2,64,balanced,0.02355072021484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,2,64,balanced,0.02317312002182007
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,1,128,power_law_1.2,0.021916160583496096
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,1,128,power_law_1.2,0.019829759597778322
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,2,64,balanced,0.02518784046173096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,2,64,balanced,0.041588478088378907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,2,64,balanced,0.039907839298248295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,2,64,balanced,0.04243072032928467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,2,64,balanced,0.044870400428771974
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,2,64,balanced,0.04360576152801514
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,2,64,balanced,0.04441599845886231
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,2,64,balanced,0.04847616195678711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,2,64,balanced,0.049292798042297366
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,2,64,balanced,0.04956416130065918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,2,64,balanced,0.05059584140777588
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,2,64,balanced,0.056721920967102046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,2,64,balanced,0.05778175830841065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,2,64,balanced,0.06480000019073487
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,2,64,balanced,0.07616127967834473
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,2,64,balanced,0.08504192352294922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,2,64,balanced,0.09790847778320313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,2,64,balanced,0.11716480255126953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,2,64,balanced,0.13837696075439454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,2,64,balanced,0.17993471145629883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,2,64,balanced,0.04993408203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,2,64,balanced,0.2237631988525391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,2,64,balanced,0.026337280273437503
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,2,64,balanced,0.025768959522247316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,2,64,balanced,0.025611519813537598
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,2,64,balanced,0.026067199707031252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,2,64,balanced,0.02682624101638794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,2,64,balanced,0.027961599826812743
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,2,64,balanced,0.03284352064132691
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,2,64,balanced,0.0337612795829773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,2,64,balanced,0.033369600772857666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,2,64,balanced,0.03463295936584473
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,2,64,balanced,0.033913600444793704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,2,64,balanced,0.03668224096298218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,2,64,balanced,0.03424384117126465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,2,64,balanced,0.040272641181945804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,2,64,balanced,0.035064320564270016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,1,128,power_law_1.2,0.032883200645446774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,2,64,balanced,0.03738368034362793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,2,64,balanced,0.03996927976608276
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,2,64,balanced,0.037047040462493894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,2,64,balanced,0.04041728019714356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,2,64,balanced,0.04262144088745117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,2,64,balanced,0.04770815849304199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,2,64,balanced,0.053579521179199216
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,2,64,balanced,0.06867199897766113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,2,64,balanced,0.08077440261840821
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,2,64,balanced,0.10651007652282715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,2,64,balanced,0.13377023696899415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,2,64,balanced,0.035715839862823485
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,1,128,power_law_1.2,0.04066431999206543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,1,128,power_law_1.2,0.07134335994720459
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,1,128,power_law_1.2,0.10912768363952637
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,1,128,power_law_1.2,0.158088960647583
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,2,64,balanced,0.01621119976043701
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,2,64,balanced,0.016145919561386106
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,2,64,balanced,0.01666432023048401
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,2,64,balanced,0.01613312005996704
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,2,64,balanced,0.016503039598464966
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,2,64,balanced,0.01692288041114807
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,2,64,balanced,0.016596479415893556
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,2,64,balanced,0.0168230402469635
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,2,64,balanced,0.01701375961303711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,2,64,balanced,0.016583679914474486
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,2,64,balanced,0.01696768045425415
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,2,64,balanced,0.016995840072631836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,2,64,balanced,0.016925439834594727
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,2,64,balanced,0.017331199645996092
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,2,64,balanced,0.01755903959274292
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,2,64,balanced,0.01615872025489807
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,2,64,balanced,0.017781759500503543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,2,64,balanced,0.016547839641571045
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,2,64,balanced,0.017254400253295898
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,2,64,balanced,0.017797119617462158
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,2,64,power_law_1.01,0.025177600383758547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,2,64,balanced,0.016523519754409792
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,2,64,power_law_1.01,0.024874238967895507
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,2,64,power_law_1.01,0.025601279735565186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,2,64,power_law_1.01,0.039521279335021975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,2,64,power_law_1.01,0.04237823963165283
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,2,64,power_law_1.01,0.042812161445617676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,2,64,power_law_1.01,0.025614080429077146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,2,64,power_law_1.01,0.04437888145446777
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,2,64,power_law_1.01,0.04439167976379395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,2,64,power_law_1.01,0.046071038246154786
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,2,64,power_law_1.01,0.0478937578201294
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,2,64,power_law_1.01,0.049305601119995116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,2,64,power_law_1.01,0.057058558464050294
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,2,64,power_law_1.01,0.056880640983581546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,2,64,power_law_1.01,0.06255743980407716
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,2,64,power_law_1.01,0.0688652801513672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,2,64,power_law_1.01,0.07165184020996093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,2,64,power_law_1.01,0.08126848220825196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,2,64,power_law_1.01,0.09611007690429688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,2,64,power_law_1.01,0.1061094379425049
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,2,64,power_law_1.01,0.12402175903320312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,2,64,power_law_1.01,0.174071044921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,2,64,power_law_1.01,0.21240575790405272
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,2,64,power_law_1.01,0.2958195114135742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,2,64,power_law_1.01,0.4188377761840821
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,2,64,power_law_1.01,0.6953958129882812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,2,64,power_law_1.01,0.027246079444885253
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,2,64,power_law_1.01,0.8383475494384764
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,2,64,power_law_1.01,0.026859519481658933
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,2,64,power_law_1.01,0.027197439670562745
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,2,64,power_law_1.01,0.027233281135559083
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,2,64,power_law_1.01,0.0325708794593811
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,2,64,power_law_1.01,0.033704960346221925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,2,64,power_law_1.01,0.03379071950912475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,2,64,power_law_1.01,0.034362881183624266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,2,64,power_law_1.01,0.03460864067077637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,2,64,power_law_1.01,0.035100159645080564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,2,64,power_law_1.01,0.0353766393661499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,2,64,power_law_1.01,0.03701247930526733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,2,64,power_law_1.01,0.03666815996170044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,2,64,power_law_1.01,0.04020736217498779
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,2,64,power_law_1.01,0.041117439270019526
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,2,64,power_law_1.01,0.04033408164978027
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,2,64,power_law_1.01,0.04070911884307861
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,2,64,power_law_1.01,0.04607232093811035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,2,64,power_law_1.01,0.0539353609085083
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,2,64,power_law_1.01,0.060523519515991206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,2,64,power_law_1.01,0.06958975791931152
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,2,64,power_law_1.01,0.09663104057312012
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,2,64,power_law_1.01,0.11174655914306642
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,2,64,power_law_1.01,0.1681331253051758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,2,64,power_law_1.01,0.22914176940917969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,2,64,power_law_1.01,0.3067967987060547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,2,64,power_law_1.01,0.4499840164184571
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,2,64,power_law_1.01,0.016992000341415407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,2,64,power_law_1.01,0.016609280109405516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,2,64,power_law_1.01,0.01665663957595825
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,2,64,power_law_1.01,0.016520960330963137
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,2,64,power_law_1.01,0.016582399606704712
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,2,64,power_law_1.01,0.01659775972366333
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,2,64,power_law_1.01,0.016659200191497803
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,2,64,power_law_1.01,0.016506880521774292
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,2,64,power_law_1.01,0.01658751964569092
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,2,64,balanced,0.016666879653930662
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,2,64,power_law_1.01,0.0169267201423645
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,2,64,power_law_1.01,0.01699839949607849
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,2,64,balanced,0.016992000341415407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,2,64,power_law_1.01,0.016933120489120483
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,2,64,power_law_1.01,0.01654528021812439
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,2,64,power_law_1.01,0.01699839949607849
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,2,64,power_law_1.01,0.017390079498291015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,2,64,power_law_1.01,0.01733504056930542
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,2,64,power_law_1.01,0.0172761595249176
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,2,64,balanced,0.020225279331207276
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,2,64,balanced,0.029278719425201417
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,2,64,balanced,0.021111040115356444
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,2,64,balanced,0.023929600715637205
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,2,64,power_law_1.01,0.018220800161361694
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,2,64,power_law_1.01,0.018800640106201173
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,2,64,power_law_1.01,0.020276479721069336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,2,64,power_law_1.01,0.021926400661468507
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,2,64,power_law_1.2,0.024843521118164062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,2,64,power_law_1.2,0.025655040740966795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,2,64,power_law_1.2,0.025830399990081788
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,2,64,power_law_1.2,0.024714241027832033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,2,64,power_law_1.2,0.04026495933532715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,2,64,power_law_1.2,0.04192768096923828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,2,64,power_law_1.2,0.043553280830383304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,2,64,power_law_1.2,0.04833024024963379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,2,64,power_law_1.2,0.04482816219329834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,2,64,power_law_1.2,0.04724095821380615
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,2,64,power_law_1.2,0.049111042022705075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,2,64,power_law_1.2,0.05877247810363769
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,2,64,power_law_1.2,0.0640064001083374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,2,64,power_law_1.2,0.0676262378692627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,2,64,power_law_1.2,0.07489408016204833
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,2,64,power_law_1.2,0.06788352012634277
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,2,64,power_law_1.2,0.0777561616897583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,2,64,power_law_1.2,0.09432319641113282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,2,64,power_law_1.2,0.10362367630004883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,2,64,power_law_1.2,0.12406911849975585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,2,64,power_law_1.2,0.14878975868225097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,2,64,power_law_1.2,0.21082752227783202
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,2,64,power_law_1.2,0.2926310348510742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,2,64,power_law_1.2,0.43765888214111326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,2,64,power_law_1.2,0.6054912185668945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,2,64,power_law_1.2,0.8770098876953124
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,2,64,power_law_1.01,0.026818559169769284
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,2,64,power_law_1.2,0.02644223928451538
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,2,64,power_law_1.2,0.02644479990005493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,2,64,power_law_1.2,1.3909260559082033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,2,64,power_law_1.2,0.02685823917388916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,2,64,power_law_1.2,0.03218175888061524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,2,64,power_law_1.2,0.032960000038146975
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,2,64,power_law_1.01,0.0415718412399292
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,2,64,power_law_1.2,0.03342080116271973
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,2,64,power_law_1.2,0.03337599992752075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,2,64,power_law_1.2,0.034173440933227536
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,2,64,power_law_1.2,0.035022079944610596
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,2,64,power_law_1.2,0.03536256074905396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,2,64,power_law_1.2,0.03619712114334107
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,2,64,power_law_1.2,0.036646399497985836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,2,64,power_law_1.2,0.03792383909225464
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,2,64,power_law_1.2,0.040527358055114746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,2,64,power_law_1.2,0.039146239757537844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,2,64,power_law_1.2,0.04034560203552246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,2,64,power_law_1.2,0.04333439826965332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,2,64,power_law_1.2,0.052803840637207035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,2,64,power_law_1.2,0.05828864097595214
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,2,64,power_law_1.2,0.06891647815704346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,2,64,power_law_1.2,0.08775936126708984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,2,64,power_law_1.2,0.11486592292785644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,2,64,power_law_1.2,0.14909567832946777
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,2,64,power_law_1.01,0.053396477699279785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,2,64,power_law_1.2,0.24485376358032224
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,2,64,power_law_1.2,0.35002113342285157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,2,64,power_law_1.2,0.016599040031433105
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,2,64,power_law_1.2,0.016893440485000612
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,2,64,power_law_1.2,0.01651968002319336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,2,64,power_law_1.2,0.48228607177734373
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,2,64,power_law_1.2,0.01657088041305542
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,2,64,power_law_1.2,0.01651455998420715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,2,64,power_law_1.2,0.5990886306762695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,2,64,power_law_1.2,0.01660159945487976
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,2,64,power_law_1.2,0.016624640226364135
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,2,64,power_law_1.2,0.016898560523986816
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,2,64,power_law_1.2,0.016753920316696168
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,2,64,power_law_1.2,0.01699839949607849
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,2,64,power_law_1.2,0.016583679914474486
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,2,64,power_law_1.2,0.01693184018135071
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,2,64,power_law_1.2,0.016984319686889647
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,2,64,power_law_1.2,0.01699839949607849
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,2,64,power_law_1.2,0.0169267201423645
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,2,64,power_law_1.2,0.017318400144577025
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,2,64,power_law_1.01,0.03160063982009888
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,2,64,power_law_1.01,0.06360191822052001
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,2,64,power_law_1.01,0.08339839935302734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,2,64,power_law_1.2,0.018981120586395263
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,2,64,power_law_1.2,0.02019711971282959
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,2,64,power_law_1.2,0.026350080966949463
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,2,64,power_law_1.2,0.018192640542984008
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,2,64,power_law_1.2,0.02254591941833496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,4,32,balanced,0.023133440017700197
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,4,32,balanced,0.02200448036193848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,4,32,balanced,0.02285696029663086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,4,32,balanced,0.021939198970794677
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,4,32,balanced,0.02319103956222534
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,4,32,balanced,0.02602112054824829
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,4,32,balanced,0.041137919425964356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,4,32,balanced,0.040584959983825684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,4,32,balanced,0.043196158409118654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,4,32,balanced,0.04285312175750732
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,4,32,balanced,0.046835198402404785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,4,32,balanced,0.04472576141357422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,4,32,balanced,0.05162623882293701
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,4,32,balanced,0.04860288143157959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,4,32,balanced,0.05503359794616699
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,4,32,balanced,0.05169407844543457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,4,32,balanced,0.059022078514099116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,4,32,balanced,0.07048575878143311
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,4,32,balanced,0.07678847789764405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,4,32,balanced,0.08736639976501465
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,2,64,power_law_1.2,0.034773759841918946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,4,32,balanced,0.10822655677795409
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,4,32,balanced,0.12340096473693847
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,4,32,balanced,0.15828607559204103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,4,32,balanced,0.19184511184692382
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,4,32,balanced,0.06498303890228271
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,4,32,balanced,0.2616819190979004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,4,32,balanced,0.025208320617675785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,4,32,balanced,0.3315532684326172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,4,32,balanced,0.02520576000213623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,4,32,balanced,0.02560256004333496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,4,32,balanced,0.025972480773925784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,4,32,balanced,0.027765760421752928
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,4,32,balanced,0.03294975996017456
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,4,32,balanced,0.0329036808013916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,4,32,balanced,0.0337779188156128
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,4,32,balanced,0.03371776103973388
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,4,32,balanced,0.03422336101531982
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,4,32,balanced,0.0337932801246643
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,4,32,balanced,0.034982399940490724
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,4,32,balanced,0.037056000232696534
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,4,32,balanced,0.025504000186920166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,4,32,balanced,0.039157760143280027
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,4,32,balanced,0.04068607807159424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,4,32,balanced,0.03698431968688965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,4,32,balanced,0.04279935836791992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,4,32,balanced,0.044673280715942384
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,4,32,balanced,0.049597439765930174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,4,32,balanced,0.06221183776855469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,4,32,balanced,0.07389440059661864
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,4,32,balanced,0.09975808143615723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,4,32,balanced,0.11954943656921387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,4,32,balanced,0.1666419219970703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,4,32,balanced,0.21302656173706053
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,4,32,balanced,0.03992063999176025
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,2,64,power_law_1.2,0.04142591953277588
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,2,64,power_law_1.2,0.054456319808959965
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,2,64,power_law_1.2,0.06769408226013184
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,2,64,power_law_1.2,0.10307583808898926
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,2,64,power_law_1.2,0.13543423652648925
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,4,32,balanced,0.01685119986534119
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,4,32,balanced,0.01861631989479065
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,4,32,balanced,0.016953599452972413
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,4,32,balanced,0.018558720350265505
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,4,32,balanced,0.018236160278320312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,4,32,balanced,0.018568960428237916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,4,32,balanced,0.01863935947418213
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,4,32,balanced,0.01828223943710327
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,4,32,balanced,0.018557440042495727
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,4,32,balanced,0.018986239433288574
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,4,32,balanced,0.018986239433288574
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,4,32,balanced,0.018978559970855714
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,4,32,balanced,0.019047679901123046
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,4,32,balanced,0.019466240406036377
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,4,32,balanced,0.019816960096359255
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,4,32,balanced,0.020383999347686768
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,4,32,balanced,0.021021440029144287
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,4,32,balanced,0.01687680006027222
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,4,32,balanced,0.017169920206069948
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,4,32,balanced,0.017647360563278196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,4,32,power_law_1.01,0.027200000286102293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,4,32,power_law_1.01,0.025579519271850586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,4,32,power_law_1.01,0.026801919937133788
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,4,32,power_law_1.01,0.02681087970733643
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,4,32,power_law_1.01,0.026807041168212892
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,4,32,power_law_1.01,0.04112895965576172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,4,32,power_law_1.01,0.04255743980407715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,4,32,power_law_1.01,0.04327936172485351
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,4,32,power_law_1.01,0.04451456069946289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,4,32,power_law_1.01,0.04680575847625733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,4,32,power_law_1.01,0.04977280139923096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,4,32,power_law_1.01,0.051696639060974124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,4,32,power_law_1.01,0.05289216041564941
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,4,32,power_law_1.01,0.06167295932769775
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,4,32,power_law_1.01,0.06414463996887207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,4,32,power_law_1.01,0.06128255844116211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,4,32,power_law_1.01,0.07062655925750733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,4,32,power_law_1.01,0.08594047546386718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,4,32,power_law_1.01,0.0941977596282959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,4,32,power_law_1.01,0.10936832427978516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,4,32,power_law_1.01,0.13147775650024413
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,4,32,power_law_1.01,0.191144962310791
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,4,32,balanced,0.018599679470062254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,4,32,power_law_1.01,0.20671615600585938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,4,32,balanced,0.018640639781951903
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,4,32,balanced,0.018457599878311158
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,4,32,power_law_1.01,0.2895462417602539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,4,32,power_law_1.01,0.02721791982650757
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,4,32,power_law_1.01,0.5664051055908204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,4,32,power_law_1.01,0.027745280265808102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,4,32,power_law_1.01,0.6987891387939453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,4,32,power_law_1.01,0.027182080745697022
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,4,32,power_law_1.01,0.02805248022079468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,4,32,power_law_1.01,0.028830718994140626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,4,32,power_law_1.01,0.03291264057159424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,4,32,power_law_1.01,0.033320960998535154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,4,32,power_law_1.01,0.037480959892272944
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,4,32,power_law_1.01,0.034142720699310306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,4,32,power_law_1.01,0.03442816019058227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,4,32,power_law_1.01,0.37949951171874996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,4,32,power_law_1.01,0.03660032033920288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,4,32,power_law_1.01,0.03862656116485595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,4,32,power_law_1.01,0.04075136184692383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,4,32,power_law_1.01,0.04079103946685791
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,4,32,power_law_1.01,0.038680319786071775
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,4,32,power_law_1.01,0.03986176013946533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,4,32,power_law_1.01,0.0435916805267334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,4,32,power_law_1.01,0.04915328025817871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,4,32,power_law_1.01,0.05382912158966065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,4,32,power_law_1.01,0.06180736064910889
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,4,32,power_law_1.01,0.08392191886901855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,4,32,power_law_1.01,0.11016960144042968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,4,32,power_law_1.01,0.13773823738098145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,4,32,power_law_1.01,0.16947584152221679
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,4,32,power_law_1.01,0.22896127700805663
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,4,32,power_law_1.01,0.01862912058830261
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,4,32,power_law_1.01,0.35972606658935546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,4,32,power_law_1.01,0.018650879859924314
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,4,32,power_law_1.01,0.01855487942695618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,4,32,power_law_1.01,0.4947366333007812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,4,32,power_law_1.01,0.01853440046310425
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,4,32,power_law_1.01,0.01776255965232849
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,4,32,power_law_1.01,0.018309119939804077
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,4,32,power_law_1.01,0.018568960428237916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,4,32,power_law_1.01,0.018403840065002442
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,4,32,power_law_1.01,0.01831168055534363
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,4,32,power_law_1.01,0.018533120155334475
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,4,32,power_law_1.01,0.018568960428237916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,4,32,power_law_1.01,0.019020800590515134
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,4,32,power_law_1.01,0.019383039474487305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,4,32,power_law_1.01,0.019096319675445554
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,4,32,power_law_1.01,0.019409919977188113
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,4,32,power_law_1.01,0.01861631989479065
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,4,32,power_law_1.01,0.018625279664993284
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,4,32,balanced,0.02437119960784912
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,4,32,balanced,0.025607678890228268
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,4,32,balanced,0.0334335994720459
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,4,32,balanced,0.04107391834259033
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,4,32,power_law_1.01,0.020587520599365233
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,4,32,power_law_1.01,0.021048319339752198
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,4,32,power_law_1.01,0.023203840255737303
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,4,32,power_law_1.2,0.02754816055297852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,4,32,power_law_1.2,0.02482176065444946
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,4,32,power_law_1.01,0.026586880683898927
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,4,32,power_law_1.2,0.02642431974411011
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,4,32,power_law_1.2,0.027668480873107908
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,4,32,power_law_1.2,0.040308480262756345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,4,32,power_law_1.2,0.043184638023376465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,4,32,power_law_1.2,0.044359679222106936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,4,32,power_law_1.2,0.04764416217803955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,4,32,power_law_1.2,0.044677119255065914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,4,32,power_law_1.2,0.0456115198135376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,4,32,power_law_1.2,0.04924543857574463
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,4,32,power_law_1.2,0.05556992053985595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,4,32,power_law_1.2,0.07068031787872314
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,4,32,power_law_1.01,0.03194495916366577
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,4,32,power_law_1.2,0.06679679870605468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,4,32,power_law_1.2,0.0682150411605835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,4,32,power_law_1.2,0.07771903991699218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,4,32,power_law_1.2,0.08848128318786622
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,4,32,power_law_1.01,0.03947648048400879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,4,32,power_law_1.2,0.09820672035217284
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,4,32,power_law_1.2,0.12469375610351563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,4,32,power_law_1.2,0.15724032402038574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,4,32,power_law_1.2,0.21116544723510744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,4,32,power_law_1.2,0.27476991653442384
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,4,32,power_law_1.2,0.06606976032257081
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,4,32,power_law_1.2,0.4087564849853516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,4,32,power_law_1.2,0.5140031814575196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,4,32,power_law_1.2,0.028701438903808597
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,4,32,power_law_1.01,0.06036479949951172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,4,32,power_law_1.2,0.8531865692138672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,4,32,power_law_1.2,0.027617280483245847
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,4,32,power_law_1.2,0.027737600803375246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,4,32,power_law_1.2,1.0673343658447265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,4,32,power_law_1.2,0.027996160984039307
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,4,32,power_law_1.2,0.03218048095703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,4,32,power_law_1.2,0.03323647975921631
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,4,32,power_law_1.2,0.03583616018295288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,4,32,power_law_1.2,0.034193921089172366
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,4,32,power_law_1.01,0.048739838600158694
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,4,32,power_law_1.2,0.03458048105239868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,4,32,power_law_1.2,0.03582079887390137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,4,32,power_law_1.2,0.037611520290374754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,4,32,power_law_1.2,0.039870719909667965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,4,32,power_law_1.2,0.03869440078735352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,4,32,power_law_1.2,0.03955199956893921
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,4,32,power_law_1.2,0.03953536033630371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,4,32,power_law_1.2,0.04074240207672119
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,4,32,power_law_1.2,0.04422272205352783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,4,32,power_law_1.2,0.050531840324401854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,4,32,power_law_1.2,0.056862721443176265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,4,32,power_law_1.2,0.07728511810302734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,4,32,power_law_1.2,0.08887167930603027
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,4,32,power_law_1.2,0.13328512191772462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,4,32,power_law_1.2,0.16002559661865234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,4,32,power_law_1.2,0.2401203155517578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,4,32,power_law_1.2,0.01860479950904846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,4,32,power_law_1.2,0.3482636642456055
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,4,32,power_law_1.2,0.01861631989479065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,4,32,power_law_1.2,0.4786572647094727
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,4,32,power_law_1.2,0.018557440042495727
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,4,32,power_law_1.2,0.01847551941871643
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,4,32,power_law_1.2,0.01821823954582214
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,4,32,power_law_1.2,0.017859840393066408
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,4,32,power_law_1.2,0.6345305633544922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,4,32,power_law_1.2,0.018627840280532836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,4,32,power_law_1.2,0.018583040237426757
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,4,32,power_law_1.2,0.018186240196228026
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,4,32,power_law_1.2,0.01835904002189636
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,4,32,power_law_1.2,0.018626559972763062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,4,32,power_law_1.2,0.018929920196533202
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,4,32,power_law_1.2,0.019031039476394652
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,4,32,power_law_1.2,0.01906559944152832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,4,32,power_law_1.2,0.019471360445022585
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,4,32,power_law_1.2,0.02075648069381714
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,4,32,power_law_1.2,0.023147521018981935
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,4,32,power_law_1.2,0.01859328031539917
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,4,32,power_law_1.01,0.08759807586669922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,4,32,power_law_1.01,0.1008844757080078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,4,32,power_law_1.2,0.019896320104598998
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,4,32,power_law_1.2,0.042305278778076175
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,4,32,power_law_1.2,0.02534656047821045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,8,16,balanced,0.02147327899932861
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,8,16,balanced,0.02162303924560547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,8,16,balanced,0.022675199508666994
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,8,16,balanced,0.02479487895965576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,8,16,balanced,0.028400640487670898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,8,16,balanced,0.0444595193862915
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,4,32,power_law_1.2,0.06540287971496582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,8,16,balanced,0.04481152057647705
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,4,32,power_law_1.2,0.046695680618286134
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,8,16,balanced,0.04650368213653565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,8,16,balanced,0.04684288024902344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,8,16,balanced,0.0506982421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,8,16,balanced,0.0529369592666626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,8,16,balanced,0.05230207920074463
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,8,16,balanced,0.05315199851989746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,8,16,balanced,0.06416384220123292
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,8,16,balanced,0.058197760581970216
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,8,16,balanced,0.06388351917266846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,8,16,balanced,0.07508096218109131
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,8,16,balanced,0.08578559875488281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,8,16,balanced,0.10259967803955079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,8,16,balanced,0.11739775657653809
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,8,16,balanced,0.1503436756134033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,8,16,balanced,0.18035327911376953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,8,16,balanced,0.2414067268371582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,8,16,balanced,0.3048857688903809
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,8,16,balanced,0.025148160457611084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,8,16,balanced,0.02144511938095093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,8,16,balanced,0.429246711730957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,8,16,balanced,0.025148160457611084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,8,16,balanced,0.025120000839233398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,8,16,balanced,0.02598400115966797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,8,16,balanced,0.5493107223510743
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,8,16,balanced,0.02768383979797363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,8,16,balanced,0.04035583972930908
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,8,16,balanced,0.03280767917633057
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,8,16,balanced,0.044893441200256345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,8,16,balanced,0.04070144176483154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,8,16,balanced,0.040735998153686524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,8,16,balanced,0.042074880599975585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,8,16,balanced,0.0431334400177002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,8,16,balanced,0.04406784057617187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,8,16,balanced,0.0399782395362854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,8,16,balanced,0.043624958992004394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,8,16,balanced,0.04564735889434814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,8,16,balanced,0.04778111934661865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,8,16,balanced,0.05255551815032959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,8,16,balanced,0.06293504238128662
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,8,16,balanced,0.07387263774871826
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,4,32,power_law_1.2,0.03010175943374634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,8,16,balanced,0.09580415725708008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,8,16,balanced,0.11534463882446289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,8,16,balanced,0.16327936172485352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,8,16,balanced,0.20268928527832034
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,8,16,balanced,0.28829824447631835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,8,16,balanced,0.04205952167510986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,8,16,balanced,0.3764851379394531
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,4,32,power_law_1.2,0.08380415916442871
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,4,32,power_law_1.2,0.12446975708007812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,4,32,power_law_1.2,0.17148544311523437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,8,16,balanced,0.017779200077056884
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,8,16,balanced,0.017830400466918944
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,8,16,balanced,0.021004800796508787
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,8,16,balanced,0.021035521030426024
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,8,16,balanced,0.021032960414886476
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,8,16,balanced,0.02106879949569702
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,8,16,balanced,0.021089279651641847
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,8,16,balanced,0.02149120092391968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,8,16,balanced,0.017785600423812865
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,8,16,balanced,0.0194649600982666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,8,16,balanced,0.021437439918518066
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,8,16,balanced,0.021573119163513184
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,8,16,balanced,0.022254080772399903
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,8,16,balanced,0.02310784101486206
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,8,16,balanced,0.018643200397491455
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,8,16,balanced,0.023425281047821045
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,8,16,balanced,0.024839680194854736
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,8,16,balanced,0.02108544111251831
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,8,16,balanced,0.02601599931716919
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,8,16,balanced,0.022255361080169678
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,8,16,power_law_1.01,0.024744958877563478
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,8,16,power_law_1.01,0.02558079957962036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,8,16,power_law_1.01,0.02599936008453369
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,8,16,power_law_1.01,0.026822400093078614
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,8,16,power_law_1.01,0.029221119880676272
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,8,16,power_law_1.01,0.04320127964019775
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,8,16,power_law_1.01,0.044042239189147944
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,8,16,power_law_1.01,0.04467584133148193
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,8,16,power_law_1.01,0.049262080192565914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,8,16,power_law_1.01,0.05247231960296631
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,8,16,power_law_1.01,0.05405312061309815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,8,16,power_law_1.01,0.05436287879943848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,8,16,power_law_1.01,0.061977601051330565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,8,16,power_law_1.01,0.06503168106079102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,8,16,power_law_1.01,0.0679475212097168
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,8,16,balanced,0.020636160373687745
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,8,16,power_law_1.01,0.07095935821533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,8,16,power_law_1.01,0.0757913589477539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,8,16,power_law_1.01,0.09258496284484863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,8,16,power_law_1.01,0.10067199707031249
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,8,16,balanced,0.02079103946685791
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,8,16,power_law_1.01,0.12395520210266113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,8,16,power_law_1.01,0.14150912284851075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,8,16,power_law_1.01,0.1693337631225586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,8,16,power_law_1.01,0.23857791900634764
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,8,16,power_law_1.01,0.3161382484436035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,8,16,power_law_1.01,0.4086041641235352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,8,16,power_law_1.01,0.026791679859161376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,8,16,power_law_1.01,0.6353036880493164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,8,16,power_law_1.01,0.030920960903167725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,8,16,power_law_1.01,0.03084800004959106
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,8,16,power_law_1.01,0.8263308715820312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,8,16,power_law_1.01,0.03126528024673462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,8,16,power_law_1.01,0.032156159877777105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,8,16,power_law_1.01,0.035033600330352785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,8,16,power_law_1.01,0.039137279987335204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,8,16,power_law_1.01,0.04108799934387207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,8,16,power_law_1.01,0.040273919105529785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,8,16,power_law_1.01,0.04159359931945801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,8,16,power_law_1.01,0.04278143882751465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,8,16,power_law_1.01,0.0448473596572876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,8,16,power_law_1.01,0.04477695941925049
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,8,16,power_law_1.01,0.04283008098602295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,8,16,power_law_1.01,0.043541760444641114
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,8,16,power_law_1.01,0.04434815883636475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,8,16,power_law_1.01,0.05599487781524658
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,8,16,power_law_1.01,0.062202877998352044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,8,16,power_law_1.01,0.08134783744812012
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,8,16,power_law_1.01,0.09478400230407716
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,8,16,power_law_1.01,0.12140288352966308
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,8,16,power_law_1.01,0.1537715244293213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,8,16,power_law_1.01,0.2143564796447754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,8,16,power_law_1.01,0.04960000038146973
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,8,16,power_law_1.01,0.3028620719909668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,8,16,power_law_1.01,0.42790401458740235
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,8,16,power_law_1.01,0.019481600522994996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,8,16,power_law_1.01,0.020665600299835205
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,8,16,power_law_1.01,0.020682239532470705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,8,16,power_law_1.01,0.5549055862426757
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,8,16,power_law_1.01,0.020632319450378418
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,8,16,power_law_1.01,0.020706560611724854
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,8,16,power_law_1.01,0.0205785608291626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,8,16,power_law_1.01,0.020689918994903564
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,8,16,power_law_1.01,0.020624639987945555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,8,16,power_law_1.01,0.020714240074157713
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,8,16,power_law_1.01,0.020508160591125486
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,8,16,power_law_1.01,0.021012480258941653
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,8,16,power_law_1.01,0.021002240180969238
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,8,16,power_law_1.01,0.02142080068588257
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,8,16,power_law_1.01,0.021356799602508546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,8,16,power_law_1.01,0.021502718925476075
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,8,16,power_law_1.01,0.02189824104309082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,8,16,balanced,0.021030399799346924
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,8,16,power_law_1.01,0.022319359779357912
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,8,16,balanced,0.062154240608215325
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,8,16,balanced,0.034584319591522215
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,8,16,balanced,0.05012735843658447
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,8,16,balanced,0.03828351974487305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,8,16,power_law_1.01,0.02430464029312134
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,8,16,power_law_1.01,0.02556544065475464
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,8,16,power_law_1.01,0.030115840435028074
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,8,16,power_law_1.01,0.03567872047424316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,8,16,power_law_1.2,0.023055360317230225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,8,16,power_law_1.2,0.026023681163787844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,8,16,power_law_1.2,0.026789119243621828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,8,16,power_law_1.2,0.027653119564056396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,8,16,power_law_1.2,0.028879361152648924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,8,16,power_law_1.2,0.046311678886413576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,8,16,power_law_1.2,0.04600831985473633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,8,16,power_law_1.2,0.04483327865600586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,8,16,power_law_1.2,0.04619008064270019
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,8,16,power_law_1.2,0.05013120174407959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,8,16,power_law_1.2,0.053221120834350585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,8,16,power_law_1.2,0.059399681091308595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,8,16,power_law_1.2,0.06270463943481445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,8,16,power_law_1.2,0.06590208053588867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,8,16,power_law_1.2,0.06968319892883301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,8,16,power_law_1.2,0.07497983932495117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,8,16,power_law_1.2,0.0805568027496338
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,8,16,power_law_1.2,0.09404159545898437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,8,16,power_law_1.2,0.10675328254699708
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,8,16,power_law_1.2,0.12434176445007324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,8,16,power_law_1.2,0.15238016128540038
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,8,16,power_law_1.2,0.20115327835083008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,8,16,power_law_1.2,0.2535487937927246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,8,16,power_law_1.2,0.3652582550048828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,8,16,power_law_1.2,0.5073855972290039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,8,16,power_law_1.2,0.7230912017822265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,8,16,power_law_1.2,0.02727423906326294
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,8,16,power_law_1.01,0.04587520122528076
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,8,16,power_law_1.2,0.031267840862274174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,8,16,power_law_1.2,0.9338829040527343
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,8,16,power_law_1.01,0.05480063915252685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,8,16,power_law_1.2,0.03092736005783081
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,8,16,power_law_1.2,0.03173120021820068
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,8,16,power_law_1.2,0.03056256055831909
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,8,16,power_law_1.2,0.03618815898895263
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,8,16,power_law_1.2,0.03960063934326172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,8,16,power_law_1.2,0.040021758079528805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,8,16,power_law_1.2,0.04070144176483154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,8,16,power_law_1.2,0.041507840156555176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,8,16,power_law_1.2,0.044048638343811036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,8,16,power_law_1.2,0.043632640838623046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,8,16,power_law_1.2,0.04222080230712891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,8,16,power_law_1.2,0.043066878318786625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,8,16,power_law_1.2,0.04396031856536865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,8,16,power_law_1.2,0.0448576021194458
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,8,16,power_law_1.2,0.05121920108795166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,8,16,power_law_1.2,0.05813504219055175
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,8,16,power_law_1.01,0.0738431978225708
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,8,16,power_law_1.2,0.06664576053619384
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,8,16,power_law_1.2,0.08316032409667969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,8,16,power_law_1.2,0.10803456306457519
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,8,16,power_law_1.2,0.14352895736694335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,8,16,power_law_1.2,0.19794815063476562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,8,16,power_law_1.2,0.29246335983276367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,8,16,power_law_1.2,0.37523071289062504
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,8,16,power_law_1.2,0.019467519521713255
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,8,16,power_law_1.2,0.020692479610443116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,8,16,power_law_1.2,0.5250649642944336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,8,16,power_law_1.2,0.020611839294433592
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,8,16,power_law_1.2,0.02079103946685791
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,8,16,power_law_1.01,0.08710271835327148
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,8,16,power_law_1.2,0.02068864107131958
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,8,16,power_law_1.2,0.02035327911376953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,8,16,power_law_1.2,0.020199680328369142
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,8,16,power_law_1.2,0.021062400341033936
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,8,16,power_law_1.2,0.020684800148010253
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,8,16,power_law_1.2,0.020670719146728515
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,8,16,power_law_1.2,0.02061311960220337
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,8,16,power_law_1.2,0.021079039573669432
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,8,16,power_law_1.2,0.02189824104309082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,8,16,power_law_1.2,0.021268479824066162
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,8,16,power_law_1.2,0.02222208023071289
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,8,16,power_law_1.2,0.02363264083862305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,8,16,power_law_1.2,0.6983910369873046
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,8,16,power_law_1.2,0.020281600952148437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,8,16,power_law_1.01,0.11991552352905273
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,8,16,power_law_1.01,0.1643391990661621
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,8,16,power_law_1.2,0.025189120769500733
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,8,16,power_law_1.2,0.037557759284973145
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,8,16,power_law_1.2,0.07228799819946288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,16,8,balanced,0.020803840160369874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,16,8,balanced,0.02229759931564331
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,8,16,power_law_1.2,0.030412800312042236
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,8,16,power_law_1.2,0.04585984230041504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,16,8,balanced,0.023146240711212157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,16,8,balanced,0.025623040199279783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,16,8,balanced,0.028828160762786863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,16,8,balanced,0.03490943908691406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,16,8,balanced,0.05174784183502197
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,16,8,balanced,0.053442559242248534
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,16,8,balanced,0.05513855934143066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,16,8,balanced,0.05810431957244873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,16,8,balanced,0.05336703777313232
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,8,16,power_law_1.2,0.053592319488525386
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,16,8,balanced,0.05752448081970215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,16,8,balanced,0.06320127964019776
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,16,8,balanced,0.06617599964141846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,16,8,balanced,0.0733350419998169
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,16,8,balanced,0.08061823844909669
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,16,8,balanced,0.08850432395935058
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,16,8,balanced,0.10404735565185548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,16,8,balanced,0.11969792366027832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,16,8,balanced,0.14946816444396974
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,16,8,balanced,0.1795212745666504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,16,8,balanced,0.2376908874511719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,16,8,balanced,0.2970495986938476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,16,8,balanced,0.4170739364624024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,16,8,balanced,0.5308095932006835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,16,8,balanced,0.025418241024017334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,16,8,balanced,0.02603775978088379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,16,8,balanced,0.7603199768066407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,16,8,balanced,0.02797696113586426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,16,8,balanced,0.030497279167175297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,16,8,balanced,0.03292927980422974
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,16,8,balanced,0.039129600524902344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,16,8,balanced,0.04576384067535401
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,16,8,balanced,0.04677248001098633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,16,8,balanced,0.048579840660095214
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,16,8,balanced,0.9926374053955078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,16,8,balanced,0.050991358757019045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,16,8,balanced,0.04607872009277344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,16,8,balanced,0.04776703834533692
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,16,8,balanced,0.04813055992126465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,16,8,balanced,0.05071872234344482
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,16,8,balanced,0.05133823871612549
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,16,8,balanced,0.05265664100646973
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,16,8,balanced,0.05719935894012451
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,16,8,balanced,0.06789504051208496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,16,8,balanced,0.07903359889984131
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,16,8,balanced,0.09769215583801269
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,16,8,balanced,0.11861503601074219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,16,8,balanced,0.15898624420166016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,16,8,balanced,0.1958451271057129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,16,8,balanced,0.28542207717895507
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,16,8,balanced,0.3604006576538086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,16,8,balanced,0.520814094543457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,16,8,balanced,0.6787110137939453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,8,16,power_law_1.2,0.09444479942321778
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,8,16,power_law_1.2,0.1090112018585205
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,8,16,power_law_1.2,0.1769215965270996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,8,16,power_law_1.2,0.2442521667480469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,16,8,balanced,0.02077440023422241
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,16,8,balanced,0.021593599319458007
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,16,8,balanced,0.027701759338378908
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,16,8,balanced,0.022693119049072265
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,16,8,balanced,0.027808001041412355
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,16,8,balanced,0.028025600910186764
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,16,8,balanced,0.028385279178619383
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,16,8,balanced,0.027823359966278077
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,16,8,balanced,0.028532478809356686
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,16,8,balanced,0.02147455930709839
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,16,8,balanced,0.027966721057891848
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,16,8,balanced,0.028517119884490967
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,16,8,balanced,0.029745919704437258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,16,8,balanced,0.03011840105056763
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,16,8,balanced,0.031783680915832516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,16,8,balanced,0.03343103885650635
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,16,8,balanced,0.03536895990371704
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,16,8,balanced,0.039261438846588136
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,16,8,balanced,0.023878400325775144
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,16,8,balanced,0.028903679847717283
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,16,8,power_law_1.01,0.025968639850616454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,16,8,power_law_1.01,0.026241281032562257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,16,8,power_law_1.01,0.0276364803314209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,16,8,power_law_1.01,0.02982016086578369
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,16,8,power_law_1.01,0.033016319274902346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,16,8,power_law_1.01,0.03468800067901611
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,16,8,power_law_1.01,0.050961918830871575
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,16,8,power_law_1.01,0.05318528175354004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,16,8,power_law_1.01,0.0562713623046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,16,8,power_law_1.01,0.05850751876831055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,16,8,power_law_1.01,0.06488959789276123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,16,8,power_law_1.01,0.06557568073272704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,16,8,power_law_1.01,0.06526336193084717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,16,8,power_law_1.01,0.06888192176818848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,16,8,power_law_1.01,0.07625599861145019
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,16,8,power_law_1.01,0.07839871883392334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,16,8,power_law_1.01,0.08559743881225586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,16,8,power_law_1.01,0.09993984222412108
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,16,8,power_law_1.01,0.11175040245056152
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,16,8,power_law_1.01,0.1376959991455078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,16,8,power_law_1.01,0.16499711990356444
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,16,8,balanced,0.027568640708923342
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,16,8,balanced,0.02814464092254639
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,16,8,power_law_1.01,0.22002048492431642
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,16,8,power_law_1.01,0.266247673034668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,16,8,power_law_1.01,0.34544254302978517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,16,8,power_law_1.01,0.46526206970214845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,16,8,power_law_1.01,0.028437759876251224
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,16,8,power_law_1.01,0.7052288055419922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,16,8,power_law_1.01,0.031315200328826905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,16,8,power_law_1.01,0.03229439973831177
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,16,8,power_law_1.01,0.901418228149414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,16,8,power_law_1.01,0.035033600330352785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,16,8,power_law_1.01,0.03902591943740845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,16,8,power_law_1.01,0.03664511919021606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,16,8,power_law_1.01,0.045171198844909666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,16,8,power_law_1.01,0.045678081512451174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,16,8,power_law_1.01,0.046570239067077634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,16,8,power_law_1.01,0.047784957885742184
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,16,8,power_law_1.01,0.05060480117797852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,16,8,power_law_1.01,0.04854911804199219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,16,8,power_law_1.01,0.04691455841064453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,16,8,power_law_1.01,0.048428797721862794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,16,8,power_law_1.01,0.05021567821502686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,16,8,power_law_1.01,0.05113215923309326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,16,8,power_law_1.01,0.05670527935028076
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,16,8,power_law_1.01,0.06427135944366455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,16,8,power_law_1.01,0.07257984161376954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,16,8,power_law_1.01,0.0931379222869873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,16,8,power_law_1.01,0.10619520187377929
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,16,8,power_law_1.01,0.14845696449279785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,16,8,power_law_1.01,0.19233791351318358
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,16,8,power_law_1.01,0.26847232818603517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,16,8,power_law_1.01,0.32083198547363284
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,16,8,power_law_1.01,0.4776166534423828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,16,8,power_law_1.01,0.02271872043609619
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,16,8,power_law_1.01,0.024316160678863524
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,16,8,power_law_1.01,0.024756479263305663
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,16,8,power_law_1.01,0.5964339065551758
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,16,8,power_law_1.01,0.024688639640808106
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,16,8,power_law_1.01,0.02669440031051636
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,16,8,power_law_1.01,0.0271347188949585
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,16,8,power_law_1.01,0.026750719547271727
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,16,8,power_law_1.01,0.02599040031433105
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,16,8,power_law_1.01,0.025944321155548095
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,16,8,power_law_1.01,0.026163198947906495
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,16,8,power_law_1.01,0.02679807901382446
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,16,8,power_law_1.01,0.027068159580230712
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,16,8,power_law_1.01,0.027463679313659665
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,16,8,power_law_1.01,0.02875391960144043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,16,8,power_law_1.01,0.028424959182739257
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,16,8,power_law_1.01,0.028444159030914306
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,16,8,power_law_1.01,0.030990080833435057
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,16,8,balanced,0.08561663627624512
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,16,8,balanced,0.06516992092132569
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,16,8,balanced,0.05893119812011719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,16,8,balanced,0.10528767585754395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,16,8,power_law_1.01,0.036176640987396244
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,16,8,power_law_1.01,0.03827199935913086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,16,8,power_law_1.01,0.058661122322082516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,16,8,power_law_1.01,0.04850048065185547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,16,8,power_law_1.2,0.023566079139709473
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,16,8,power_law_1.2,0.026063361167907716
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,16,8,power_law_1.2,0.026419200897216798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,16,8,power_law_1.2,0.02845952033996582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,16,8,power_law_1.2,0.03263999938964844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,16,8,power_law_1.01,0.11977472305297851
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,16,8,power_law_1.2,0.032951040267944334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,16,8,power_law_1.01,0.07585792064666748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,16,8,power_law_1.2,0.052801280021667486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,16,8,power_law_1.2,0.05306496143341064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,16,8,power_law_1.2,0.054042878150939944
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,16,8,power_law_1.2,0.05719295978546143
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,16,8,power_law_1.01,0.0877734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,16,8,power_law_1.2,0.05874176025390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,16,8,power_law_1.2,0.057164797782897944
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,16,8,power_law_1.2,0.06279295921325684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,16,8,power_law_1.2,0.06988671779632569
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,16,8,power_law_1.2,0.07497983932495117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,16,8,power_law_1.2,0.07888383865356445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,16,8,power_law_1.2,0.08649472236633302
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,16,8,power_law_1.2,0.09966848373413087
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,16,8,power_law_1.2,0.11789695739746094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,16,8,power_law_1.2,0.13631232261657716
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,16,8,power_law_1.2,0.17533695220947265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,16,8,power_law_1.2,0.2342924880981445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,16,8,power_law_1.2,0.284847354888916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,16,8,power_law_1.2,0.40040447235107424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,16,8,power_law_1.2,0.5168102264404297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,16,8,power_law_1.2,0.026794240474700925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,16,8,power_law_1.2,0.03100543975830078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,16,8,power_law_1.2,0.7667033386230468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,16,8,power_law_1.2,0.0312550401687622
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,16,8,power_law_1.2,0.03335936069488525
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,16,8,power_law_1.2,0.9172198486328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,16,8,power_law_1.2,0.035411200523376464
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,16,8,power_law_1.2,0.03824127912521362
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,16,8,power_law_1.2,0.044440321922302246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,16,8,power_law_1.2,0.04444799900054931
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,16,8,power_law_1.2,0.04649343967437744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,16,8,power_law_1.2,0.04779520034790039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,16,8,power_law_1.2,0.05064576148986817
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,16,8,power_law_1.2,0.04727551937103271
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,16,8,power_law_1.2,0.04813439846038818
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,16,8,power_law_1.2,0.04908160209655762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,16,8,power_law_1.2,0.05017727851867676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,16,8,power_law_1.2,0.05834239959716797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,16,8,power_law_1.2,0.06696063995361327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,16,8,power_law_1.2,0.0781606388092041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,16,8,power_law_1.2,0.09806464195251466
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,16,8,power_law_1.2,0.11122559547424316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,16,8,power_law_1.2,0.05012991905212403
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,16,8,power_law_1.2,0.15705599784851074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,16,8,power_law_1.2,0.20767871856689454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,16,8,power_law_1.2,0.2763520050048828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,16,8,power_law_1.2,0.38542465209960936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,16,8,power_law_1.2,0.5284070587158203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,16,8,power_law_1.2,0.022639360427856445
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,16,8,power_law_1.2,0.023927040100097656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,16,8,power_law_1.2,0.02537600040435791
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,16,8,power_law_1.2,0.026617600917816165
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,16,8,power_law_1.2,0.026414079666137697
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,16,8,power_law_1.2,0.025991680622100832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,16,8,power_law_1.2,0.02592639923095703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,16,8,power_law_1.2,0.02433151960372925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,16,8,power_law_1.2,0.7250956726074219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,16,8,power_law_1.2,0.026681599617004392
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,16,8,power_law_1.2,0.028856320381164553
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,16,8,power_law_1.2,0.02634880065917969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,16,8,power_law_1.2,0.02727168083190918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,16,8,power_law_1.2,0.028188159465789796
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,16,8,power_law_1.2,0.02879487991333008
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,16,8,power_law_1.2,0.030385921001434325
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,16,8,power_law_1.2,0.03860352039337158
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,16,8,power_law_1.01,0.14065152168273926
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,16,8,power_law_1.01,0.21340415954589842
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,16,8,power_law_1.01,0.2728908729553223
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,16,8,power_law_1.2,0.0337932801246643
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,16,8,power_law_1.2,0.054398717880249026
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,16,8,power_law_1.2,0.048277759552001955
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,16,8,power_law_1.2,0.07193855762481689
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,16,8,power_law_1.2,0.09895551681518555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,32,4,balanced,0.021108479499816896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,32,4,balanced,0.022728960514068603
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,32,4,balanced,0.024785919189453127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,32,4,balanced,0.028887040615081787
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,16,8,power_law_1.2,0.10943103790283204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,32,4,balanced,0.03339263916015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,32,4,balanced,0.03703423976898194
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,32,4,balanced,0.03753983974456787
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,32,4,balanced,0.05846784114837646
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,32,4,balanced,0.06040959835052491
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,32,4,balanced,0.07113344192504882
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,32,4,balanced,0.07039487838745118
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,32,4,balanced,0.07137792110443116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,32,4,balanced,0.03632256031036377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,32,4,balanced,0.08220671653747559
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,32,4,balanced,0.08172800064086913
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,32,4,balanced,0.09669504165649415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,32,4,balanced,0.11192192077636717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,32,4,balanced,0.12735872268676757
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,32,4,balanced,0.1578700828552246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,32,4,balanced,0.18686464309692383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,32,4,balanced,0.08700672149658203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,32,4,balanced,0.2450547218322754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,32,4,balanced,0.30277759552001954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,32,4,balanced,0.4207628631591797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,32,4,balanced,0.5367091369628907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,32,4,balanced,0.7642265319824219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,32,4,balanced,0.025582079887390134
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,32,4,balanced,0.025615360736846927
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,32,4,balanced,1.0129241943359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,32,4,balanced,0.02887808084487915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,32,4,balanced,0.03539583921432495
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,32,4,balanced,0.03915136098861695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,32,4,balanced,0.05305471897125245
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,32,4,balanced,0.04815616130828858
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,32,4,balanced,0.04803328037261963
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,32,4,balanced,0.058382081985473636
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,32,4,balanced,0.058798079490661626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,32,4,balanced,0.0591923189163208
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,32,4,balanced,0.060483841896057135
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,32,4,balanced,0.0613644790649414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,32,4,balanced,0.06288256168365478
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,32,4,balanced,0.06462592124938965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,32,4,balanced,0.0682150411605835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,32,4,balanced,0.06614655971527099
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,32,4,balanced,0.08124799728393554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,32,4,balanced,0.0913715171813965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,32,4,balanced,0.11040896415710448
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,32,4,balanced,0.12886528015136717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,32,4,balanced,0.1673151969909668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,32,4,balanced,0.20484352111816406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,32,4,balanced,0.29973888397216797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,32,4,balanced,0.3746611022949219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,32,4,balanced,0.5536640167236329
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,32,4,power_law_1.01,0.02197119951248169
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,32,4,balanced,0.7201561737060547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,32,4,power_law_1.01,0.024976639747619628
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,32,4,power_law_1.01,0.027653119564056396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,32,4,power_law_1.01,0.029441280364990236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,32,4,power_law_1.01,0.030909440517425536
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,32,4,power_law_1.01,0.032821760177612305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,16,8,power_law_1.2,0.15379584312438965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,32,4,power_law_1.01,0.0370688009262085
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,16,8,power_law_1.2,0.20567552566528322
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,32,4,power_law_1.01,0.059745278358459476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,32,4,power_law_1.01,0.06305408000946044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,32,4,power_law_1.01,0.0658355188369751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,32,4,power_law_1.01,0.06477568149566651
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,32,4,power_law_1.01,0.07353343963623046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,32,4,power_law_1.01,0.07932032108306884
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,32,4,power_law_1.01,0.08292991638183593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,32,4,power_law_1.01,0.0349452805519104
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,32,4,power_law_1.01,0.08774911880493164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,32,4,power_law_1.01,0.09590656280517577
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,32,4,power_law_1.01,0.1254220771789551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,32,4,power_law_1.01,0.15308799743652343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,32,4,power_law_1.01,0.18220415115356445
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,16,8,power_law_1.2,0.3170252799987793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,32,4,power_law_1.01,0.24042495727539062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,32,4,power_law_1.01,0.3010393524169922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,32,4,power_law_1.01,0.41677185058593746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,32,4,power_law_1.01,0.539257583618164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,32,4,power_law_1.01,0.025966079235076905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,32,4,power_law_1.01,0.032572159767150874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,32,4,power_law_1.01,0.030324480533599853
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,32,4,power_law_1.01,0.7695820617675782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,32,4,power_law_1.01,0.03873280048370361
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,32,4,power_law_1.01,0.04073088169097901
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,32,4,power_law_1.01,0.04234752178192139
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,32,4,power_law_1.01,0.04495999813079834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,32,4,power_law_1.01,0.0452569580078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,32,4,power_law_1.01,0.0542848014831543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,32,4,power_law_1.01,0.05591296195983887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,32,4,power_law_1.01,0.05629695892333984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,32,4,power_law_1.01,0.10945664405822755
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,32,4,power_law_1.01,0.0591974401473999
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,32,4,power_law_1.01,0.06170623779296876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,32,4,power_law_1.01,0.060673279762268065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,32,4,power_law_1.01,0.06454400062561036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,32,4,power_law_1.01,0.06815999984741211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,32,4,power_law_1.01,0.07671807765960693
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,32,4,power_law_1.01,1.0132428741455077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,32,4,power_law_1.01,0.09115776062011718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,32,4,power_law_1.01,0.11142271995544432
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,32,4,power_law_1.01,0.13729280471801758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,32,4,power_law_1.01,0.17727359771728515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,32,4,power_law_1.01,0.22407936096191405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,32,4,power_law_1.01,0.056417279243469245
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,32,4,power_law_1.01,0.30792831420898437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,32,4,power_law_1.01,0.41007102966308595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,32,4,power_law_1.2,0.023572480678558348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,32,4,power_law_1.2,0.02374783992767334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,32,4,power_law_1.01,0.5782617568969727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,32,4,power_law_1.2,0.025180160999298096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,32,4,power_law_1.2,0.02801408052444458
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,32,4,power_law_1.2,0.03092992067337036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,32,4,power_law_1.2,0.035600640773773194
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,32,4,power_law_1.2,0.03296128034591675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,32,4,power_law_1.2,0.03722111940383911
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,32,4,power_law_1.2,0.059109120368957525
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,32,4,power_law_1.2,0.06499199867248535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,32,4,power_law_1.2,0.05797120094299316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,32,4,power_law_1.2,0.07344128131866455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,32,4,power_law_1.2,0.07193471908569335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,32,4,power_law_1.2,0.08006015777587891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,32,4,power_law_1.2,0.08359295845031738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,32,4,power_law_1.2,0.08857088088989258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,32,4,power_law_1.2,0.0964748764038086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,32,4,power_law_1.2,0.11061375617980958
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,32,4,power_law_1.2,0.12820096015930177
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,32,4,power_law_1.2,0.15453439712524414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,32,4,power_law_1.2,0.1903027153015137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,32,4,power_law_1.2,0.2482329559326172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,32,4,power_law_1.01,0.7609024047851562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,32,4,power_law_1.2,0.2988467216491699
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,32,4,power_law_1.2,0.42789249420166015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,32,4,power_law_1.2,0.026023681163787844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,32,4,power_law_1.2,0.5474060821533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,32,4,power_law_1.2,0.030046720504760743
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,32,4,power_law_1.2,0.03133311986923218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,32,4,power_law_1.2,0.037047040462493894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,32,4,power_law_1.2,0.7807078552246094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,32,4,power_law_1.2,0.04213247776031494
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,32,4,power_law_1.2,0.04505856037139892
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,16,8,power_law_1.2,0.46937473297119137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,32,4,power_law_1.2,0.05378943920135498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,32,4,power_law_1.2,0.054666237831115724
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,32,4,power_law_1.2,0.0568064022064209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,32,4,power_law_1.2,0.055379199981689456
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,32,4,power_law_1.2,0.06002560138702393
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,32,4,power_law_1.2,0.060803837776184075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,32,4,power_law_1.2,0.06277376174926758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,32,4,power_law_1.2,0.06493567943572998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,32,4,power_law_1.2,0.06951168060302734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,32,4,power_law_1.2,0.040261120796203614
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,32,4,power_law_1.2,0.08073344230651855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,32,4,power_law_1.2,0.09304960250854492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,32,4,power_law_1.2,1.0649407958984374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,32,4,power_law_1.2,0.04479231834411621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,32,4,power_law_1.2,0.13853568077087403
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,32,4,power_law_1.2,0.18523647308349608
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,32,4,power_law_1.2,0.23612159729003906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,32,4,power_law_1.2,0.3295948791503906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,32,4,power_law_1.2,0.42143104553222654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,32,4,power_law_1.2,0.6003763198852539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,32,4,power_law_1.2,0.1154355239868164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,128,balanced,0.02802175998687744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,128,balanced,0.024309759140014646
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,128,balanced,0.02603775978088379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,128,balanced,0.02807552099227905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,32,4,power_law_1.2,0.802862091064453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,128,balanced,0.027233281135559083
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,128,balanced,0.042342400550842284
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,128,balanced,0.045990400314331055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,128,balanced,0.04686592102050781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,128,balanced,0.047196159362792964
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,128,balanced,0.04778880119323731
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,128,balanced,0.04663296222686768
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,128,balanced,0.04842239856719971
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,128,balanced,0.05221888065338135
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,128,balanced,0.051635198593139656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,128,balanced,0.044510722160339355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,128,balanced,0.04731391906738282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,128,balanced,0.053415679931640626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,128,balanced,0.06040575981140137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,128,balanced,0.056892161369323735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,128,balanced,0.06619135856628418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,128,balanced,0.07179008007049562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,128,balanced,0.06687232017517089
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,128,balanced,0.08603391647338868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,128,balanced,0.10766592025756835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,128,balanced,0.12580351829528807
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,128,balanced,0.043948798179626464
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,128,balanced,0.044860157966613766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,1,128,balanced,0.017008639574050903
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,1,128,balanced,0.016947200298309324
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,1,128,balanced,0.017020159959793092
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,1,128,balanced,0.017328640222549437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,1,128,balanced,0.017031680345535278
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,1,128,balanced,0.017045760154724122
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,1,128,balanced,0.01698815941810608
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,1,128,balanced,0.017318400144577025
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,1,128,balanced,0.01701632022857666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,1,128,balanced,0.016974079608917236
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,1,128,balanced,0.01701248049736023
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,1,128,balanced,0.017308160066604614
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,1,128,balanced,0.017460479736328124
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,1,128,balanced,0.016948479413986205
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,1,128,balanced,0.016980479955673217
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,1,128,balanced,0.01738368034362793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,1,128,balanced,0.017419519424438475
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,1,128,balanced,0.017422080039978027
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,1,128,balanced,0.01692800045013428
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,1,128,balanced,0.016760319471359253
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,1,128,balanced,0.016179200410842896
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,1,128,balanced,0.018243839740753172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,128,power_law_1.01,0.02309760093688965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,128,power_law_1.01,0.023093760013580322
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,1,128,balanced,0.019063040018081664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,1,128,balanced,0.017850879430770874
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,1,128,balanced,0.01903615951538086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,128,power_law_1.01,0.04526847839355469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,1,128,balanced,0.021173760890960694
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,1,128,balanced,0.025218560695648196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,128,power_law_1.01,0.04690432071685791
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,128,power_law_1.01,0.048636159896850585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,128,power_law_1.01,0.05051008224487304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,128,power_law_1.01,0.05206528186798096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,128,power_law_1.01,0.05341184139251709
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,128,power_law_1.01,0.05850368022918702
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,128,power_law_1.01,0.02563456058502197
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,128,power_law_1.01,0.06568448066711426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,128,power_law_1.01,0.055669760704040526
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,128,power_law_1.01,0.08309887886047364
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,128,power_law_1.01,0.10395392417907715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,128,power_law_1.01,0.06139135837554932
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,128,power_law_1.01,0.11476863861083983
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,128,power_law_1.01,0.1745919990539551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,128,power_law_1.01,0.3024857521057129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,128,power_law_1.01,0.4399705505371093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,128,power_law_1.01,0.04733312129974365
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,128,power_law_1.01,0.044803838729858395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,1,128,power_law_1.01,0.016992000341415407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,128,power_law_1.01,0.042997760772705076
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,128,power_law_1.01,0.6279449462890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,1,128,power_law_1.01,0.01691264033317566
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,1,128,power_law_1.01,0.01698304057121277
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,1,128,power_law_1.01,0.016620800495147706
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,1,128,power_law_1.01,0.01693184018135071
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,1,128,power_law_1.01,0.017326079607009888
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,1,128,power_law_1.01,0.0169651198387146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,128,power_law_1.01,0.795068130493164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,1,128,power_law_1.01,0.016921600103378294
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,1,128,power_law_1.01,0.017392640113830564
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,1,128,power_law_1.01,0.017356799840927122
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,1,128,power_law_1.01,0.01734272003173828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,128,power_law_1.01,0.049387521743774414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,1,128,power_law_1.01,0.017295360565185547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,1,128,power_law_1.01,0.01736191987991333
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,1,128,power_law_1.01,0.017419519424438475
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,1,128,power_law_1.01,0.016993279457092288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,128,power_law_1.01,0.06871679782867432
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,1,128,power_law_1.01,0.016977920532226562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,128,power_law_1.01,0.20794240951538087
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,1,128,power_law_1.01,0.017822719812393188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,1,128,power_law_1.01,0.018257919549942016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,1,128,power_law_1.01,0.01901952028274536
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,1,128,power_law_1.01,0.020599040985107422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,128,power_law_1.2,0.024218881130218507
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,128,power_law_1.2,0.024387838840484618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,128,power_law_1.2,0.04037375926971436
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,1,128,power_law_1.01,0.022917120456695556
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,1,128,power_law_1.01,0.030725119113922117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,128,power_law_1.2,0.04648191928863525
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,128,power_law_1.2,0.0448524808883667
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,128,power_law_1.2,0.04732160091400146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,128,power_law_1.2,0.0489408016204834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,128,power_law_1.2,0.050762238502502444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,128,power_law_1.2,0.05213183879852294
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,128,power_law_1.2,0.05262335777282715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,128,power_law_1.2,0.05327744007110595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,128,power_law_1.2,0.05500544071197509
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,128,power_law_1.2,0.05977600097656251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,128,power_law_1.2,0.06352511882781983
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,128,power_law_1.2,0.06530432224273681
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,128,power_law_1.2,0.0706496000289917
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,128,power_law_1.2,0.08596223831176758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,128,power_law_1.2,0.09961088180541992
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,1,128,power_law_1.01,0.033653759956359865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,128,power_law_1.2,0.1226905632019043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,128,power_law_1.2,0.1713484764099121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,128,power_law_1.2,0.23636863708496092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,128,power_law_1.2,0.3765542221069336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,128,power_law_1.2,0.4796672058105469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,128,power_law_1.2,0.6680780792236328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,1,128,power_law_1.2,0.01701120018959045
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,1,128,power_law_1.2,0.016972800493240355
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,1,128,power_law_1.2,0.016956160068511965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,128,power_law_1.2,0.04324351787567139
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,1,128,power_law_1.2,0.0168230402469635
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,1,128,power_law_1.2,0.01696895956993103
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,1,128,power_law_1.2,0.016954879760742187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,1,128,power_law_1.2,0.01690623998641968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,1,128,power_law_1.2,0.017067519426345826
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,1,128,power_law_1.2,0.01699839949607849
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,1,128,power_law_1.2,0.017017600536346437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,1,128,power_law_1.2,0.016939519643783568
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,1,128,power_law_1.2,0.017375999689102174
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,1,128,power_law_1.2,0.017396479845046997
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,1,128,power_law_1.2,0.017375999689102174
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,1,128,power_law_1.2,0.017337599992752077
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,1,128,power_law_1.2,0.01777791976928711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,128,power_law_1.2,1.2464473724365235
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,1,128,power_law_1.01,0.04528639793395996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,128,power_law_1.2,1.6371839904785155
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,1,128,power_law_1.01,0.059420161247253425
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,1,128,power_law_1.01,0.08616703987121582
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,1,128,power_law_1.01,0.10307583808898926
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,1,128,power_law_1.2,0.018323199748992918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,1,128,power_law_1.2,0.019033600091934205
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,1,128,power_law_1.2,0.020724480152130124
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,1,128,power_law_1.2,0.02512768030166626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,1,128,power_law_1.2,0.031153919696807864
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,64,balanced,0.02314879894256592
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,64,balanced,0.02310143947601318
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,64,balanced,0.023165440559387206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,64,balanced,0.0361075210571289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,64,balanced,0.037482240200042725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,64,balanced,0.03829119920730591
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,64,balanced,0.03952896118164063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,64,balanced,0.021848320960998535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,64,balanced,0.039930880069732666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,64,balanced,0.040364799499511717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,64,balanced,0.04162943840026855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,64,balanced,0.04280831813812256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,64,balanced,0.04322815895080566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,64,balanced,0.04487167835235596
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,64,balanced,0.04359424114227295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,64,balanced,0.04477695941925049
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,64,balanced,0.039064319133758546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,64,balanced,0.04681471824645996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,64,balanced,0.05019008159637452
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,64,balanced,0.056907520294189454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,64,balanced,0.05425663948059082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,64,balanced,0.05692416191101074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,64,balanced,0.07047552108764649
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,64,balanced,0.07738239765167236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,64,balanced,0.09177087783813477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,64,balanced,0.11741056442260742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,64,balanced,0.1395353603363037
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,1,128,power_law_1.2,0.04306431770324707
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,1,128,power_law_1.2,0.05642496109008789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,1,128,power_law_1.2,0.07957759857177735
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,1,128,power_law_1.2,0.10077695846557617
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,1,128,power_law_1.2,0.1440998363494873
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,1,128,power_law_1.2,0.19716224670410157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,2,64,balanced,0.01701632022857666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,2,64,balanced,0.01705471992492676
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,2,64,balanced,0.017390079498291015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,2,64,balanced,0.017437440156936646
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,2,64,balanced,0.017418240308761594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,2,64,balanced,0.01740288019180298
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,2,64,balanced,0.01740288019180298
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,2,64,balanced,0.017381119728088378
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,2,64,balanced,0.016573439836502075
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,2,64,balanced,0.017447680234909058
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,2,64,balanced,0.01734272003173828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,2,64,balanced,0.01740031957626343
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,2,64,balanced,0.017073919773101805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,2,64,balanced,0.017731839418411256
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,2,64,balanced,0.01782912015914917
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,2,64,balanced,0.01778432011604309
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,2,64,balanced,0.017786879539489747
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,2,64,balanced,0.01734784007072449
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,2,64,balanced,0.01741312026977539
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,2,64,balanced,0.017396479845046997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,64,power_law_1.01,0.023589119911193848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,64,power_law_1.01,0.023104000091552734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,2,64,balanced,0.017390079498291015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,64,power_law_1.01,0.023521280288696288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,64,power_law_1.01,0.03665791988372803
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,64,power_law_1.01,0.037620480060577395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,64,power_law_1.01,0.03910144090652466
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,64,power_law_1.01,0.03950592041015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,64,power_law_1.01,0.041184000968933106
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,64,power_law_1.01,0.04079872131347656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,64,power_law_1.01,0.0432038402557373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,64,power_law_1.01,0.043575038909912114
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,64,power_law_1.01,0.04489727973937988
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,64,power_law_1.01,0.045758719444274905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,64,power_law_1.01,0.04769536018371582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,64,power_law_1.01,0.051046400070190436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,64,power_law_1.01,0.054991359710693355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,64,power_law_1.01,0.055614719390869136
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,64,power_law_1.01,0.0559116792678833
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,64,power_law_1.01,0.06603263854980469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,64,power_law_1.01,0.08536447525024414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,64,power_law_1.01,0.10261631965637208
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,64,power_law_1.01,0.14617856025695802
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,64,power_law_1.01,0.165614070892334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,64,power_law_1.01,0.23131776809692384
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,64,power_law_1.01,0.3371084976196289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,64,power_law_1.01,0.5034329605102539
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,2,64,power_law_1.01,0.01741055965423584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,64,power_law_1.01,0.7194303894042969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,2,64,power_law_1.01,0.017341439723968503
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,2,64,power_law_1.01,0.017395199537277223
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,2,64,power_law_1.01,0.01741696000099182
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,2,64,power_law_1.01,0.0170905601978302
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,2,64,power_law_1.01,0.01737215995788574
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,2,64,power_law_1.01,0.01738368034362793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,2,64,power_law_1.01,0.01740928053855896
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,2,64,power_law_1.01,0.017375999689102174
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,2,64,power_law_1.01,0.017487360239028932
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,2,64,power_law_1.01,0.01739392042160034
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,2,64,power_law_1.01,0.017382400035858152
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,2,64,power_law_1.01,0.017789440155029295
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,2,64,power_law_1.01,0.017794560194015503
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,2,64,power_law_1.01,0.0177292799949646
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,2,64,power_law_1.01,0.01745151996612549
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,2,64,balanced,0.018640639781951903
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,2,64,balanced,0.019086079597473143
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,2,64,balanced,0.020665600299835205
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,2,64,balanced,0.020198400020599368
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,2,64,balanced,0.02676352024078369
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,2,64,balanced,0.028485119342803955
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,2,64,power_law_1.01,0.018561279773712157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,2,64,power_law_1.01,0.019059200286865235
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,2,64,power_law_1.01,0.019842560291290286
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,2,64,power_law_1.01,0.021817600727081297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,2,64,power_law_1.01,0.024231679439544677
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,64,power_law_1.2,0.023927040100097656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,64,power_law_1.2,0.022690560817718506
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,64,power_law_1.2,0.023121919631958008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,64,power_law_1.2,0.03665791988372803
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,64,power_law_1.2,0.03868288040161133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,64,power_law_1.2,0.03898751974105835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,64,power_law_1.2,0.0400051212310791
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,2,64,power_law_1.01,0.028849918842315674
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,2,64,power_law_1.01,0.035174400806427
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,64,power_law_1.2,0.04037375926971436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,64,power_law_1.2,0.04119552135467529
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,64,power_law_1.2,0.04500607967376709
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,64,power_law_1.2,0.04578944206237793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,64,power_law_1.2,0.047362561225891116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,64,power_law_1.2,0.05051519870758057
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,64,power_law_1.2,0.05377151966094971
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,64,power_law_1.2,0.05641344070434571
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,64,power_law_1.2,0.05616000175476075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,64,power_law_1.2,0.07358848094940186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,64,power_law_1.2,0.08191360473632812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,64,power_law_1.2,0.09412223815917968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,64,power_law_1.2,0.12569600105285644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,64,power_law_1.2,0.19133312225341798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,64,power_law_1.2,0.2759347152709961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,64,power_law_1.2,0.04360320091247559
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,64,power_law_1.2,0.31813631057739256
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,2,64,power_law_1.2,0.017056000232696534
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,64,power_law_1.2,0.44575870513916016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,2,64,power_law_1.2,0.017379839420318604
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,2,64,power_law_1.2,0.01736832022666931
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,64,power_law_1.2,0.8155161285400391
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,2,64,power_law_1.2,0.01739904046058655
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,2,64,power_law_1.2,0.017295360565185547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,2,64,power_law_1.2,0.017375999689102174
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,2,64,power_law_1.2,0.017375999689102174
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,2,64,power_law_1.2,0.017415679693222046
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,2,64,power_law_1.2,0.01733504056930542
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,2,64,power_law_1.2,0.017446399927139283
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,2,64,power_law_1.2,0.017378560304641723
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,2,64,power_law_1.2,0.017791999578475954
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,2,64,power_law_1.2,0.017724159955978393
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,2,64,power_law_1.2,0.01778432011604309
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,2,64,power_law_1.2,0.017825280427932737
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,2,64,power_law_1.2,0.01820672035217285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,64,power_law_1.2,1.1097100830078124
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,2,64,power_law_1.01,0.04337664127349854
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,2,64,power_law_1.01,0.06066688060760498
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,2,64,power_law_1.01,0.10707200050354004
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,2,64,power_law_1.01,0.0806873607635498
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,2,64,power_law_1.2,0.01862912058830261
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,2,64,power_law_1.2,0.020430080890655518
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,2,64,power_law_1.2,0.02025599956512451
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,2,64,power_law_1.2,0.022405118942260743
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,2,64,power_law_1.2,0.03043839931488037
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,4,32,balanced,0.019882240295410157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,4,32,balanced,0.020206079483032227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,4,32,balanced,0.02140928030014038
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,4,32,balanced,0.01988736033439636
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,4,32,balanced,0.03445247888565063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,4,32,balanced,0.03586944103240967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,4,32,balanced,0.03657344102859497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,4,32,balanced,0.034982399940490724
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,4,32,balanced,0.03777535915374756
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,4,32,balanced,0.04001279830932617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,4,32,balanced,0.04113024234771728
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,4,32,balanced,0.04159999847412109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,4,32,balanced,0.0416755199432373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,4,32,balanced,0.04312448024749756
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,4,32,balanced,0.0446131181716919
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,4,32,balanced,0.04607744216918945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,4,32,balanced,0.036180479526519774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,4,32,balanced,0.05115776062011719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,4,32,balanced,0.04895232200622558
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,4,32,balanced,0.051837439537048335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,4,32,balanced,0.05510015964508057
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,4,32,balanced,0.06574719905853271
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,4,32,balanced,0.07957888126373291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,4,32,balanced,0.10488448143005372
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,4,32,balanced,0.12435327529907228
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,4,32,balanced,0.16120447158813475
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,2,64,power_law_1.2,0.04108799934387207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,4,32,balanced,0.2045542335510254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,4,32,power_law_1.01,0.021422080993652344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,4,32,power_law_1.01,0.02152064085006714
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,2,64,power_law_1.2,0.04411007881164551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,4,32,power_law_1.01,0.022369279861450195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,4,32,power_law_1.01,0.02225152015686035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,4,32,power_law_1.01,0.03544575929641723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,4,32,power_law_1.01,0.036803839206695554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,4,32,power_law_1.01,0.03675519943237305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,4,32,power_law_1.01,0.038565120697021484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,4,32,power_law_1.01,0.04031360149383545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,4,32,power_law_1.01,0.042237439155578614
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,4,32,power_law_1.01,0.04424191951751709
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,4,32,power_law_1.01,0.04448256015777588
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,4,32,power_law_1.01,0.0470143985748291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,4,32,power_law_1.01,0.05057024002075196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,4,32,power_law_1.01,0.05925119876861572
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,4,32,power_law_1.01,0.051006717681884764
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,4,32,power_law_1.01,0.06139264106750488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,4,32,power_law_1.01,0.06632192134857177
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,4,32,power_law_1.01,0.08530048370361329
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,4,32,power_law_1.01,0.11002880096435548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,4,32,power_law_1.01,0.13559167861938476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,4,32,power_law_1.01,0.17067007064819334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,4,32,power_law_1.01,0.2738688087463379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,4,32,power_law_1.01,0.04271232128143311
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,4,32,power_law_1.01,0.3400089645385742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,4,32,power_law_1.2,0.021048319339752198
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,4,32,power_law_1.01,0.4695564651489258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,4,32,power_law_1.2,0.02189055919647217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,4,32,power_law_1.2,0.022749440670013426
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,2,64,power_law_1.2,0.0653056001663208
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,4,32,power_law_1.2,0.022480640411376953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,4,32,power_law_1.2,0.03615744113922119
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,4,32,power_law_1.01,0.6288921737670898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,4,32,power_law_1.2,0.035783679485321046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,4,32,power_law_1.2,0.04035967826843262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,4,32,power_law_1.2,0.03752448081970215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,4,32,power_law_1.2,0.04143360137939453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,4,32,power_law_1.2,0.04214144229888916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,4,32,power_law_1.2,0.04279679775238037
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,4,32,power_law_1.2,0.04531455993652343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,4,32,power_law_1.2,0.0476697587966919
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,4,32,power_law_1.2,0.05623040199279785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,4,32,power_law_1.2,0.05218048095703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,4,32,power_law_1.2,0.052883200645446785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,4,32,power_law_1.2,0.07894271850585938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,4,32,power_law_1.2,0.09789440155029297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,4,32,power_law_1.2,0.11673600196838378
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,4,32,power_law_1.2,0.16458751678466796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,4,32,power_law_1.2,0.1955891227722168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,4,32,power_law_1.2,0.30050432205200195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,4,32,power_law_1.2,0.05001599788665771
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,2,64,power_law_1.2,0.08486656188964845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,4,32,power_law_1.2,0.06765567779541015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,8,16,balanced,0.022242560386657714
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,8,16,balanced,0.023120639324188234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,8,16,balanced,0.02143872022628784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,8,16,balanced,0.027242240905761717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,8,16,balanced,0.04315904140472412
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,8,16,balanced,0.04408063888549805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,8,16,balanced,0.044094719886779786
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,8,16,balanced,0.04442880153656006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,8,16,balanced,0.04695551872253418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,8,16,balanced,0.047957758903503414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,4,32,power_law_1.2,0.43557502746582033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,8,16,balanced,0.050113282203674316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,8,16,balanced,0.051360001564025884
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,8,16,balanced,0.054897918701171874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,8,16,balanced,0.056733441352844236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,8,16,balanced,0.05260672092437744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,8,16,balanced,0.05426559925079346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,8,16,balanced,0.0582041597366333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,8,16,balanced,0.07867136001586914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,8,16,balanced,0.08948224067687989
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,8,16,balanced,0.1209984016418457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,8,16,balanced,0.024545280933380126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,8,16,balanced,0.13613056182861327
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,2,64,power_law_1.2,0.1227302360534668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,8,16,balanced,0.18052480697631837
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,2,64,power_law_1.2,0.1730508804321289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,8,16,balanced,0.21891712188720702
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,4,32,power_law_1.2,0.6694143676757813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,8,16,power_law_1.01,0.024778239727020264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,8,16,balanced,0.3023616027832031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,8,16,power_law_1.01,0.028460800647735596
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,8,16,power_law_1.01,0.028860158920288086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,8,16,balanced,0.38737407684326175
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,8,16,power_law_1.01,0.04322815895080566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,8,16,power_law_1.01,0.041418237686157225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,8,16,power_law_1.01,0.04389503955841064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,8,16,power_law_1.01,0.045907201766967776
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,8,16,power_law_1.01,0.04723968029022217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,8,16,power_law_1.01,0.05009791851043701
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,8,16,power_law_1.01,0.051466240882873535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,8,16,power_law_1.01,0.05383039951324463
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,8,16,power_law_1.01,0.057145600318908696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,8,16,power_law_1.01,0.05925504207611084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,8,16,power_law_1.01,0.05686528205871581
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,8,16,power_law_1.01,0.06070015907287598
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,8,16,power_law_1.01,0.08102272033691407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,8,16,power_law_1.01,0.08826239585876465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,8,16,balanced,0.04906496047973633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,8,16,power_law_1.01,0.10379008293151855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,8,16,power_law_1.01,0.12819456100463866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,8,16,power_law_1.01,0.18374015808105468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,8,16,power_law_1.01,0.19468288421630858
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,8,16,power_law_1.01,0.30917247772216794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,8,16,power_law_1.01,0.3691046524047851
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,8,16,power_law_1.2,0.02477695941925049
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,8,16,power_law_1.01,0.04770944118499756
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,8,16,power_law_1.2,0.02723072052001953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,8,16,power_law_1.2,0.029614078998565673
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,8,16,power_law_1.2,0.042145280838012694
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,8,16,power_law_1.01,0.6686592102050781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,8,16,power_law_1.2,0.04640255928039551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,8,16,power_law_1.2,0.045241599082946775
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,8,16,power_law_1.2,0.04728576183319092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,8,16,power_law_1.01,0.026912000179290775
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,8,16,power_law_1.2,0.05051263809204102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,8,16,power_law_1.2,0.05177343845367431
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,8,16,power_law_1.2,0.056858878135681155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,8,16,power_law_1.2,0.05696256160736084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,8,16,power_law_1.2,0.058132481575012204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,8,16,power_law_1.2,0.06459136009216308
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,8,16,power_law_1.01,0.575175666809082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,8,16,power_law_1.2,0.0822719955444336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,8,16,power_law_1.2,0.09095423698425294
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,8,16,power_law_1.2,0.04402815818786621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,8,16,power_law_1.2,0.11170816421508789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,8,16,power_law_1.2,0.14698752403259277
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,8,16,power_law_1.2,0.04907904148101806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,8,16,power_law_1.2,0.1916147232055664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,8,16,power_law_1.2,0.2136895942687988
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,8,16,power_law_1.2,0.3550041580200195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,16,8,balanced,0.022731521129608155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,4,32,power_law_1.2,0.9246387481689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,16,8,balanced,0.02313215970993042
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,8,16,power_law_1.2,0.4323148727416992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,16,8,balanced,0.023168001174926758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,16,8,balanced,0.031000320911407468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,16,8,balanced,0.05076863765716553
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,16,8,balanced,0.04805759906768799
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,16,8,balanced,0.05179776191711426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,16,8,balanced,0.052255997657775874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,16,8,balanced,0.05386367797851562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,16,8,balanced,0.05877888202667236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,16,8,balanced,0.06041600227355957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,8,16,power_law_1.2,0.7265843200683594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,16,8,balanced,0.05506944179534912
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,16,8,balanced,0.058535680770874024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,16,8,balanced,0.0555072021484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,16,8,balanced,0.05873536109924317
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,16,8,balanced,0.0706547212600708
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,16,8,balanced,0.08130816459655762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,8,16,power_law_1.2,0.027619841098785403
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,16,8,balanced,0.11420160293579103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,16,8,balanced,0.12775168418884278
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,16,8,balanced,0.053415679931640626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,16,8,balanced,0.1677516746520996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,16,8,balanced,0.19648639678955077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,8,16,power_law_1.2,0.05382143974304199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,16,8,balanced,0.27385343551635744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,16,8,balanced,0.3431500625610352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,16,8,power_law_1.01,0.028006401062011722
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,16,8,balanced,0.5160780715942382
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,16,8,power_law_1.01,0.02720383882522583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,16,8,balanced,0.659805450439453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,16,8,power_law_1.01,0.03131135940551758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,16,8,power_law_1.01,0.04574079990386963
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,16,8,power_law_1.01,0.049262080192565914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,16,8,power_law_1.01,0.04893055915832519
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,16,8,power_law_1.01,0.05074304103851318
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,8,16,power_law_1.2,0.9649574279785156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,16,8,power_law_1.01,0.05427840232849122
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,16,8,power_law_1.01,0.05309951782226563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,16,8,power_law_1.01,0.05666560173034668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,16,8,power_law_1.01,0.05984127998352051
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,16,8,power_law_1.01,0.05870848178863526
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,16,8,power_law_1.01,0.05745279788970947
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,16,8,power_law_1.01,0.057863678932189945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,16,8,power_law_1.01,0.06662528038024902
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,16,8,power_law_1.01,0.0843609619140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,16,8,power_law_1.01,0.09726207733154298
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,16,8,power_law_1.01,0.11847040176391602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,16,8,power_law_1.01,0.1350540828704834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,16,8,power_law_1.01,0.1779840087890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,16,8,power_law_1.01,0.23857664108276366
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,16,8,power_law_1.01,0.32917247772216796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,16,8,power_law_1.2,0.02601855993270874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,16,8,power_law_1.01,0.4312601470947266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,16,8,power_law_1.2,0.02642431974411011
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,16,8,power_law_1.01,0.03171072006225586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,16,8,power_law_1.2,0.027553279399871827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,16,8,power_law_1.2,0.029304320812225344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,16,8,power_law_1.01,0.6416998291015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,16,8,power_law_1.2,0.04565887928009033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,16,8,power_law_1.2,0.03177088022232056
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,16,8,power_law_1.2,0.04948480129241943
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,16,8,power_law_1.01,0.8370047760009764
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,16,8,power_law_1.2,0.05131392002105713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,16,8,power_law_1.2,0.05275263786315918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,16,8,power_law_1.2,0.05500415802001953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,16,8,power_law_1.2,0.057607679367065436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,16,8,power_law_1.2,0.05901311874389649
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,16,8,power_law_1.2,0.05731967926025391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,16,8,power_law_1.2,0.05875072002410888
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,16,8,power_law_1.2,0.057816319465637214
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,16,8,power_law_1.2,0.06947711944580079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,16,8,power_law_1.2,0.08542976379394532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,16,8,power_law_1.2,0.09801600456237794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,16,8,power_law_1.2,0.1223782444000244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,16,8,power_law_1.2,0.13525376319885254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,16,8,power_law_1.2,0.205928955078125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,16,8,power_law_1.2,0.27099775314331054
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,32,4,balanced,0.020899839401245117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,16,8,power_law_1.2,0.048288002014160156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,32,4,balanced,0.022623999118804933
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,16,8,power_law_1.2,0.5279296112060547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,32,4,balanced,0.025011200904846192
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,32,4,balanced,0.029277439117431643
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,16,8,balanced,0.025617918968200683
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,32,4,balanced,0.037131519317626954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,32,4,balanced,0.04316671848297119
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,32,4,balanced,0.05705471992492676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,32,4,balanced,0.05963520050048828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,32,4,balanced,0.0607155179977417
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,32,4,balanced,0.06530176162719727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,32,4,balanced,0.06717440128326416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,16,8,power_law_1.2,0.39676929473876954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,32,4,balanced,0.05990655899047852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,32,4,balanced,0.06077184200286865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,32,4,balanced,0.06451072216033935
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,32,4,balanced,0.0641587209701538
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,32,4,balanced,0.0688422393798828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,32,4,balanced,0.08596223831176758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,32,4,balanced,0.09879296302795411
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,32,4,balanced,0.13532416343688963
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,32,4,balanced,0.15207039833068847
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,16,8,power_law_1.2,0.7232563018798828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,32,4,balanced,0.20456192016601565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,32,4,balanced,0.24419456481933594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,32,4,balanced,0.34265342712402347
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,32,4,power_law_1.01,0.02315648078918457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,32,4,balanced,0.42962558746337887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,32,4,power_law_1.01,0.027676160335540774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,32,4,balanced,0.6445132446289062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,32,4,power_law_1.01,0.031978240013122564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,32,4,power_law_1.01,0.03381632089614868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,16,8,power_law_1.01,0.0235647988319397
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,32,4,power_law_1.01,0.040341758728027345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,32,4,balanced,0.8349107360839844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,32,4,power_law_1.01,0.055255041122436524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,32,4,power_law_1.01,0.059159040451049805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,32,4,power_law_1.01,0.0591487979888916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,32,4,power_law_1.01,0.06224768161773682
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,32,4,power_law_1.01,0.0649228811264038
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,32,4,power_law_1.01,0.0620198392868042
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,32,4,power_law_1.01,0.06166143894195557
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,32,4,power_law_1.01,0.06243584156036377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,32,4,power_law_1.01,0.06473087787628173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,32,4,power_law_1.01,0.0702400016784668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,32,4,power_law_1.01,0.027668480873107908
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,32,4,power_law_1.01,0.09303680419921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,32,4,power_law_1.01,0.10484607696533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,32,4,power_law_1.01,0.12647040367126466
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,32,4,power_law_1.01,0.15599743843078612
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,32,4,power_law_1.01,0.20195840835571288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,32,4,power_law_1.01,0.26222591400146483
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,32,4,power_law_1.2,0.024683520793914795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,32,4,power_law_1.01,0.35140094757080076
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,32,4,power_law_1.01,0.476945915222168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,32,4,power_law_1.2,0.026644480228424073
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,32,4,power_law_1.2,0.027630081176757814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,32,4,power_law_1.2,0.030209279060363768
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,32,4,power_law_1.01,0.6946701049804688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,32,4,power_law_1.2,0.03379839897155762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,32,4,power_law_1.2,0.0544652795791626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,32,4,power_law_1.2,0.055505919456481936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,32,4,power_law_1.01,0.914434585571289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,32,4,power_law_1.2,0.05791999816894531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,32,4,power_law_1.2,0.05880576133728027
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,32,4,power_law_1.2,0.06277120113372803
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,32,4,power_law_1.2,0.0653324794769287
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,32,4,power_law_1.2,0.060761599540710455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,32,4,power_law_1.2,0.0638592004776001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,32,4,power_law_1.2,0.06548480033874512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,32,4,power_law_1.2,0.06677375793457031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,32,4,power_law_1.2,0.07331456184387207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,32,4,power_law_1.01,0.055505919456481936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,32,4,power_law_1.2,0.09110143661499023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,32,4,power_law_1.2,0.10756863594055174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,32,4,power_law_1.2,0.13204352378845213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,32,4,power_law_1.2,0.15571840286254884
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,32,4,power_law_1.2,0.2535500717163086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,32,4,power_law_1.2,0.042224640846252444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,128,balanced,0.03649408102035522
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,32,4,power_law_1.2,0.5069734573364257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,128,balanced,0.036788480281829836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,128,balanced,0.03542144060134887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,32,4,power_law_1.2,0.7103871917724609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,128,balanced,0.07200767993927001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,32,4,balanced,0.05839744091033936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,128,balanced,0.1200704002380371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,128,balanced,0.12240384101867677
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,128,balanced,0.12559359550476074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,128,balanced,0.12981760025024414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,128,balanced,0.1293990421295166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,128,balanced,0.1355519962310791
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,128,balanced,0.15095680236816406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,128,balanced,0.14580608367919923
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,16,8,power_law_1.2,1.0166092681884766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,128,balanced,0.1540595245361328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,128,balanced,0.16750463485717773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,128,balanced,0.18786815643310545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,32,4,power_law_1.2,0.39796478271484376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,128,balanced,0.21016704559326174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,128,balanced,0.22668287277221683
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,128,balanced,0.26152191162109373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,128,balanced,0.13963775634765624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,128,balanced,0.4300057601928711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,128,balanced,0.14799615859985352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,128,balanced,0.5822220611572265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,128,balanced,0.1301081562042236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,128,balanced,0.7229708862304688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,128,balanced,0.034508800506591795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,128,balanced,0.04596992015838623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,128,balanced,1.052294387817383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,128,balanced,0.05634304046630859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,128,balanced,0.05748223781585693
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,128,balanced,0.06038911819458008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,128,balanced,1.4270541381835937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,128,balanced,0.06363264083862305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,128,balanced,0.062186241149902344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,128,balanced,0.03460736036300659
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,128,balanced,0.06333439826965333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,128,balanced,0.06942975997924805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,128,balanced,0.07332863807678222
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,128,balanced,0.07668735980987548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,128,balanced,0.0715776014328003
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,128,balanced,0.06969600200653077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,128,balanced,0.07248767852783203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,128,balanced,0.0941478443145752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,128,balanced,0.08157567977905274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,128,balanced,0.10960000038146973
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,128,balanced,0.361712646484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,128,balanced,0.11666048049926758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,128,balanced,0.1367296028137207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,128,balanced,0.18920064926147462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,128,balanced,0.23054847717285157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,128,balanced,0.3108992004394531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,128,balanced,0.3839475250244141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,32,4,power_law_1.2,0.21034624099731447
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,128,balanced,0.5592371368408203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,128,balanced,0.7160192108154296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,128,balanced,0.03436928033828736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,32,4,power_law_1.2,0.963589096069336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,1,128,balanced,0.019787520170211792
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,1,128,balanced,0.019850239753723145
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,1,128,balanced,0.026010880470275877
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,1,128,balanced,0.019833600521087645
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,1,128,balanced,0.02113408088684082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,1,128,balanced,0.025966079235076905
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,1,128,balanced,0.02633984088897705
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,1,128,balanced,0.026109440326690676
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,1,128,balanced,0.026087679862976075
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,1,128,balanced,0.02634495973587036
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,1,128,balanced,0.026466560363769532
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,1,128,balanced,0.02601471900939941
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,1,128,balanced,0.027644159793853762
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,1,128,balanced,0.02772864103317261
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,1,128,balanced,0.02807039976119995
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,1,128,balanced,0.02925823926925659
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,1,128,balanced,0.02600831985473633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,1,128,balanced,0.026990079879760744
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,1,128,balanced,0.026178560256958007
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,1,128,balanced,0.025811200141906736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,128,power_law_1.01,0.07539072036743164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,1,128,balanced,0.02620032072067261
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,128,power_law_1.01,0.0761894416809082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,128,power_law_1.01,0.11737728118896484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,128,power_law_1.01,0.11989888191223144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,128,power_law_1.01,0.07800320148468018
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,128,power_law_1.01,0.12391679763793946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,128,power_law_1.01,0.13004032135009766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,128,power_law_1.01,0.1418227195739746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,128,power_law_1.01,0.1528012752532959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,128,power_law_1.01,0.14893695831298828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,128,power_law_1.01,0.13147135734558107
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,128,power_law_1.01,0.15126399993896483
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,128,power_law_1.01,0.16252416610717774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,128,power_law_1.01,0.17414911270141603
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,128,power_law_1.01,0.1943155288696289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,128,power_law_1.01,0.2069926452636719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,128,power_law_1.01,0.24029184341430665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,128,power_law_1.01,0.3017292785644531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,128,power_law_1.01,0.3715251159667969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,128,power_law_1.01,0.49286144256591796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,128,power_law_1.01,0.5933516693115235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,128,power_law_1.01,0.8890201568603515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,128,power_law_1.01,1.1246092987060545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,128,power_law_1.01,1.8423423767089844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,128,power_law_1.01,2.421361846923828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,128,power_law_1.01,3.6364376831054686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,128,power_law_1.01,0.050335998535156246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,128,power_law_1.01,0.050677762031555176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,128,power_law_1.01,0.05234560012817383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,128,power_law_1.01,0.054448637962341305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,128,power_law_1.01,0.055854082107543945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,128,power_law_1.01,0.060837121009826654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,128,power_law_1.01,5.142942504882813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,128,power_law_1.01,0.06327807903289795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,128,power_law_1.01,0.06593023777008057
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,128,power_law_1.01,0.0722649621963501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,128,power_law_1.01,0.07788159847259521
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,128,power_law_1.01,0.07507840156555176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,128,power_law_1.01,0.07356671810150146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,128,power_law_1.01,0.07673984050750733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,128,power_law_1.01,0.08860544204711915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,128,power_law_1.01,0.09727999687194824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,128,power_law_1.01,0.11000320434570313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,128,power_law_1.01,0.1255948829650879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,128,power_law_1.01,0.15012351989746092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,128,power_law_1.01,0.20333055496215818
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,128,power_law_1.01,0.24453887939453126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,128,power_law_1.01,0.3207231903076172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,128,power_law_1.01,0.45284095764160154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,128,power_law_1.01,0.5656166458129883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,128,power_law_1.01,0.8763827514648437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,128,power_law_1.01,1.2248461151123047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,128,power_law_1.01,1.756910095214844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,1,128,power_law_1.01,0.02590208053588867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,1,128,balanced,0.040967679023742674
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,1,128,balanced,0.05266047954559326
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,1,128,power_law_1.01,0.025946879386901857
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,1,128,balanced,0.04406911849975586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,1,128,power_law_1.01,0.025195519924163818
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,1,128,power_law_1.01,0.023637759685516357
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,1,128,power_law_1.01,0.02443648099899292
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,1,128,power_law_1.01,0.026177279949188233
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,128,power_law_1.01,2.3921075439453126
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,1,128,power_law_1.01,0.025958399772644043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,1,128,power_law_1.01,0.025483520030975343
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,1,128,power_law_1.01,0.02519808053970337
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,1,128,power_law_1.01,0.026689279079437255
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,1,128,power_law_1.01,0.025452799797058105
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,1,128,power_law_1.01,0.026010880470275877
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,1,128,balanced,0.05534080028533935
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,1,128,power_law_1.01,0.026767361164093017
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,1,128,power_law_1.01,0.026928639411926268
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,1,128,power_law_1.01,0.02716928005218506
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,1,128,power_law_1.01,0.03720319986343384
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,1,128,balanced,0.07471744060516358
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,1,128,balanced,0.08724096298217773
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,1,128,power_law_1.01,0.024931840896606446
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,1,128,power_law_1.01,0.03284735918045044
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,1,128,power_law_1.01,0.05368703842163086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,1,128,power_law_1.01,0.04070144176483154
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,1,128,power_law_1.01,0.07519999980926514
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,1,128,power_law_1.01,0.058613758087158206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,128,power_law_1.2,0.0752614402770996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,1,128,power_law_1.01,0.09601920127868653
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,128,power_law_1.2,0.07570176124572754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,128,power_law_1.2,0.11546879768371583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,128,power_law_1.2,0.12027520179748534
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,128,power_law_1.2,0.12256768226623535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,128,power_law_1.2,0.13002623558044432
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,128,power_law_1.2,0.1374131202697754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,128,power_law_1.2,0.1523020839691162
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,128,power_law_1.2,0.14399616241455077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,128,power_law_1.2,0.14833536148071289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,128,power_law_1.2,0.16389888763427735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,128,power_law_1.2,0.17455488204956054
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,128,power_law_1.2,0.19555456161499024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,128,power_law_1.2,0.21358720779418947
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,1,128,power_law_1.01,0.12406528472900391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,128,power_law_1.2,0.23042816162109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,128,power_law_1.2,0.2662348747253418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,128,power_law_1.2,0.3283276748657227
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,1,128,power_law_1.01,0.161582088470459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,128,power_law_1.2,0.3989593505859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,128,power_law_1.2,0.6349977493286133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,128,power_law_1.2,0.8751219177246095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,128,power_law_1.2,0.07644800186157227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,128,power_law_1.2,1.3556005859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,128,power_law_1.2,0.05326848030090332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,128,power_law_1.2,0.052358398437499995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,128,power_law_1.2,0.053835520744323725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,128,power_law_1.2,2.7643341064453124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,128,power_law_1.2,0.05519360065460205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,128,power_law_1.2,0.06082816123962402
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,128,power_law_1.2,3.782557373046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,128,power_law_1.2,0.05532159805297852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,128,power_law_1.2,0.07593472003936767
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,128,power_law_1.2,0.06951551914215089
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,128,power_law_1.2,0.07405568122863769
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,128,power_law_1.2,0.07576191902160645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,128,power_law_1.2,0.07838592052459717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,128,power_law_1.2,0.08384384155273437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,128,power_law_1.2,0.09345151901245116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,128,power_law_1.2,0.06448383808135986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,128,power_law_1.2,0.11941760063171387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,128,power_law_1.2,0.13490303993225097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,128,power_law_1.2,0.16825088500976562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,128,power_law_1.2,0.23358720779418946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,128,power_law_1.2,0.35092094421386716
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,128,power_law_1.2,0.11089152336120604
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,128,power_law_1.2,5.893429565429687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,128,power_law_1.2,0.4039948654174805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,128,power_law_1.2,0.584769287109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,128,power_law_1.2,0.9828237152099609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,128,power_law_1.2,1.2605337524414062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,128,power_law_1.2,1.4933132934570312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,1,128,power_law_1.2,0.026165759563446044
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,1,128,power_law_1.2,0.025912320613861083
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,1,128,power_law_1.2,0.025992960929870607
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,1,128,power_law_1.2,0.025944321155548095
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,1,128,power_law_1.2,0.022746880054473877
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,1,128,power_law_1.2,0.025169920921325684
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,1,128,power_law_1.2,0.025167360305786136
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,1,128,power_law_1.2,0.025491199493408202
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,1,128,power_law_1.2,0.025497601032257077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,128,power_law_1.2,1.6899392700195315
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,1,128,power_law_1.2,0.026000640392303466
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,1,128,power_law_1.2,0.02602112054824829
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,1,128,power_law_1.2,0.023717119693756103
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,1,128,power_law_1.2,0.026136319637298584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,1,128,power_law_1.2,0.026698238849639892
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,1,128,power_law_1.2,0.02837503910064697
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,1,128,power_law_1.2,0.034895360469818115
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,128,power_law_1.2,2.719851379394531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,128,power_law_1.2,7.849774169921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,128,power_law_1.2,3.703278198242187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,1,128,power_law_1.01,0.3284620666503906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,1,128,power_law_1.01,0.23752319335937502
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,1,128,power_law_1.2,0.04005631923675537
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,1,128,power_law_1.2,0.06294655799865723
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,1,128,power_law_1.2,0.03092351913452148
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,1,128,power_law_1.2,0.0509503984451294
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,1,128,power_law_1.2,0.07761407852172851
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,1,128,power_law_1.2,0.0991385555267334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,64,balanced,0.0311078405380249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,64,balanced,0.05183616161346436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,64,balanced,0.06649856090545654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,64,balanced,0.10742015838623047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,64,balanced,0.11150719642639159
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,64,balanced,0.11462271690368653
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,64,balanced,0.12014080047607423
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,64,balanced,0.12034048080444334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,64,balanced,0.13528575897216796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,64,balanced,0.1229759979248047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,64,balanced,0.1282367992401123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,64,balanced,0.12934783935546873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,64,balanced,0.1444416046142578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,64,balanced,0.15179776191711425
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,64,balanced,0.15807231903076172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,64,balanced,0.17724416732788087
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,64,balanced,0.18642816543579102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,64,balanced,0.0278272008895874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,64,balanced,0.2088217544555664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,64,balanced,0.26708736419677737
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,64,balanced,0.316176643371582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,64,balanced,0.4364352035522461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,64,balanced,0.5384998321533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,64,balanced,0.7533197021484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,64,balanced,0.9625421142578124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,64,balanced,0.030842878818511964
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,64,balanced,0.032067840099334714
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,64,balanced,0.0320038390159607
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,64,balanced,1.4054054260253905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,64,balanced,0.04207231998443604
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,64,balanced,0.058822398185729985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,64,balanced,0.06402048110961914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,64,balanced,1.8751820373535157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,64,balanced,0.0687718391418457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,64,balanced,0.07000832080841064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,64,balanced,0.07401343822479248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,64,balanced,0.06470272064208984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,64,balanced,0.0677836799621582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,64,balanced,0.06909311771392822
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,64,balanced,0.07776512145996094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,1,128,power_law_1.2,0.1306150436401367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,64,balanced,0.08040703773498535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,64,balanced,0.08681344032287598
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,64,balanced,0.09693568229675294
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,64,balanced,0.10183808326721192
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,1,128,power_law_1.2,0.1857344055175781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,64,balanced,0.11635456085205079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,64,balanced,0.14605183601379396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,64,balanced,0.17932544708251955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,64,balanced,0.2539507293701172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,64,balanced,0.31282304763793944
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,64,balanced,0.4409088134765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,64,balanced,0.548875503540039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,64,balanced,0.8053695678710937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,64,balanced,1.0424857330322266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,64,balanced,0.056501760482788085
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,1,128,power_law_1.2,0.2727001571655273
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,1,128,power_law_1.2,0.3493107223510742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,1,128,power_law_1.2,0.5567257690429688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,2,64,balanced,0.018982399702072144
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,2,64,balanced,0.029290239810943603
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,2,64,balanced,0.029105920791625977
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,2,64,balanced,0.029361920356750486
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,2,64,balanced,0.019463679790496825
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,2,64,balanced,0.029521920680999757
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,2,64,balanced,0.030426878929138184
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,2,64,balanced,0.03004415988922119
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,2,64,balanced,0.02030080080032349
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,2,64,balanced,0.03078144073486328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,2,64,balanced,0.030196480751037598
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,2,64,balanced,0.0301580810546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,2,64,balanced,0.03053056001663208
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,2,64,balanced,0.030942718982696533
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,2,64,balanced,0.031381759643554684
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,2,64,balanced,0.034236159324646
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,2,64,balanced,0.023576319217681885
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,2,64,balanced,0.029360640048980712
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,2,64,balanced,0.029047040939331054
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,2,64,balanced,0.02976639986038208
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,64,power_law_1.01,0.06834559917449952
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,64,power_law_1.01,0.06757376194000245
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,64,power_law_1.01,0.06793727874755859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,64,power_law_1.01,0.07589119911193848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,64,power_law_1.01,0.10399488449096679
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,64,power_law_1.01,0.1084620761871338
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,64,power_law_1.01,0.11372287750244141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,64,power_law_1.01,0.12224512100219727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,64,power_law_1.01,0.12718208312988283
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,64,power_law_1.01,0.12575231552124025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,64,power_law_1.01,0.13540096282958985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,64,power_law_1.01,0.1401254367828369
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,64,power_law_1.01,0.15016448020935058
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,2,64,balanced,0.028885760307312013
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,64,power_law_1.01,0.16397951126098634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,64,power_law_1.01,0.17213184356689454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,64,power_law_1.01,0.18584064483642576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,64,power_law_1.01,0.19841791152954102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,64,power_law_1.01,0.26269952774047856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,64,power_law_1.01,0.30586368560791016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,64,power_law_1.01,0.42066814422607424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,64,power_law_1.01,0.536129264831543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,64,power_law_1.01,0.7884480285644531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,64,power_law_1.01,0.9186073303222656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,2,64,balanced,0.046545920372009275
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,2,64,balanced,0.049108481407165526
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,2,64,balanced,0.06161791801452636
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,64,power_law_1.01,1.5046450805664062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,64,power_law_1.01,0.04998015880584717
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,2,64,balanced,0.06383103847503661
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,64,power_law_1.01,0.04986495971679687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,64,power_law_1.01,0.04951680183410644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,64,power_law_1.01,1.9992472839355468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,64,power_law_1.01,0.0516044807434082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,64,power_law_1.01,0.05294335842132568
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,64,power_law_1.01,0.059077119827270506
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,64,power_law_1.01,0.07217023849487306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,64,power_law_1.01,0.06808703899383545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,64,power_law_1.01,0.06675327777862548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,64,power_law_1.01,0.07105408191680909
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,64,power_law_1.01,0.07775872230529786
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,64,power_law_1.01,0.07809279918670654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,64,power_law_1.01,0.09026176452636718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,64,power_law_1.01,0.09364224433898925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,64,power_law_1.01,3.222411499023438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,64,power_law_1.01,0.105600004196167
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,64,power_law_1.01,0.1075660800933838
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,64,power_law_1.01,0.14269311904907228
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,64,power_law_1.01,0.16694015502929688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,64,power_law_1.01,0.24129663467407228
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,64,power_law_1.01,0.2805452728271484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,64,power_law_1.01,3.6100250244140626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,64,power_law_1.01,0.443059196472168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,64,power_law_1.01,0.5589081573486328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,2,64,power_law_1.01,0.0293120002746582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,64,power_law_1.01,0.8043865966796876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,2,64,power_law_1.01,0.029268479347229003
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,2,64,power_law_1.01,0.028913919925689695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,64,power_law_1.01,1.238608627319336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,2,64,power_law_1.01,0.02929663896560669
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,2,64,power_law_1.01,0.02512383937835693
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,2,64,power_law_1.01,0.02636928081512451
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,64,power_law_1.01,1.5240127563476562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,2,64,power_law_1.01,0.027566080093383787
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,2,64,power_law_1.01,0.02764928102493286
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,2,64,power_law_1.01,0.02722048044204712
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,2,64,power_law_1.01,0.029484798908233644
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,2,64,power_law_1.01,0.029054720401763916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,2,64,power_law_1.01,0.02849024057388306
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,2,64,power_law_1.01,0.029605119228363036
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,2,64,power_law_1.01,0.02964992046356201
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,2,64,power_law_1.01,0.029256958961486817
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,64,power_law_1.01,2.1547506713867186
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,2,64,balanced,0.08745856285095215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,64,power_law_1.01,0.06623104095458984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,2,64,power_law_1.01,0.03130111932754517
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,2,64,balanced,0.10298624038696289
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,2,64,power_law_1.01,0.0427507209777832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,2,64,power_law_1.01,0.03901312112808227
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,2,64,power_law_1.01,0.052517118453979494
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,2,64,power_law_1.01,0.05516799926757813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,64,power_law_1.2,0.06876031875610351
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,2,64,power_law_1.01,0.07826943874359131
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,64,power_law_1.2,0.06950655937194825
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,64,power_law_1.2,0.06994688034057617
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,2,64,power_law_1.01,0.03384576082229614
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,2,64,power_law_1.01,0.095098876953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,64,power_law_1.2,0.07451136112213134
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,64,power_law_1.2,0.10552448272705077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,64,power_law_1.2,0.12090368270874023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,64,power_law_1.2,0.11027199745178222
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,64,power_law_1.2,0.12685183525085447
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,64,power_law_1.2,0.13227264404296873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,64,power_law_1.2,0.12442496299743652
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,64,power_law_1.2,0.14787839889526366
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,64,power_law_1.2,0.15925632476806642
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,64,power_law_1.2,0.17748479843139647
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,64,power_law_1.2,0.19070720672607422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,64,power_law_1.2,0.18943744659423828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,64,power_law_1.2,0.20574207305908204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,64,power_law_1.2,0.3168716812133789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,64,power_law_1.2,0.11597824096679688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,64,power_law_1.2,0.3648896026611328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,64,power_law_1.2,0.49531391143798825
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,64,power_law_1.2,0.6136153411865235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,64,power_law_1.2,1.044501724243164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,64,power_law_1.2,1.3444863891601562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,64,power_law_1.2,1.9383628845214844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,64,power_law_1.2,2.3977702331542967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,64,power_law_1.2,0.04855936050415039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,64,power_law_1.2,0.049697279930114746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,64,power_law_1.2,0.04827136039733887
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,2,64,power_law_1.01,0.11633279800415039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,64,power_law_1.2,0.052255997657775874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,64,power_law_1.2,0.06049280166625977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,64,power_law_1.2,0.0648691177368164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,64,power_law_1.2,0.07230976104736328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,64,power_law_1.2,0.0728768014907837
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,64,power_law_1.2,0.06906112194061279
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,64,power_law_1.2,0.07228799819946288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,64,power_law_1.2,0.07703423976898194
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,64,power_law_1.2,0.08187392234802246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,64,power_law_1.2,0.09438336372375489
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,64,power_law_1.2,0.10683391571044923
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,64,power_law_1.2,0.10920448303222656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,64,power_law_1.2,0.1155174446105957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,64,power_law_1.2,0.15599360466003417
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,64,power_law_1.2,0.2101823997497559
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,64,power_law_1.2,3.998269348144531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,64,power_law_1.2,0.28581632614135744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,64,power_law_1.2,0.3400153732299805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,64,power_law_1.2,0.5401190567016602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,64,power_law_1.2,5.599959106445313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,64,power_law_1.2,1.0932518768310548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,64,power_law_1.2,0.04787456035614014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,64,power_law_1.2,0.7660211181640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,64,power_law_1.2,1.2523583984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,2,64,power_law_1.2,0.029448959827423092
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,2,64,power_law_1.2,0.029283840656280518
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,2,64,power_law_1.2,0.028884479999542235
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,2,64,power_law_1.2,0.027635200023651125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,2,64,power_law_1.2,0.025569279193878174
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,2,64,power_law_1.2,0.026854400634765623
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,2,64,power_law_1.2,0.028053760528564453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,2,64,power_law_1.2,0.028058879375457764
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,2,64,power_law_1.2,0.02852992057800293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,2,64,power_law_1.2,0.028161280155181885
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,2,64,power_law_1.2,0.028453121185302733
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,2,64,power_law_1.2,0.02801919937133789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,2,64,power_law_1.2,0.02880000114440918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,2,64,power_law_1.2,0.030574080944061277
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,2,64,power_law_1.2,0.031870720386505125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,64,power_law_1.2,2.056830749511719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,2,64,power_law_1.2,0.036177918910980225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,64,power_law_1.2,3.0597042846679687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,2,64,power_law_1.01,0.14928256034851076
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,2,64,power_law_1.01,0.2467276763916016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,2,64,power_law_1.01,0.3106534385681152
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,2,64,power_law_1.2,0.03403007984161377
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,2,64,power_law_1.2,0.04950911998748779
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,2,64,power_law_1.2,0.05771135807037353
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,2,64,power_law_1.2,0.04114560127258301
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,2,64,power_law_1.2,0.07467391967773437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,4,32,balanced,0.027304959297180176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,4,32,balanced,0.04589439868927002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,4,32,balanced,0.06417664051055907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,4,32,balanced,0.07857920169830322
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,4,32,balanced,0.10771967887878417
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,4,32,balanced,0.13140224456787108
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,4,32,balanced,0.11680000305175782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,4,32,balanced,0.12050432205200194
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,4,32,balanced,0.12128767967224122
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,4,32,balanced,0.02888959884643555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,4,32,balanced,0.13474047660827637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,4,32,balanced,0.14133503913879394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,4,32,balanced,0.14823295593261718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,4,32,balanced,0.16538368225097658
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,4,32,balanced,0.1737740707397461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,4,32,balanced,0.16499839782714845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,4,32,balanced,0.1860032081604004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,4,32,balanced,0.22525568008422853
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,4,32,balanced,0.2673049545288086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,4,32,balanced,0.34787841796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,4,32,balanced,0.42720001220703124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,4,32,balanced,0.6047398376464843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,4,32,balanced,0.7586176300048828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,2,64,power_law_1.2,0.08931072235107422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,4,32,balanced,1.1038822174072265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,4,32,balanced,1.4277516174316407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,4,32,balanced,0.029344000816345216
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,4,32,balanced,0.03133568048477173
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,2,64,power_law_1.2,0.11359616279602051
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,4,32,balanced,0.03297408103942871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,4,32,balanced,0.05960447788238525
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,2,64,power_law_1.2,0.1724684715270996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,4,32,balanced,0.06705535888671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,4,32,balanced,0.07766528129577636
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,4,32,balanced,0.06655360221862792
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,4,32,balanced,0.0692300796508789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,4,32,balanced,2.1097138977050784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,4,32,balanced,0.07037312030792237
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,4,32,balanced,0.0778764820098877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,4,32,balanced,0.08466943740844726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,4,32,balanced,0.09745920181274415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,4,32,balanced,0.10399999618530273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,4,32,balanced,0.09643648147583009
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,4,32,balanced,0.046256642341613766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,4,32,balanced,2.7692044067382815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,4,32,balanced,0.10796671867370607
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,4,32,balanced,0.13228799819946288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,4,32,balanced,0.1622271919250488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,4,32,balanced,0.2065126419067383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,4,32,balanced,0.2576959991455078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,4,32,balanced,0.37881473541259764
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,4,32,balanced,0.4764006423950195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,4,32,balanced,0.6767961883544922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,4,32,balanced,0.8624114990234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,4,32,balanced,1.2866546630859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,4,32,balanced,0.08057344436645508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,4,32,balanced,1.6704051208496093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,2,64,power_law_1.2,0.23154815673828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,2,64,power_law_1.2,0.34157695770263674
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,2,64,power_law_1.2,0.5100825500488282
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,4,32,balanced,0.01986176013946533
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,4,32,balanced,0.035427839756011964
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,4,32,balanced,0.026027519702911377
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,4,32,balanced,0.03548160076141357
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,4,32,balanced,0.022723839282989503
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,4,32,balanced,0.03532032012939453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,4,32,balanced,0.03498111963272095
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,4,32,balanced,0.035397119522094726
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,4,32,balanced,0.035791358947753905
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,4,32,balanced,0.03598207950592041
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,4,32,balanced,0.03625472068786621
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,4,32,balanced,0.0364134407043457
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,4,32,balanced,0.03518464088439942
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,4,32,balanced,0.03693311929702759
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,4,32,balanced,0.03889663934707642
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,4,32,balanced,0.021071360111236573
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,4,32,balanced,0.04010240077972412
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,4,32,balanced,0.04269440174102783
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,4,32,balanced,0.03543936014175415
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,4,32,balanced,0.03528320074081421
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,4,32,power_law_1.01,0.07041664123535156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,4,32,power_law_1.01,0.07100416183471679
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,4,32,power_law_1.01,0.07180031776428222
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,4,32,power_law_1.01,0.07498112201690674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,4,32,power_law_1.01,0.07574528217315674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,4,32,power_law_1.01,0.10590208053588868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,4,32,power_law_1.01,0.11958399772644043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,4,32,power_law_1.01,0.1227660846710205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,4,32,power_law_1.01,0.11787775993347167
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,4,32,power_law_1.01,0.12090496063232421
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,4,32,power_law_1.01,0.13077119827270506
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,4,32,power_law_1.01,0.134137601852417
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,4,32,power_law_1.01,0.1536192035675049
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,4,32,power_law_1.01,0.16494592666625976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,4,32,power_law_1.01,0.17229696273803713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,4,32,power_law_1.01,0.16976896286010743
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,4,32,power_law_1.01,0.1935910415649414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,4,32,power_law_1.01,0.24181247711181642
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,4,32,power_law_1.01,0.2903385543823242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,4,32,power_law_1.01,0.39108097076416015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,4,32,balanced,0.03517568111419678
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,4,32,power_law_1.01,0.5362521743774413
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,4,32,power_law_1.01,0.7397465515136719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,4,32,power_law_1.01,1.0039667510986328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,4,32,power_law_1.01,1.3582566833496093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,4,32,balanced,0.07843711853027344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,4,32,balanced,0.08315648078918457
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,4,32,balanced,0.05761663913726807
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,4,32,power_law_1.01,0.054504961967468256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,4,32,power_law_1.01,0.052299518585205075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,4,32,power_law_1.01,2.0295321655273435
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,4,32,power_law_1.01,0.054817280769348144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,4,32,power_law_1.01,0.05625472068786621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,4,32,power_law_1.01,0.06365824222564698
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,4,32,balanced,0.06069632053375244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,4,32,power_law_1.01,0.07121535778045654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,4,32,power_law_1.01,0.07242239952087402
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,4,32,power_law_1.01,0.0671782398223877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,4,32,power_law_1.01,0.06792191982269287
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,4,32,power_law_1.01,0.07396224021911621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,4,32,power_law_1.01,0.0790771198272705
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,4,32,balanced,0.1174732780456543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,4,32,power_law_1.01,0.08324480056762695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,4,32,power_law_1.01,0.09671168327331543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,4,32,power_law_1.01,0.10421248435974122
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,4,32,power_law_1.01,0.10438783645629883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,4,32,power_law_1.01,0.111843843460083
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,4,32,power_law_1.01,0.15072768211364745
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,4,32,power_law_1.01,0.05713151931762696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,4,32,power_law_1.01,0.1804787254333496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,4,32,power_law_1.01,0.2269580841064453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,4,32,power_law_1.01,0.28728960037231444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,4,32,power_law_1.01,2.9697894287109374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,4,32,power_law_1.01,0.42893184661865236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,4,32,power_law_1.01,0.5513075256347657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,4,32,power_law_1.01,0.034548480510711674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,4,32,power_law_1.01,0.8100518035888673
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,4,32,power_law_1.01,3.5571405029296876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,4,32,power_law_1.01,0.035861759185791015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,4,32,power_law_1.01,0.03458303928375244
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,4,32,power_law_1.01,0.03532671928405762
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,4,32,power_law_1.01,0.032494080066680905
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,4,32,power_law_1.01,0.031230719089508058
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,4,32,power_law_1.01,1.139383010864258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,4,32,power_law_1.01,0.03292288064956665
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,4,32,power_law_1.01,0.03264512062072754
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,4,32,power_law_1.01,0.03296128034591675
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,4,32,power_law_1.01,0.0332915210723877
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,4,32,power_law_1.01,0.03380991935729981
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,4,32,power_law_1.01,1.4785830688476562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,4,32,power_law_1.01,0.036392960548400875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,4,32,power_law_1.01,0.034663679599761965
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,4,32,power_law_1.01,0.037374720573425294
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,4,32,power_law_1.01,0.03712512016296386
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,4,32,power_law_1.01,0.040166401863098146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,4,32,power_law_1.01,2.3050086975097654
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,4,32,power_law_1.01,0.03325311899185181
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,4,32,balanced,0.14073984146118163
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,4,32,power_law_1.01,0.06276351928710937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,4,32,power_law_1.01,0.06978047847747802
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,4,32,power_law_1.01,0.04636672019958496
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,4,32,power_law_1.01,0.09638527870178223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,4,32,power_law_1.2,0.06988671779632569
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,4,32,power_law_1.01,0.11576319694519044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,4,32,power_law_1.2,0.07154304027557373
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,4,32,power_law_1.01,0.15142016410827636
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,4,32,power_law_1.2,0.07160448074340821
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,4,32,power_law_1.01,0.04965888023376465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,4,32,power_law_1.2,0.07562111854553223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,4,32,power_law_1.2,0.11498880386352539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,4,32,power_law_1.2,0.10684032440185547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,4,32,power_law_1.2,0.11936896324157716
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,4,32,power_law_1.2,0.11784704208374024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,4,32,power_law_1.2,0.12254976272583007
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,4,32,power_law_1.2,0.13017855644226076
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,4,32,power_law_1.2,0.14594559669494628
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,4,32,power_law_1.2,0.14827903747558593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,4,32,power_law_1.2,0.1657151985168457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,4,32,power_law_1.2,0.17681407928466797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,4,32,power_law_1.2,0.18003583908081056
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,4,32,power_law_1.2,0.2048384094238281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,4,32,power_law_1.2,0.27336959838867186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,4,32,power_law_1.2,0.29712383270263676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,4,32,power_law_1.2,0.4668864059448242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,4,32,power_law_1.2,0.6025395202636719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,4,32,power_law_1.2,0.8426636505126954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,4,32,power_law_1.2,1.0990438079833984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,4,32,power_law_1.2,0.0748198413848877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,4,32,power_law_1.2,0.05149312019348144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,4,32,power_law_1.2,1.4694989013671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,4,32,power_law_1.2,0.052133121490478516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,4,32,power_law_1.2,0.054113278388977046
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,4,32,power_law_1.01,0.19480960845947265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,4,32,power_law_1.2,0.05553791999816895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,4,32,power_law_1.2,0.05811200141906738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,4,32,power_law_1.2,0.06258304119110107
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,4,32,power_law_1.2,0.07102464199066162
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,4,32,power_law_1.2,0.07050879955291747
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,4,32,power_law_1.2,0.06622975826263428
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,4,32,power_law_1.2,0.07590528011322022
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,4,32,power_law_1.2,0.08475008010864257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,4,32,power_law_1.2,0.08702719688415528
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,4,32,power_law_1.2,0.10226431846618653
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,4,32,power_law_1.2,0.1047052764892578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,4,32,power_law_1.2,3.229964904785156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,4,32,power_law_1.2,0.10339455604553223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,4,32,power_law_1.2,0.12264191627502441
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,4,32,power_law_1.2,0.06876287937164308
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,4,32,power_law_1.2,0.16448383331298827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,4,32,power_law_1.2,0.19763456344604494
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,4,32,power_law_1.2,0.27590911865234374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,4,32,power_law_1.2,0.3429286575317383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,4,32,power_law_1.2,0.47399551391601563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,4,32,power_law_1.2,0.6625292968750001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,4,32,power_law_1.2,2.1958822631835937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,4,32,power_law_1.2,0.03537408113479614
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,4,32,power_law_1.2,0.978698272705078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,4,32,power_law_1.2,0.03582207918167114
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,4,32,power_law_1.2,0.03551615953445435
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,4,32,power_law_1.2,1.1594432067871092
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,4,32,power_law_1.2,0.03638015985488892
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,4,32,power_law_1.2,0.03275264024734497
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,4,32,power_law_1.2,0.03191551923751831
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,4,32,power_law_1.2,0.03259392023086548
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,4,32,power_law_1.2,0.032940800189971926
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,4,32,power_law_1.2,0.032920320034027103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,4,32,power_law_1.2,4.326549682617188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,4,32,power_law_1.2,1.7179533386230468
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,4,32,power_law_1.2,0.03319551944732666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,4,32,power_law_1.2,0.03343487977981567
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,4,32,power_law_1.2,0.035411200523376464
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,4,32,power_law_1.2,0.036078081130981446
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,4,32,power_law_1.2,0.03868288040161133
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,4,32,power_law_1.2,0.043683838844299314
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,4,32,power_law_1.2,0.03753599882125854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,4,32,power_law_1.2,2.5088922119140626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,4,32,power_law_1.01,0.2969587135314941
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,4,32,power_law_1.01,0.34535552978515627
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,4,32,power_law_1.2,0.04251776218414306
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,4,32,power_law_1.2,0.05075712203979492
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,4,32,power_law_1.2,0.054675202369689944
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,4,32,power_law_1.2,0.0706879997253418
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,4,32,power_law_1.2,0.08702719688415528
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,8,16,balanced,0.02892159938812256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,8,16,balanced,0.0459276819229126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,8,16,balanced,0.05857151985168457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,8,16,balanced,0.07519231796264648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,8,16,balanced,0.09124223709106445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,8,16,balanced,0.11539584159851075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,8,16,balanced,0.12024703979492188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,8,16,balanced,0.1300800037384033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,8,16,balanced,0.13396608352661132
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,4,32,power_law_1.2,0.11246720314025879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,8,16,balanced,0.1405504035949707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,8,16,balanced,0.1627814483642578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,8,16,balanced,0.1846681594848633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,8,16,balanced,0.025952000617980957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,8,16,balanced,0.16864896774291993
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,8,16,balanced,0.20685184478759763
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,8,16,balanced,0.26039552688598633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,8,16,balanced,0.31199232101440433
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,8,16,balanced,0.42816638946533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,8,16,balanced,0.15476223945617676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,8,16,balanced,0.532495346069336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,8,16,balanced,0.1819443130493164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,8,16,balanced,0.7758719635009765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,8,16,balanced,0.9769087982177733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,8,16,balanced,1.438592071533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,4,32,power_law_1.2,0.1508019161224365
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,8,16,balanced,0.03029632091522217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,8,16,balanced,0.033807361125946046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,8,16,balanced,1.8736370849609376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,8,16,balanced,0.0365068793296814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,8,16,balanced,0.043550720214843755
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,8,16,balanced,0.059520001411437984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,8,16,balanced,0.07688576221466065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,8,16,balanced,0.07493120193481445
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,4,32,power_law_1.2,0.22014463424682615
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,8,16,balanced,0.07612671852111816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,8,16,balanced,0.080762882232666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,8,16,balanced,0.08252415657043458
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,8,16,balanced,0.08982272148132324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,8,16,balanced,0.09596927642822266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,8,16,balanced,0.10479743957519531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,8,16,balanced,0.11897215843200684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,8,16,balanced,0.10661760330200196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,8,16,balanced,0.11652223587036133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,8,16,balanced,0.13110272407531737
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,8,16,balanced,0.16813823699951172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,8,16,balanced,0.1992844772338867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,8,16,balanced,0.26274431228637696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,8,16,balanced,2.783078308105469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,8,16,balanced,0.33362560272216796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,8,16,balanced,0.4935129547119141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,8,16,balanced,0.6183692932128906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,8,16,balanced,0.8935359954833985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,8,16,balanced,3.6663552856445314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,8,16,power_law_1.01,0.0625433588027954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,8,16,balanced,1.1485273742675781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,8,16,power_law_1.01,0.07473408222198487
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,8,16,power_law_1.01,0.08125951766967773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,8,16,balanced,1.7168960571289062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,8,16,power_law_1.01,0.082807035446167
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,8,16,power_law_1.01,0.07333888053894043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,8,16,power_law_1.01,0.11531776428222655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,8,16,balanced,2.242655944824219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,8,16,power_law_1.01,0.1196671962738037
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,8,16,power_law_1.01,0.12391167640686036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,8,16,power_law_1.01,0.13518591880798342
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,8,16,power_law_1.01,0.13972607612609864
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,8,16,power_law_1.01,0.15556480407714843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,8,16,power_law_1.01,0.17350528717041017
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,8,16,power_law_1.01,0.1651468849182129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,8,16,power_law_1.01,0.17105535507202146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,8,16,power_law_1.01,0.1992959976196289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,8,16,power_law_1.01,0.24114048004150393
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,8,16,power_law_1.01,0.30605695724487303
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,8,16,power_law_1.01,0.4117939376831055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,8,16,power_law_1.01,0.5476070404052734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,8,16,power_law_1.01,0.12314751625061035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,8,16,power_law_1.01,0.07597824096679687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,8,16,power_law_1.01,0.7849996948242188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,8,16,power_law_1.01,0.9555494689941406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,8,16,power_law_1.01,0.04362368106842041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,8,16,power_law_1.01,1.4526336669921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,8,16,power_law_1.01,0.05958528041839599
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,8,16,power_law_1.01,0.06007296085357666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,8,16,power_law_1.01,1.8985061645507812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,8,16,power_law_1.01,0.06250495910644531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,8,16,power_law_1.01,0.06623104095458984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,8,16,power_law_1.01,0.08113920211791992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,8,16,power_law_1.01,0.06868991851806641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,8,16,power_law_1.01,0.07203711986541747
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,8,16,power_law_1.01,0.07426432132720948
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,8,16,power_law_1.01,0.0793446397781372
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,8,16,power_law_1.01,0.0828377628326416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,8,16,power_law_1.01,0.08956416130065918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,8,16,power_law_1.01,0.09668224334716798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,8,16,power_law_1.01,0.11323136329650878
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,8,16,power_law_1.01,0.10552319526672363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,8,16,power_law_1.01,0.10729727745056153
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,8,16,power_law_1.01,0.12468735694885254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,8,16,power_law_1.01,0.165032958984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,8,16,power_law_1.01,0.19851264953613282
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,4,32,power_law_1.2,0.2916748809814453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,8,16,power_law_1.01,0.25463552474975587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,8,16,power_law_1.01,3.7944705200195314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,8,16,power_law_1.01,0.33257598876953126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,8,16,power_law_1.01,0.48350975036621097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,8,16,power_law_1.01,0.5816076660156251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,8,16,power_law_1.2,0.06253183841705322
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,8,16,power_law_1.01,0.8527833557128905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,8,16,power_law_1.2,0.07136640071868897
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,8,16,power_law_1.2,0.07303167819976807
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,8,16,power_law_1.01,1.1530226898193359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,8,16,power_law_1.01,2.8652197265624997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,8,16,power_law_1.2,0.07780863761901854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,8,16,power_law_1.2,0.07971072196960449
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,8,16,power_law_1.2,0.08209280014038087
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,8,16,power_law_1.2,0.1216089630126953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,4,32,power_law_1.2,0.4760345458984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,8,16,power_law_1.2,0.11697152137756348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,8,16,power_law_1.01,1.7459609985351563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,8,16,power_law_1.2,0.12254719734191895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,8,16,power_law_1.2,0.12609408378601075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,8,16,power_law_1.2,0.13334912300109864
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,8,16,power_law_1.2,0.14454400062561035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,8,16,power_law_1.2,0.15501312255859373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,8,16,power_law_1.2,0.17863552093505858
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,8,16,power_law_1.2,0.16709760665893553
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,8,16,power_law_1.2,0.1719923210144043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,8,16,power_law_1.01,2.197068786621094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,8,16,power_law_1.2,0.20888063430786136
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,8,16,power_law_1.2,0.25627519607543947
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,8,16,power_law_1.2,0.30028287887573246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,8,16,power_law_1.2,0.45147262573242186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,8,16,power_law_1.2,0.576316146850586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,8,16,power_law_1.2,0.8279782104492188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,8,16,power_law_1.2,1.0306841278076173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,8,16,power_law_1.2,0.043105278015136715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,8,16,power_law_1.2,0.05798272132873535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,8,16,power_law_1.2,0.059884800910949706
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,8,16,power_law_1.2,1.5379533386230468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,8,16,power_law_1.2,0.06400000095367432
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,8,16,power_law_1.2,0.0636185598373413
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,8,16,power_law_1.2,0.07097983837127686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,8,16,power_law_1.2,2.01064453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,8,16,power_law_1.2,0.08016639709472656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,8,16,power_law_1.2,0.0755020809173584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,8,16,power_law_1.2,0.07919231891632081
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,8,16,power_law_1.2,0.08318464279174805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,8,16,power_law_1.2,0.0943718433380127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,8,16,power_law_1.2,0.09766783714294433
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,8,16,power_law_1.2,0.11529600143432617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,8,16,power_law_1.2,0.1045414352416992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,8,16,power_law_1.2,0.11065343856811523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,8,16,power_law_1.2,0.13205120086669922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,8,16,power_law_1.2,0.16718847274780274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,8,16,power_law_1.2,0.20275583267211914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,8,16,power_law_1.2,0.2829030418395996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,8,16,power_law_1.2,0.36219905853271483
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,8,16,power_law_1.2,2.948927917480469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,8,16,power_law_1.2,0.07206399917602539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,8,16,power_law_1.2,0.5057344055175781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,16,8,balanced,0.028390400409698487
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,8,16,power_law_1.2,0.656390380859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,16,8,balanced,0.03461503982543945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,16,8,balanced,0.07819136142730712
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,8,16,power_law_1.2,0.9365184020996093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,16,8,balanced,0.11304191589355468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,16,8,balanced,0.1526707172393799
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,16,8,balanced,0.14291584014892578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,8,16,power_law_1.2,1.2126041412353517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,16,8,balanced,0.1566361618041992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,16,8,balanced,0.1981337547302246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,16,8,balanced,0.20419328689575195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,16,8,balanced,0.20988672256469726
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,4,32,power_law_1.2,0.6435391998291016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,16,8,balanced,0.22421503067016602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,16,8,balanced,0.23313407897949218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,16,8,balanced,0.2497920036315918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,8,16,power_law_1.2,1.7628044128417968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,16,8,balanced,0.24657152175903319
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,16,8,balanced,0.2338457679748535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,16,8,balanced,0.27139328002929686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,8,16,power_law_1.2,2.339271697998047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,16,8,balanced,0.32162559509277344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,16,8,balanced,0.3687513732910156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,16,8,balanced,0.05542272090911865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,16,8,balanced,0.5418726348876953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,16,8,balanced,0.6364108657836913
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,16,8,balanced,0.06819200038909912
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,16,8,balanced,0.9466944122314453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,16,8,balanced,0.08768768310546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,16,8,balanced,1.1239730834960937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,16,8,balanced,0.21319936752319335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,16,8,balanced,1.6764378356933594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,16,8,balanced,0.45384063720703127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,16,8,balanced,0.4671833419799805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,16,8,balanced,0.4877926254272461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,16,8,balanced,0.5718041610717773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,16,8,balanced,2.1648077392578124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,8,16,power_law_1.2,3.9561676025390624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,16,8,balanced,0.7196147155761718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,16,8,balanced,0.6303744125366211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,16,8,balanced,0.7197119903564453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,16,8,balanced,3.18919677734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,16,8,balanced,0.7697062683105469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,16,8,balanced,0.6545318603515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,16,8,balanced,0.7346534729003906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,16,8,balanced,0.672779541015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,16,8,balanced,0.7509375762939453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,16,8,balanced,0.8059993743896484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,16,8,balanced,4.134608764648438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,16,8,balanced,0.913597412109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,16,8,balanced,0.8862016296386719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,16,8,balanced,1.5443263244628906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,16,8,balanced,1.9374374389648437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,16,8,power_law_1.01,0.0631987190246582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,16,8,balanced,2.3643353271484377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,16,8,power_law_1.01,0.10165632247924805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,16,8,power_law_1.01,0.10777600288391113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,16,8,balanced,3.1617279052734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,16,8,power_law_1.01,0.11753727912902831
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,16,8,power_law_1.01,0.12500991821289062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,16,8,power_law_1.01,0.13022848129272463
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,16,8,balanced,4.200593872070312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,16,8,power_law_1.01,0.13735296249389647
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,16,8,power_law_1.01,0.0922060775756836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,16,8,power_law_1.01,0.18592512130737304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,16,8,power_law_1.01,0.18680831909179688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,16,8,power_law_1.01,0.2095961570739746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,16,8,power_law_1.01,0.219866886138916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,16,8,power_law_1.01,0.2397542381286621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,16,8,power_law_1.01,0.22902143478393552
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,16,8,power_law_1.01,0.24761728286743162
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,16,8,power_law_1.01,0.2760793685913086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,16,8,power_law_1.01,0.4094668960571289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,16,8,balanced,5.582676391601562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,16,8,power_law_1.01,0.5243174362182617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,16,8,power_law_1.01,0.6800268554687501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,16,8,power_law_1.01,1.2426175689697265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,16,8,power_law_1.01,0.1963327980041504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,16,8,power_law_1.01,1.7501388549804688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,16,8,power_law_1.01,0.34661376953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,16,8,power_law_1.01,0.25056640625000004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,16,8,power_law_1.01,2.331260223388672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,16,8,power_law_1.01,0.43348350524902346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,16,8,power_law_1.01,0.9623206329345703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,16,8,power_law_1.01,0.4008281707763672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,16,8,power_law_1.01,0.4931532669067383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,16,8,power_law_1.01,3.499521179199219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,16,8,power_law_1.01,0.4729548645019531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,16,8,power_law_1.01,4.546222229003907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,16,8,balanced,0.15055871963500977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,16,8,balanced,0.6074342346191406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,16,8,balanced,0.6626124572753906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,16,8,power_law_1.01,0.36255104064941407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,16,8,power_law_1.01,0.4696499252319336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,16,8,power_law_1.01,0.4156454467773438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,16,8,power_law_1.01,0.6414463806152344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,16,8,power_law_1.01,0.7284262084960937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,16,8,power_law_1.01,0.539870719909668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,16,8,power_law_1.01,0.867151336669922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,16,8,power_law_1.01,0.8532057952880858
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,16,8,power_law_1.01,1.1764403533935546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,16,8,power_law_1.01,1.1771596527099608
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,16,8,power_law_1.01,1.8345074462890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,16,8,power_law_1.01,1.9057792663574218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,16,8,power_law_1.01,0.6128371047973633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,16,8,power_law_1.01,3.10210693359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,16,8,power_law_1.01,3.3232626342773437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,16,8,power_law_1.2,0.061122560501098634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,16,8,power_law_1.2,0.08366975784301758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,16,8,power_law_1.2,0.09111295700073242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,16,8,power_law_1.2,0.10604160308837891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,16,8,power_law_1.01,0.5699814224243165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,16,8,power_law_1.2,0.12379136085510253
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,16,8,power_law_1.01,0.6435699462890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,16,8,power_law_1.2,0.12951040267944336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,16,8,power_law_1.2,0.18213504791259766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,16,8,power_law_1.2,0.18829824447631835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,16,8,power_law_1.2,0.19942527770996094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,16,8,power_law_1.01,5.352394409179688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,16,8,power_law_1.2,0.21508863449096677
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,16,8,power_law_1.2,0.2252134323120117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,16,8,power_law_1.2,0.2505855941772461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,16,8,power_law_1.2,0.12544768333435058
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,16,8,power_law_1.2,0.23909759521484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,16,8,power_law_1.2,0.2535500717163086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,16,8,power_law_1.2,0.28674816131591796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,16,8,power_law_1.2,0.3529446411132812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,16,8,power_law_1.2,0.4210137557983399
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,16,8,power_law_1.2,0.7100492858886719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,16,8,power_law_1.2,0.9865715026855468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,16,8,power_law_1.01,7.802675170898437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,16,8,power_law_1.2,0.14208640098571776
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,16,8,power_law_1.2,1.2793740844726562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,16,8,power_law_1.2,1.8076441955566405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,16,8,power_law_1.2,0.18792448043823243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,16,8,power_law_1.2,0.5633459091186523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,16,8,power_law_1.2,2.407887420654297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,16,8,power_law_1.2,0.2670425605773926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,16,8,power_law_1.2,0.3937279891967774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,16,8,power_law_1.2,0.401550064086914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,16,8,power_law_1.2,0.4821593475341797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,16,8,power_law_1.2,0.41318912506103517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,16,8,power_law_1.2,0.543641586303711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,16,8,power_law_1.2,0.5561305618286132
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,16,8,power_law_1.2,0.5708198547363281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,16,8,power_law_1.2,0.6027148818969726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,16,8,power_law_1.01,0.7120281219482422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,16,8,power_law_1.2,3.619263916015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,16,8,power_law_1.2,0.6179929733276367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,16,8,power_law_1.2,0.7422335815429688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,16,8,power_law_1.2,4.746676330566406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,16,8,power_law_1.2,0.6715122985839843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,16,8,power_law_1.2,0.6367526245117188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,16,8,power_law_1.01,0.606163215637207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,16,8,power_law_1.01,0.6311193466186523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,16,8,power_law_1.2,0.8217369842529297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,16,8,power_law_1.2,0.8070028686523438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,16,8,power_law_1.2,0.942956771850586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,16,8,power_law_1.2,1.0708236694335938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,16,8,power_law_1.2,1.2238438415527344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,16,8,power_law_1.2,1.7470208740234376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,16,8,power_law_1.2,2.4314968872070315
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,32,4,balanced,0.034979839324951176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,32,4,balanced,0.0632908821105957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,32,4,balanced,0.09389311790466308
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,16,8,power_law_1.2,3.1112139892578123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,32,4,balanced,0.1449843215942383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,32,4,balanced,0.19104639053344727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,32,4,balanced,0.20115583419799804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,32,4,balanced,0.20494464874267576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,16,8,power_law_1.2,3.5603057861328127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,32,4,balanced,0.2150694465637207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,32,4,balanced,0.22043392181396487
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,32,4,balanced,0.22990335464477538
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,32,4,balanced,0.25256320953369144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,32,4,balanced,0.25816831588745115
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,32,4,balanced,0.27712255477905273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,32,4,balanced,0.26198015213012693
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,32,4,balanced,0.27348096847534176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,32,4,balanced,0.2967897605895996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,32,4,balanced,0.3476812744140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,32,4,balanced,0.3935334396362305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,32,4,balanced,0.6198847961425782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,16,8,power_law_1.2,5.062301330566406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,32,4,balanced,0.7142310333251953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,32,4,balanced,1.0314765167236328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,32,4,balanced,0.03909631967544556
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,32,4,balanced,1.2251558685302735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,32,4,balanced,0.06171775817871093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,32,4,balanced,0.09642111778259277
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,32,4,balanced,1.85351806640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,32,4,balanced,0.14200703620910643
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,32,4,balanced,0.26854143142700193
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,32,4,balanced,2.35849853515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,32,4,balanced,0.4396825790405273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,16,8,power_law_1.2,7.298095092773437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,32,4,balanced,0.595712013244629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,32,4,balanced,0.6095193481445312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,32,4,balanced,0.7486029052734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,32,4,balanced,0.6126489639282227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,32,4,balanced,0.7331622314453126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,32,4,balanced,0.6291993713378906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,32,4,balanced,3.45583740234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,32,4,balanced,1.0906355285644531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,32,4,balanced,0.7846310424804688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,32,4,balanced,0.8875456237792969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,32,4,balanced,0.7819391632080078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,32,4,balanced,0.9326067352294922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,32,4,balanced,0.8729305267333984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,32,4,balanced,1.1915955352783203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,32,4,balanced,1.0087296295166017
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,32,4,balanced,0.9860057830810547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,32,4,balanced,1.2443276977539062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,32,4,balanced,1.8507122802734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,32,4,balanced,4.474827575683594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,32,4,balanced,2.7789709472656248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,32,4,power_law_1.01,0.04522111892700195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,32,4,balanced,3.139624938964844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,32,4,power_law_1.01,0.07969279766082764
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,32,4,power_law_1.01,0.08551679611206055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,32,4,balanced,3.370670166015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,32,4,power_law_1.01,0.12392831802368165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,32,4,power_law_1.01,0.17219072341918945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,32,4,power_law_1.01,0.19156736373901367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,32,4,power_law_1.01,0.2010905647277832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,32,4,power_law_1.01,0.225164794921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,32,4,power_law_1.01,0.11329536437988282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,32,4,power_law_1.01,0.23853696823120116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,32,4,power_law_1.01,0.16223232269287108
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,32,4,power_law_1.01,0.24672128677368166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,32,4,power_law_1.01,0.18238336563110352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,32,4,power_law_1.01,0.25776639938354495
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,32,4,power_law_1.01,0.2727129554748535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,32,4,power_law_1.01,0.3025369644165039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,16,8,power_law_1.2,0.3460262298583984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,32,4,power_law_1.01,0.4168588638305664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,32,4,balanced,6.033062133789063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,32,4,power_law_1.01,0.4785715103149414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,32,4,power_law_1.01,0.6045388793945312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,16,8,power_law_1.2,0.4608755111694336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,32,4,power_law_1.01,1.0361177825927734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,32,4,power_law_1.01,1.3298675537109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,32,4,balanced,6.481392822265624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,32,4,power_law_1.01,0.1322764778137207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,32,4,power_law_1.01,0.2679974365234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,32,4,power_law_1.01,0.22387840270996096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,32,4,power_law_1.01,1.8578215026855467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,32,4,power_law_1.01,0.22321792602539064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,32,4,power_law_1.01,0.30955776214599606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,32,4,power_law_1.01,0.3853977584838867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,32,4,power_law_1.01,2.4979763793945313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,32,4,power_law_1.01,0.6266342544555664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,32,4,power_law_1.01,0.4872832107543945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,32,4,power_law_1.01,0.5071820831298828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,32,4,power_law_1.01,0.5284940719604492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,32,4,power_law_1.01,0.7546137237548829
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,32,4,power_law_1.01,0.6653427124023438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,32,4,power_law_1.01,0.667462387084961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,32,4,power_law_1.01,0.7535743713378906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,32,4,power_law_1.01,0.7147264099121093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,32,4,power_law_1.01,3.6093057250976566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,32,4,power_law_1.01,0.8442649841308594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,32,4,power_law_1.01,1.0935897827148438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,32,4,power_law_1.01,0.8882470703124999
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,32,4,power_law_1.01,4.83625732421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,32,4,power_law_1.01,1.4055859375000002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,32,4,power_law_1.01,1.7953062438964842
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,32,4,power_law_1.01,1.9964685058593752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,32,4,power_law_1.01,4.2177560424804685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,32,4,power_law_1.01,4.131196289062499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,32,4,power_law_1.2,0.07295872211456299
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,32,4,power_law_1.2,0.08541312217712402
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,32,4,power_law_1.2,0.11012607574462892
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,32,4,power_law_1.2,0.04673535823822021
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,32,4,power_law_1.01,5.470183715820313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,32,4,power_law_1.2,0.13195648193359374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,32,4,power_law_1.2,0.16676607131958007
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,32,4,power_law_1.2,0.1738342475891113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,32,4,power_law_1.2,0.18001279830932618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,32,4,power_law_1.2,0.1943129539489746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,32,4,power_law_1.2,0.2069183921813965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,32,4,power_law_1.2,0.23250944137573243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,32,4,power_law_1.2,0.25005311965942384
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,32,4,power_law_1.2,0.2770432090759277
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,32,4,power_law_1.01,0.73404541015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,32,4,power_law_1.01,0.7537369537353515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,32,4,power_law_1.2,0.2632179260253906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,32,4,power_law_1.2,0.27836416244506834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,32,4,power_law_1.2,0.33803009033203124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,32,4,power_law_1.2,0.42140033721923825
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,32,4,power_law_1.2,0.48823806762695315
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,32,4,power_law_1.2,0.6155161666870117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,32,4,power_law_1.2,0.7970214080810547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,32,4,power_law_1.2,1.0495372772216798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,32,4,power_law_1.01,1.3832691955566407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,32,4,power_law_1.2,0.24094976425170897
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,32,4,power_law_1.2,0.18409088134765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,32,4,power_law_1.2,1.8928973388671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,32,4,power_law_1.01,10.191654663085938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,32,4,power_law_1.2,0.25129600524902346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,32,4,power_law_1.2,1.3518873596191407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,32,4,power_law_1.2,0.31778175354003907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,32,4,power_law_1.2,0.38766849517822266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,32,4,power_law_1.2,0.38724224090576176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,32,4,power_law_1.2,2.5756454467773438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,32,4,power_law_1.01,2.628608093261719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,32,4,power_law_1.2,0.5125324630737305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,32,4,power_law_1.2,0.5981721496582031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,32,4,power_law_1.2,0.6491264343261719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,32,4,power_law_1.2,0.7128383636474609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,32,4,power_law_1.2,0.8056614685058594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,32,4,power_law_1.2,3.6844635009765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,32,4,power_law_1.2,0.5861708831787109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,32,4,power_law_1.2,1.1348966217041014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,32,4,power_law_1.2,0.760926742553711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,32,4,power_law_1.2,0.841760025024414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,32,4,power_law_1.2,0.8307532501220702
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,32,4,power_law_1.2,4.838425598144531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,32,4,power_law_1.2,1.2380467224121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,32,4,power_law_1.2,1.3587776184082032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,32,4,power_law_1.2,1.5139814758300782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,32,4,power_law_1.2,1.5433740234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,32,4,power_law_1.2,1.898961944580078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,32,8,balanced,0.02396415948867798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,32,8,balanced,0.02372096061706543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,32,8,balanced,0.023960320949554442
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,32,8,balanced,0.043169279098510746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,32,8,balanced,0.0456601619720459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,32,4,power_law_1.2,2.663285827636719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,32,8,balanced,0.046058239936828616
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,32,8,balanced,0.04604159832000733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,32,8,balanced,0.046554880142211916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,32,8,balanced,0.047307519912719725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,32,8,balanced,0.048139519691467285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,32,8,balanced,0.04815231800079346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,32,8,balanced,0.049132800102233885
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,32,8,balanced,0.04931839942932129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,32,8,balanced,0.05074304103851318
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,32,8,balanced,0.051503357887268064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,32,8,balanced,0.052230401039123533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,32,8,balanced,0.05281919956207275
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,32,8,balanced,0.055904002189636225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,32,4,power_law_1.2,3.1521524047851566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,32,8,balanced,0.057498879432678215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,32,8,balanced,0.06411007881164552
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,32,8,balanced,0.06115200042724609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,32,8,balanced,0.07230720043182373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,32,8,balanced,0.08808704376220702
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,32,8,balanced,0.10626048088073729
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,32,8,balanced,0.12118016242980958
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,32,8,balanced,0.1543552017211914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,32,8,balanced,0.026201601028442385
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,32,8,balanced,0.20195840835571288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,32,8,balanced,0.026828799247741696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,32,8,balanced,0.02636672019958496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,32,8,balanced,0.034955520629882816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,32,8,balanced,0.03504767894744873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,32,8,balanced,0.034965760707855224
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,32,8,balanced,0.03540992021560669
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,32,8,balanced,0.03510656118392944
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,32,8,balanced,0.0354150390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,32,8,balanced,0.03549952030181885
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,32,8,balanced,0.03580928087234497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,32,8,balanced,0.03625472068786621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,32,8,balanced,0.03607167959213257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,32,8,balanced,0.03664000034332275
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,32,8,balanced,0.037061119079589845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,32,8,balanced,0.0377292799949646
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,32,8,balanced,0.03910527944564819
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,32,4,power_law_1.2,0.5334783935546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,32,8,balanced,0.04034815788269043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,32,8,balanced,0.04158080101013183
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,32,8,balanced,0.04558720111846924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,32,8,balanced,0.04360703945159912
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,32,8,balanced,0.04926464080810547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,32,8,balanced,0.0558835220336914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,32,8,balanced,0.06826496124267578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,32,8,balanced,0.08111743927001953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,32,8,balanced,0.10444928169250489
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,32,8,balanced,0.13181568145751954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,32,4,power_law_1.2,8.205360717773438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,32,4,power_law_1.2,8.177611694335937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,32,4,power_law_1.2,5.468820190429687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,32,8,balanced,0.015331840515136719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,32,8,balanced,0.015532799959182738
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,32,8,balanced,0.015699199438095092
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,32,8,balanced,0.01573632001876831
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,32,8,balanced,0.01533951997756958
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,32,8,balanced,0.015764479637145994
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,32,8,balanced,0.015692800283432007
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,32,8,balanced,0.01577087998390198
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,32,8,balanced,0.015768320560455323
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,32,8,balanced,0.0157695996761322
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,32,8,balanced,0.015751680135726927
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,32,8,balanced,0.016083199977874756
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,32,8,balanced,0.016176639795303344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,32,8,balanced,0.015336960554122925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,32,4,power_law_1.2,0.5887871932983398
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,32,8,balanced,0.015750399827957152
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,32,8,balanced,0.016098560094833375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,32,8,balanced,0.015754239559173582
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,32,8,balanced,0.016930559873580935
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,32,8,balanced,0.01693439960479736
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,32,8,balanced,0.01769343972206116
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,32,8,balanced,0.01882240056991577
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,32,8,power_law_1.01,0.023956480026245116
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,32,8,balanced,0.020400640964508058
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,32,8,power_law_1.01,0.04420608043670654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,32,8,power_law_1.01,0.041113600730896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,32,8,power_law_1.01,0.045607681274414065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,32,8,power_law_1.01,0.04732800006866455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,32,8,power_law_1.01,0.04851840019226074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,32,8,power_law_1.01,0.04935935974121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,32,8,power_law_1.01,0.04445824146270752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,32,8,power_law_1.01,0.050149121284484864
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,32,8,power_law_1.01,0.050922241210937504
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,32,8,balanced,0.021594879627227784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,32,8,power_law_1.01,0.05330431938171387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,32,8,power_law_1.01,0.05462143898010254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,32,8,power_law_1.01,0.05887616157531739
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,32,8,balanced,0.02422015905380249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,32,8,power_law_1.01,0.06303232192993165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,32,8,power_law_1.01,0.06329343795776368
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,32,8,power_law_1.01,0.0575270414352417
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,32,8,power_law_1.01,0.08646143913269043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,32,8,power_law_1.01,0.1050649642944336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,32,8,power_law_1.01,0.06780672073364258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,32,8,power_law_1.01,0.1193446445465088
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,32,8,power_law_1.01,0.17839359283447265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,32,8,power_law_1.01,0.052229118347167966
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,32,8,power_law_1.01,0.1498483180999756
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,32,8,power_law_1.01,0.026338560581207277
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,32,8,balanced,0.027672319412231444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,32,8,power_law_1.01,0.035024640560150144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,32,8,power_law_1.01,0.035000319480896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,32,8,power_law_1.01,0.03454976081848145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,32,8,power_law_1.01,0.035148799419403076
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,32,8,power_law_1.01,0.03576832056045533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,32,8,power_law_1.01,0.03619839906692505
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,32,8,power_law_1.01,0.036231679916381834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,32,8,power_law_1.01,0.036618239879608154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,32,8,power_law_1.01,0.03783807992935181
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,32,8,power_law_1.01,0.038638079166412355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,32,8,power_law_1.01,0.039548161029815676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,32,8,power_law_1.01,0.04077824115753174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,32,8,power_law_1.01,0.2775436782836914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,32,8,power_law_1.01,0.04356351852416992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,32,8,power_law_1.01,0.04445695877075195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,32,8,power_law_1.01,0.04796031951904297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,32,8,power_law_1.01,0.3562995147705078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,32,8,power_law_1.01,0.5293952178955078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,32,8,power_law_1.01,0.03534719944000244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,32,8,power_law_1.01,0.7078451538085938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,32,8,power_law_1.01,0.05468160152435303
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,32,8,power_law_1.01,0.06597119808197022
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,32,8,power_law_1.01,0.04558720111846924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,32,8,power_law_1.01,0.2220479965209961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,32,8,power_law_1.01,0.10055295944213867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,32,8,balanced,0.03860608100891113
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,32,8,balanced,0.0450380802154541
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,32,8,power_law_1.01,0.01571455955505371
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,32,8,power_law_1.01,0.015772160291671753
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,32,8,power_law_1.01,0.015691519975662233
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,32,8,power_law_1.01,0.01544319987297058
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,32,8,power_law_1.01,0.01567103981971741
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,32,8,power_law_1.01,0.0778713607788086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,32,8,power_law_1.01,0.015767040252685546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,32,8,power_law_1.01,0.01575935959815979
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,32,8,power_law_1.01,0.015760639905929567
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,32,8,power_law_1.01,0.01578495979309082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,32,8,power_law_1.01,0.015864319801330566
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,32,8,power_law_1.01,0.016179200410842896
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,32,8,power_law_1.01,0.016669440269470214
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,32,8,power_law_1.01,0.01692415952682495
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,32,8,power_law_1.01,0.41187583923339843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,32,8,power_law_1.01,0.11563263893127443
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,32,8,power_law_1.01,0.19913600921630858
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,32,8,power_law_1.01,0.35387905120849605
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,32,8,power_law_1.01,0.017760000228881835
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,32,8,power_law_1.01,0.017433600425720213
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,32,8,power_law_1.01,0.016954879760742187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,32,8,power_law_1.01,0.018946559429168702
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,32,8,power_law_1.01,0.04260096073150635
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,32,8,power_law_1.01,0.03584511995315552
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,32,8,power_law_1.01,0.020235519409179687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,32,8,power_law_1.2,0.02319999933242798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,32,8,power_law_1.2,0.042375679016113284
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,32,8,power_law_1.2,0.04568960189819336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,32,8,power_law_1.2,0.04441472053527832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,32,8,power_law_1.01,0.0214899206161499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,32,8,power_law_1.2,0.04622335910797119
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,32,8,power_law_1.2,0.047713279724121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,32,8,power_law_1.2,0.04890751838684082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,32,8,power_law_1.2,0.04929920196533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,32,8,power_law_1.2,0.05066495895385742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,32,8,power_law_1.2,0.0512499189376831
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,32,8,power_law_1.2,0.052748799324035645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,32,8,power_law_1.2,0.053451519012451175
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,32,8,power_law_1.2,0.054850559234619144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,32,8,power_law_1.2,0.05688831806182861
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,32,8,power_law_1.2,0.059953918457031255
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,32,8,power_law_1.2,0.06311039924621582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,32,8,power_law_1.2,0.06246655941009521
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,32,8,power_law_1.2,0.0659494400024414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,32,8,power_law_1.2,0.0863321590423584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,32,8,power_law_1.2,0.10476799964904786
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,32,8,power_law_1.01,0.023503360748291017
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,32,8,power_law_1.2,0.12082431793212892
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,32,8,power_law_1.2,0.15139455795288087
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,32,8,power_law_1.2,0.2023936080932617
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,32,8,power_law_1.01,0.02716671943664551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,32,8,power_law_1.2,0.2803264045715332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,32,8,power_law_1.2,0.027226879596710208
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,32,8,power_law_1.2,0.03530751943588257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,32,8,power_law_1.2,0.03556864023208618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,32,8,power_law_1.2,0.03584511995315552
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,32,8,power_law_1.01,0.0702297592163086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,32,8,power_law_1.2,0.035416319370269775
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,32,8,power_law_1.2,0.03557375907897949
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,32,8,power_law_1.2,0.03584511995315552
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,32,8,power_law_1.2,0.036474881172180174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,32,8,power_law_1.2,0.037052159309387205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,32,8,power_law_1.2,0.03709952116012573
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,32,8,power_law_1.2,0.0382860803604126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,32,8,power_law_1.2,0.039509758949279786
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,32,8,power_law_1.2,0.04024703979492188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,32,8,power_law_1.2,0.04158336162567139
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,32,8,power_law_1.2,0.04404096126556396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,32,8,power_law_1.2,0.04563839912414551
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,32,8,power_law_1.01,0.06213759899139404
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,32,8,power_law_1.2,0.04361728191375732
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,32,8,power_law_1.2,0.04768256187438965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,32,8,power_law_1.2,0.35503616333007815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,32,8,power_law_1.2,0.6814310455322266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,32,8,power_law_1.2,0.5510758590698243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,32,8,power_law_1.2,0.05538815975189208
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,32,8,power_law_1.2,0.06498688220977783
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,32,8,power_law_1.01,0.09500288009643555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,32,8,power_law_1.2,0.07775872230529786
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,32,8,power_law_1.2,0.12962047576904295
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,32,8,power_law_1.2,0.015751680135726927
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,32,8,power_law_1.2,0.01585536003112793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,32,8,power_law_1.2,0.01573248028755188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,32,8,power_law_1.2,0.01574399948120117
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,32,8,power_law_1.2,0.015727360248565674
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,32,8,power_law_1.2,0.01608191967010498
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,32,8,power_law_1.2,0.015750399827957152
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,32,8,power_law_1.2,0.016019200086593625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,32,8,power_law_1.2,0.01579519987106323
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,32,8,power_law_1.2,0.016165119409561158
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,32,8,power_law_1.2,0.01615231990814209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,32,8,power_law_1.2,0.18137344360351562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,32,8,power_law_1.2,0.10301952362060547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,32,8,power_law_1.2,0.25282432556152346
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,32,8,power_law_1.2,0.017804800271987914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,32,8,power_law_1.2,0.3441676712036133
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,32,8,power_law_1.2,0.016583679914474486
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,32,8,power_law_1.2,0.0202726411819458
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,32,8,power_law_1.2,0.01700096011161804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,32,8,power_law_1.2,0.46844417572021485
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,32,8,power_law_1.2,0.016947200298309324
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,32,8,power_law_1.2,0.01778048038482666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,32,8,power_law_1.2,0.01854591965675354
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,32,8,power_law_1.01,0.13575936317443849
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,32,8,power_law_1.2,0.02148736000061035
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,32,8,power_law_1.2,0.02393728017807007
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,32,8,balanced,0.030105600357055663
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,32,8,balanced,0.03009536027908325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,32,8,balanced,0.030083839893341062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,32,8,balanced,0.055228161811828616
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,32,8,balanced,0.057112321853637696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,32,8,balanced,0.06002943992614747
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,32,8,balanced,0.06121088027954101
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,32,8,balanced,0.06212351799011231
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,32,8,balanced,0.06286975860595703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,32,8,balanced,0.06326784133911133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,32,8,balanced,0.06449920177459717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,32,8,balanced,0.0662220811843872
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,32,8,balanced,0.0671283197402954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,32,8,balanced,0.06833920001983643
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,32,8,balanced,0.06965760231018067
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,32,8,power_law_1.2,0.02705535888671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,32,8,balanced,0.07030655860900878
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,32,8,balanced,0.07159423828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,32,8,balanced,0.07593728065490722
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,32,8,balanced,0.07833727836608886
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,32,8,balanced,0.08748543739318848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,32,8,balanced,0.08788991928100585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,32,8,balanced,0.11782784461975097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,32,8,balanced,0.1298534393310547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,32,8,balanced,0.1574630355834961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,32,8,balanced,0.1806643295288086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,32,8,balanced,0.23437824249267578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,32,8,balanced,0.31283071517944333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,32,8,balanced,0.030117120742797855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,32,8,balanced,0.030969600677490234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,32,8,balanced,0.030507519245147708
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,32,8,balanced,0.04078976154327393
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,32,8,balanced,0.04070655822753906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,32,8,balanced,0.04076032161712646
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,32,8,balanced,0.040828161239624024
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,32,8,power_law_1.2,0.03689088106155396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,32,8,balanced,0.04080512046813965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,32,8,balanced,0.04073984146118164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,32,8,balanced,0.041152000427246094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,32,8,balanced,0.0407257604598999
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,32,8,balanced,0.04153088092803955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,32,8,balanced,0.041556482315063474
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,32,8,balanced,0.041991682052612306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,32,8,balanced,0.04188159942626953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,32,8,balanced,0.042365441322326655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,32,8,balanced,0.04408576011657715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,32,8,balanced,0.04644864082336426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,32,8,balanced,0.04897791862487793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,32,8,balanced,0.055128321647644044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,32,8,balanced,0.06281599998474122
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,32,8,balanced,0.07262335777282715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,32,8,balanced,0.0892403221130371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,32,8,balanced,0.1061734390258789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,32,8,balanced,0.14287872314453126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,32,8,balanced,0.18387327194213868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,32,8,balanced,0.052291841506958
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,32,8,power_law_1.2,0.04306431770324707
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,32,8,power_law_1.2,0.062300162315368654
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,32,8,power_law_1.2,0.10244352340698241
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,32,8,power_law_1.2,0.07443583965301513
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,32,8,balanced,0.01740288019180298
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,32,8,balanced,0.01701120018959045
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,32,8,balanced,0.017431039810180664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,32,8,balanced,0.017308160066604614
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,32,8,balanced,0.0174399995803833
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,32,8,balanced,0.01736191987991333
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,32,8,balanced,0.017429759502410887
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,32,8,balanced,0.017163519859313962
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,32,8,balanced,0.017465599775314332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,32,8,balanced,0.01736448049545288
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,32,8,balanced,0.01741696000099182
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,32,8,balanced,0.017820160388946533
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,32,8,balanced,0.016940799951553346
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,32,8,balanced,0.01698815941810608
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,32,8,balanced,0.016921600103378294
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,32,8,power_law_1.2,0.13426688194274902
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,32,8,balanced,0.017171200513839722
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,32,8,balanced,0.01691264033317566
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,32,8,balanced,0.019059200286865235
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,32,8,balanced,0.019018239974975586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,32,8,balanced,0.019808000326156615
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,32,8,power_law_1.01,0.030461440086364745
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,32,8,power_law_1.01,0.05379327774047852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,32,8,power_law_1.01,0.05592703819274902
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,32,8,power_law_1.01,0.05699071884155273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,32,8,power_law_1.01,0.059552001953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,32,8,power_law_1.01,0.06281856060028077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,32,8,power_law_1.01,0.06452095985412598
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,32,8,balanced,0.021661438941955567
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,32,8,power_law_1.01,0.06664959907531738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,32,8,power_law_1.01,0.06803967952728271
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,32,8,power_law_1.01,0.06870783805847168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,32,8,power_law_1.01,0.07060863971710205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,32,8,power_law_1.01,0.0725158405303955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,32,8,power_law_1.01,0.07386112213134766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,32,8,power_law_1.01,0.07616511821746827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,32,8,power_law_1.01,0.07946879863739013
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,32,8,power_law_1.01,0.08302592277526856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,32,8,power_law_1.01,0.09571840286254882
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,32,8,power_law_1.01,0.11761792182922363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,32,8,power_law_1.01,0.12324607849121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,32,8,power_law_1.01,0.17548416137695313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,32,8,power_law_1.01,0.15135231971740723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,32,8,power_law_1.01,0.21286016464233398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,32,8,power_law_1.01,0.3637158584594727
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,32,8,balanced,0.022759680747985837
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,32,8,power_law_1.01,0.3037286376953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,32,8,balanced,0.02517888069152832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,32,8,power_law_1.01,0.7725555419921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,32,8,power_law_1.01,0.03091200113296509
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,32,8,power_law_1.01,0.040677118301391604
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,32,8,power_law_1.01,0.04070911884307861
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,32,8,power_law_1.01,0.04060160160064698
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,32,8,power_law_1.01,0.041347842216491695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,32,8,power_law_1.01,0.04107647895812988
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,32,8,power_law_1.01,0.041141757965087886
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,32,8,power_law_1.01,0.04159232139587402
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,32,8,power_law_1.01,0.04196479797363281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,32,8,power_law_1.01,0.04281856060028076
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,32,8,balanced,0.029057281017303465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,32,8,power_law_1.01,0.04363135814666748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,32,8,power_law_1.01,0.04723455905914307
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,32,8,power_law_1.01,0.05009920120239257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,32,8,power_law_1.01,0.05262335777282715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,32,8,power_law_1.01,0.04601088047027588
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,32,8,power_law_1.01,0.054192638397216795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,32,8,power_law_1.01,0.062325758934020994
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,32,8,power_law_1.01,0.07041408061981201
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,32,8,power_law_1.01,0.03904128074645996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,32,8,balanced,0.03424256086349488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,32,8,power_law_1.01,0.08017536163330077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,32,8,power_law_1.01,0.10334336280822753
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,32,8,power_law_1.01,0.5735359954833984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,32,8,balanced,0.05033472061157227
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,32,8,balanced,0.06099967956542969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,32,8,power_law_1.01,0.01739776015281677
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,32,8,power_law_1.01,0.017118719816207887
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,32,8,power_law_1.01,0.01741696000099182
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,32,8,power_law_1.01,0.016992000341415407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,32,8,power_law_1.01,0.017062400579452512
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,32,8,power_law_1.01,0.017404160499572753
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,32,8,power_law_1.01,0.01737727999687195
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,32,8,power_law_1.01,0.017363200187683107
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,32,8,power_law_1.01,0.017382400035858152
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,32,8,power_law_1.01,0.1322265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,32,8,power_law_1.01,0.017369600534439086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,32,8,power_law_1.01,0.017404160499572753
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,32,8,power_law_1.01,0.018197760581970215
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,32,8,power_law_1.01,0.019028480052948
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,32,8,power_law_1.01,0.18412927627563475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,32,8,power_law_1.01,0.27513599395751953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,32,8,power_law_1.01,0.3164569664001465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,32,8,power_law_1.01,0.49422592163085943
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,32,8,power_law_1.01,1.1113164520263672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,32,8,power_law_1.01,0.5895897674560546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,32,8,power_law_1.01,0.0183078396320343
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,32,8,power_law_1.01,0.019494400024414063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,32,8,power_law_1.01,0.019489279985427856
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,32,8,power_law_1.01,0.02100735902786255
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,32,8,power_law_1.01,0.022690560817718506
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,32,8,power_law_1.01,0.058090238571167
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,32,8,power_law_1.01,0.024416000843048097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,32,8,power_law_1.2,0.030536959171295164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,32,8,power_law_1.2,0.05388160228729248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,32,8,power_law_1.2,0.05709695816040039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,32,8,power_law_1.2,0.05547391891479492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,32,8,power_law_1.2,0.05962495803833008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,32,8,power_law_1.2,0.06288640022277832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,32,8,power_law_1.2,0.06518400192260743
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,32,8,power_law_1.2,0.06705408096313477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,32,8,power_law_1.2,0.0688972806930542
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,32,8,power_law_1.2,0.06914559841156007
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,32,8,power_law_1.2,0.07068416118621826
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,32,8,power_law_1.2,0.07306623935699463
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,32,8,power_law_1.2,0.07449855804443359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,32,8,power_law_1.2,0.07687935829162598
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,32,8,power_law_1.2,0.07965184211730957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,32,8,power_law_1.2,0.08310015678405762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,32,8,power_law_1.2,0.09569151878356934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,32,8,power_law_1.2,0.11912063598632812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,32,8,power_law_1.2,0.1277836799621582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,32,8,power_law_1.2,0.15292927742004395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,32,8,power_law_1.2,0.17079168319702148
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,32,8,power_law_1.01,0.028120319843292236
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,32,8,power_law_1.01,0.03354367971420288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,32,8,power_law_1.2,0.23313663482666014
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,32,8,power_law_1.01,0.048800001144409175
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,32,8,power_law_1.2,0.30965887069702147
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,32,8,power_law_1.01,0.133570556640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,32,8,power_law_1.2,0.44659584045410156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,32,8,power_law_1.2,0.030918400287628173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,32,8,power_law_1.2,0.04033664226531982
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,32,8,power_law_1.2,0.04074111938476562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,32,8,power_law_1.2,0.04068352222442627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,32,8,power_law_1.2,0.04074111938476562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,32,8,power_law_1.2,0.040839681625366206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,32,8,power_law_1.2,0.04114431858062744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,32,8,power_law_1.2,0.04115839958190918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,32,8,power_law_1.2,0.041890559196472166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,32,8,power_law_1.2,0.04197760105133057
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,32,8,power_law_1.2,0.0429747200012207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,32,8,power_law_1.2,0.04447360038757324
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,32,8,power_law_1.01,0.10362751960754395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,32,8,power_law_1.2,0.045382399559021
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,32,8,power_law_1.01,0.07619455814361573
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,32,8,power_law_1.2,0.05044095993041993
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,32,8,power_law_1.2,0.05221888065338135
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,32,8,power_law_1.2,0.054850559234619144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,32,8,power_law_1.2,0.060737280845642096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,32,8,power_law_1.2,0.04775040149688721
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,32,8,power_law_1.2,0.07018752098083496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,32,8,power_law_1.2,0.08702336311340332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,32,8,power_law_1.2,0.10624256134033203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,32,8,power_law_1.2,0.5897830581665039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,32,8,power_law_1.2,0.12920703887939453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,32,8,power_law_1.2,0.18398975372314452
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,32,8,power_law_1.2,0.01654016017913818
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,32,8,power_law_1.2,0.016915199756622316
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,32,8,power_law_1.2,0.016919039487838745
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,32,8,power_law_1.2,0.0171289598941803
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,32,8,power_law_1.2,0.016993279457092288
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,32,8,power_law_1.2,0.01731711983680725
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,32,8,power_law_1.2,0.01698559999465942
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,32,8,power_law_1.2,0.017392640113830564
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,32,8,power_law_1.2,0.017326079607009888
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,32,8,power_law_1.2,0.017422080039978027
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,32,8,power_law_1.2,0.27512191772460937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,32,8,power_law_1.2,0.31435903549194333
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,32,8,power_law_1.2,0.018640639781951903
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,32,8,power_law_1.2,0.018885120153427123
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,32,8,power_law_1.2,0.017460479736328124
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,32,8,power_law_1.2,0.019880959987640383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,32,8,power_law_1.2,0.8126169586181641
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,32,8,power_law_1.2,0.022672638893127442
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,32,8,power_law_1.2,1.0560768127441407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,32,8,power_law_1.2,0.02886399984359741
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,32,8,power_law_1.2,0.5157632064819336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,32,8,power_law_1.01,0.18315391540527343
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,32,8,power_law_1.2,0.06112639904022217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,32,8,power_law_1.2,0.6798194885253906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,32,8,power_law_1.2,0.018318079710006714
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,32,8,power_law_1.2,0.019621119499206544
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,32,8,power_law_1.2,0.02067199945449829
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,256,balanced,0.06484096050262451
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,256,balanced,0.0647488021850586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,256,balanced,0.0648307180404663
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,256,balanced,0.06477824211120606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,256,balanced,0.06474368095397949
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,256,balanced,0.06506752014160157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,256,balanced,0.11953280448913575
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,256,balanced,0.1160857582092285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,256,balanced,0.12116607666015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,256,balanced,0.12152064323425293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,256,balanced,0.12171263694763183
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,256,balanced,0.12058367729187011
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,256,balanced,0.12106623649597167
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,256,balanced,0.12107263565063478
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,32,8,power_law_1.2,0.024797439575195312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,256,balanced,0.1239680004119873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,256,balanced,0.12014335632324218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,256,balanced,0.13036160469055175
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,256,balanced,0.13395968437194825
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,256,balanced,0.14303232192993165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,256,balanced,0.14148223876953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,256,balanced,0.14721920013427733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,256,balanced,0.16033151626586914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,256,balanced,0.15523839950561524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,256,balanced,0.1669081687927246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,256,balanced,0.1810905647277832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,256,balanced,0.19913215637207032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,256,balanced,0.22410240173339843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,256,balanced,0.03781888008117676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,256,balanced,0.037904639244079594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,256,balanced,0.037849600315094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,256,balanced,0.03774847984313965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,256,balanced,0.03788543939590454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,256,balanced,0.04853759765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,256,balanced,0.03783040046691895
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,32,8,power_law_1.2,0.09533568382263183
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,256,balanced,0.04868864059448243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,256,balanced,0.04850560188293457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,256,balanced,0.04824063777923584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,256,balanced,0.04900479793548584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,256,balanced,0.04845312118530273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,256,balanced,0.04978303909301758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,256,balanced,0.04968063831329346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,256,balanced,0.05098368167877197
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,256,balanced,0.050115838050842285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,256,balanced,0.04813055992126465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,256,balanced,0.05466879844665527
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,256,balanced,0.05253759860992432
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,256,balanced,0.05210752010345458
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,256,balanced,0.055015678405761714
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,256,balanced,0.059288320541381834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,256,balanced,0.0584332799911499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,256,balanced,0.0718233585357666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,256,balanced,0.07846399784088134
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,256,balanced,0.09089152336120605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,256,balanced,0.10496128082275391
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,32,8,power_law_1.2,0.03458559989929199
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,32,8,power_law_1.2,0.045830402374267575
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,32,8,power_law_1.2,0.13576576232910156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,32,8,power_law_1.2,0.07941120147705079
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,1,256,balanced,0.01897984027862549
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,1,256,balanced,0.019311360120773315
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,1,256,balanced,0.019020800590515134
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,1,256,balanced,0.019095040559768676
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,1,256,balanced,0.018967039585113525
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,1,256,balanced,0.01924479961395264
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,1,256,balanced,0.01903615951538086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,1,256,balanced,0.018986239433288574
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,1,256,balanced,0.01897984027862549
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,1,256,balanced,0.01938815951347351
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,1,256,balanced,0.01930367946624756
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,1,256,balanced,0.019048960208892823
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,1,256,balanced,0.01944831967353821
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,1,256,balanced,0.019482879638671874
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,1,256,balanced,0.019425280094146728
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,1,256,balanced,0.019400960206985472
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,1,256,balanced,0.019838720560073853
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,1,256,balanced,0.020206079483032227
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,1,256,balanced,0.018996479511260985
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,1,256,balanced,0.01898751974105835
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,1,256,balanced,0.01904255986213684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,256,power_law_1.01,0.06375936031341553
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,256,power_law_1.01,0.06367487907409668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,256,power_law_1.01,0.06363647937774658
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,256,power_law_1.01,0.06412543773651122
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,256,power_law_1.01,0.11811584472656249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,256,power_law_1.01,0.12765567779541015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,1,256,balanced,0.019454720020294188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,256,power_law_1.01,0.12835840225219725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,256,power_law_1.01,0.13180800437927248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,256,power_law_1.01,0.12917247772216797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,256,power_law_1.01,0.13532544136047364
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,256,power_law_1.01,0.1313203239440918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,256,power_law_1.01,0.1428006362915039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,256,power_law_1.01,0.13567487716674803
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,256,power_law_1.01,0.1404032039642334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,256,power_law_1.01,0.14440192222595213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,256,power_law_1.01,0.14896896362304687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,256,power_law_1.01,0.1588326358795166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,256,power_law_1.01,0.16038400650024415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,256,power_law_1.01,0.1745267105102539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,256,power_law_1.01,0.19159168243408203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,256,power_law_1.01,0.2134105682373047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,256,power_law_1.01,0.28051712036132814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,256,power_law_1.01,0.4096332931518555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,256,power_law_1.01,0.5747430419921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,256,power_law_1.01,0.7530585479736328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,256,power_law_1.01,1.1914444732666016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,256,power_law_1.01,0.03790335893630982
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,256,power_law_1.01,0.03823487997055054
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,256,power_law_1.01,1.4463641357421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,256,power_law_1.01,0.037989120483398434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,256,power_law_1.01,0.037465600967407225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,256,power_law_1.01,0.04889855861663818
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,256,power_law_1.01,0.048552961349487306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,256,power_law_1.01,0.05015423774719239
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,32,8,power_law_1.2,0.1825382423400879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,256,power_law_1.01,0.051842560768127446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,256,power_law_1.01,0.05348095893859863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,256,power_law_1.01,0.05429376125335693
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,256,power_law_1.01,0.05678847789764404
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,256,power_law_1.01,0.058734078407287595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,256,power_law_1.01,0.051521282196044925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,256,power_law_1.01,0.053465600013732906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,256,power_law_1.01,0.05547776222229004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,256,power_law_1.01,0.04894591808319092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,256,power_law_1.01,0.058348798751831056
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,1,256,balanced,0.019475200176239015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,256,power_law_1.01,0.06500351905822754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,256,power_law_1.01,0.074901762008667
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,256,power_law_1.01,0.08512895584106446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,256,power_law_1.01,0.09924863815307618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,256,power_law_1.01,0.13228416442871094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,256,power_law_1.01,0.16847488403320313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,256,power_law_1.01,0.247955207824707
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,1,256,power_law_1.01,0.01943295955657959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,256,power_law_1.01,0.3472204971313476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,256,power_law_1.01,0.4906496047973633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,1,256,power_law_1.01,0.019182080030441286
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,1,256,power_law_1.01,0.019408639669418335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,256,power_law_1.01,0.6697702026367187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,1,256,power_law_1.01,0.019530240297317505
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,1,256,power_law_1.01,0.019472639560699462
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,1,256,power_law_1.01,0.019014400243759156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,1,256,power_law_1.01,0.01980031967163086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,1,256,power_law_1.01,0.019470080137252807
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,1,256,power_law_1.01,0.019032319784164427
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,1,256,power_law_1.01,0.0193612802028656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,1,256,power_law_1.01,0.019485440254211426
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,1,256,power_law_1.01,0.019528959989547727
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,1,256,power_law_1.01,0.019687680006027223
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,1,256,power_law_1.01,0.019425280094146728
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,1,256,power_law_1.01,0.019965440034866333
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,1,256,power_law_1.01,0.019819519519805907
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,1,256,balanced,0.022247679233551025
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,1,256,balanced,0.02183039903640747
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,1,256,balanced,0.025029120445251463
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,1,256,balanced,0.0265446400642395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,1,256,power_law_1.01,0.019500800371170045
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,1,256,power_law_1.01,0.02183552026748657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,1,256,power_law_1.01,0.021584639549255373
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,1,256,power_law_1.01,0.024766719341278075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,256,power_law_1.2,0.04484479904174805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,256,power_law_1.2,0.04513535976409912
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,256,power_law_1.2,0.044442877769470215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,256,power_law_1.2,0.10746239662170409
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,256,power_law_1.2,0.1195263957977295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,256,power_law_1.2,0.1317568016052246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,256,power_law_1.2,0.12307200431823731
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,256,power_law_1.2,0.1278809642791748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,256,power_law_1.2,0.12290687561035156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,256,power_law_1.2,0.1279475212097168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,256,power_law_1.2,0.1338150405883789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,256,power_law_1.2,0.13593600273132325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,256,power_law_1.2,0.1415385627746582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,256,power_law_1.2,0.1456332778930664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,256,power_law_1.2,0.1534886360168457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,256,power_law_1.2,0.15935999870300294
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,256,power_law_1.2,0.16436351776123045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,256,power_law_1.2,0.1782771110534668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,256,power_law_1.2,0.1964556884765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,256,power_law_1.2,0.23196544647216796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,256,power_law_1.2,0.30469120025634766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,256,power_law_1.2,0.4623961639404297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,256,power_law_1.2,0.6471641540527344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,1,256,power_law_1.01,0.02814079999923706
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,1,256,power_law_1.01,0.0379532790184021
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,256,power_law_1.2,0.9873868560791015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,1,256,power_law_1.01,0.043342080116271976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,256,power_law_1.2,1.4756568908691405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,256,power_law_1.2,0.03788160085678101
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,256,power_law_1.2,0.03782144069671631
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,256,power_law_1.2,0.04807040214538574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,256,power_law_1.2,0.0480614423751831
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,256,power_law_1.2,0.048889598846435546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,256,power_law_1.2,0.03780735969543457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,256,power_law_1.2,2.414392395019531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,256,power_law_1.2,0.04937600135803223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,256,power_law_1.2,0.05123968124389648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,256,power_law_1.2,0.05019904136657714
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,256,power_law_1.2,0.05461503982543945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,256,power_law_1.2,0.051925759315490726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,256,power_law_1.2,0.05099008083343506
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,256,power_law_1.2,0.05420928001403809
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,256,power_law_1.2,0.056707839965820316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,256,power_law_1.2,0.04846975803375244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,256,power_law_1.2,0.05919360160827637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,256,power_law_1.2,0.06574719905853271
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,256,power_law_1.2,3.8173797607421873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,256,power_law_1.2,0.08730624198913575
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,256,power_law_1.2,0.1039475154876709
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,256,power_law_1.2,0.13610367774963378
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,256,power_law_1.2,0.22403072357177733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,256,power_law_1.2,0.27335935592651367
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,1,256,power_law_1.01,0.06289663791656494
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,256,power_law_1.2,0.44740478515625004
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,1,256,power_law_1.2,0.01914368033409119
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,256,power_law_1.2,0.07672959804534912
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,1,256,power_law_1.2,0.019435520172119143
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,256,power_law_1.2,0.8601382446289062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,1,256,power_law_1.2,0.019435520172119143
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,1,256,power_law_1.2,0.018982399702072144
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,1,256,power_law_1.2,0.019426560401916503
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,1,256,power_law_1.2,0.018937599658966065
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,1,256,power_law_1.2,0.01941632032394409
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,1,256,power_law_1.2,0.01903488039970398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,256,power_law_1.2,0.6314918518066406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,1,256,power_law_1.2,0.019481600522994996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,1,256,power_law_1.2,0.019397120475769043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,1,256,power_law_1.2,0.01898751974105835
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,1,256,power_law_1.2,0.01948415994644165
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,1,256,power_law_1.2,0.01944831967353821
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,1,256,power_law_1.2,0.019404799938201905
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,1,256,power_law_1.2,0.02025343894958496
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,1,256,power_law_1.2,0.021003520488739012
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,1,256,power_law_1.2,0.024687359333038332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,256,power_law_1.2,1.440185546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,1,256,power_law_1.2,0.019367680549621583
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,1,256,power_law_1.01,0.08789504051208497
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,1,256,power_law_1.01,0.11724160194396974
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,1,256,power_law_1.01,0.1515558433532715
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,1,256,power_law_1.2,0.021940479278564455
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,1,256,power_law_1.2,0.03297663927078247
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,128,balanced,0.03453824043273926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,128,balanced,0.03420799970626831
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,1,256,power_law_1.2,0.04089856147766113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,128,balanced,0.0342412805557251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,128,balanced,0.04040832042694092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,128,balanced,0.03453056097030639
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,1,256,power_law_1.2,0.07803520202636718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,128,balanced,0.06162432193756103
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,1,256,power_law_1.2,0.05674240112304687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,128,balanced,0.09470335960388183
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,128,balanced,0.11320320129394532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,128,balanced,0.11576576232910156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,128,balanced,0.10495360374450684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,128,balanced,0.1139622402191162
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,128,balanced,0.11746432304382323
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,128,balanced,0.11346048355102538
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,128,balanced,0.11933312416076661
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,128,balanced,0.12077440261840819
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,128,balanced,0.12648703575134276
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,128,balanced,0.12422143936157226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,128,balanced,0.12935680389404297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,128,balanced,0.13255552291870115
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,128,balanced,0.1448960018157959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,128,balanced,0.13943167686462404
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,128,balanced,0.15130751609802245
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,128,balanced,0.1605939292907715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,128,balanced,0.17470207214355468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,128,balanced,0.19387903213500976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,128,balanced,0.23557247161865233
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,128,balanced,0.0350272011756897
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,128,balanced,0.03820544004440308
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,128,balanced,0.03498752117156982
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,128,balanced,0.03827327966690063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,128,balanced,0.035000319480896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,128,balanced,0.03713792085647583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,128,balanced,0.04547840118408203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,128,balanced,0.04541952133178711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,128,balanced,0.04639743804931641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,128,balanced,0.045438718795776364
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,128,balanced,0.04573823928833008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,128,balanced,0.04727935791015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,128,balanced,0.046474242210388185
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,128,balanced,0.04684927940368652
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,128,balanced,0.05017727851867676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,128,balanced,0.04815872192382813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,128,balanced,0.049295358657836914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,128,balanced,0.052634878158569334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,128,balanced,0.05575424194335937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,128,balanced,0.05462783813476563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,128,balanced,0.062492160797119146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,128,balanced,0.0702681589126587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,128,balanced,0.08104703903198243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,128,balanced,0.08981247901916503
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,128,balanced,0.11371775627136231
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,128,balanced,0.13805439949035644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,128,balanced,0.2769036865234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,128,balanced,0.04951935768127441
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,1,256,power_law_1.2,0.10666367530822754
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,1,256,power_law_1.2,0.1273318386077881
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,1,256,power_law_1.2,0.23667327880859376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,1,256,power_law_1.2,0.31673856735229494
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,2,128,balanced,0.0196070396900177
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,2,128,balanced,0.01980288028717041
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,2,128,balanced,0.019864319562911986
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,2,128,balanced,0.01978111982345581
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,2,128,balanced,0.019869439601898194
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,2,128,balanced,0.019402240514755247
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,2,128,balanced,0.019375360012054442
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,2,128,balanced,0.019870719909667968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,2,128,balanced,0.01988479971885681
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,2,128,balanced,0.01995136022567749
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,2,128,balanced,0.020254719257354736
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,2,128,balanced,0.02064255952835083
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,2,128,balanced,0.0202726411819458
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,2,128,balanced,0.020261120796203614
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,2,128,balanced,0.02035840034484863
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,2,128,balanced,0.020709118843078612
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,2,128,balanced,0.02065664052963257
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,2,128,balanced,0.021505279541015627
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,2,128,balanced,0.02025984048843384
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,2,128,balanced,0.019495680332183837
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,128,power_law_1.01,0.061768960952758786
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,128,power_law_1.01,0.06157440185546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,128,power_law_1.01,0.06210944175720214
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,128,power_law_1.01,0.061815037727355956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,128,power_law_1.01,0.11678079605102538
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,2,128,balanced,0.020160000324249267
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,128,power_law_1.01,0.12240768432617188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,128,power_law_1.01,0.11904000282287597
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,128,power_law_1.01,0.1251852798461914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,128,power_law_1.01,0.12477312088012696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,128,power_law_1.01,0.12853759765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,128,power_law_1.01,0.1304307174682617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,128,power_law_1.01,0.12894463539123535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,128,power_law_1.01,0.12892288208007813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,128,power_law_1.01,0.1347929573059082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,128,power_law_1.01,0.14300288200378417
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,128,power_law_1.01,0.1462169647216797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,128,power_law_1.01,0.14682623863220215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,128,power_law_1.01,0.15668352127075197
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,128,power_law_1.01,0.1619046401977539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,128,power_law_1.01,0.1857574462890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,128,power_law_1.01,0.20121599197387696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,128,power_law_1.01,0.25023231506347654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,128,power_law_1.01,0.3388608169555664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,128,power_law_1.01,0.5015795135498047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,2,128,balanced,0.019868160486221316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,128,power_law_1.01,0.6960153961181641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,128,power_law_1.01,0.8933004760742188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,128,power_law_1.01,0.03673471927642823
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,128,power_law_1.01,1.3009843444824218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,128,power_law_1.01,0.036819200515747066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,128,power_law_1.01,0.03788671970367431
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,128,power_law_1.01,0.0367193603515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,128,power_law_1.01,0.04615551948547363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,128,power_law_1.01,0.046045441627502445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,128,power_law_1.01,0.04823935985565185
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,128,power_law_1.01,0.04963327884674072
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,128,power_law_1.01,0.046856961250305175
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,128,power_law_1.01,0.0522380781173706
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,128,power_law_1.01,0.05465856075286866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,128,power_law_1.01,0.04925568103790283
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,128,power_law_1.01,0.05100031852722168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,128,power_law_1.01,0.05177599906921386
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,128,power_law_1.01,0.054269437789917
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,128,power_law_1.01,0.05664256095886231
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,128,power_law_1.01,0.05818624019622802
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,128,power_law_1.01,0.06654208183288575
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,128,power_law_1.01,0.07145984172821045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,128,power_law_1.01,0.0923136043548584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,128,power_law_1.01,0.10559359550476075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,128,power_law_1.01,0.1300595188140869
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,128,power_law_1.01,0.1686502456665039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,128,power_law_1.01,0.25004671096801756
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,128,power_law_1.01,0.31309823989868163
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,128,power_law_1.01,0.5099929428100586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,2,128,power_law_1.01,0.020101120471954344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,128,power_law_1.01,0.6967846679687499
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,2,128,power_law_1.01,0.019848959445953367
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,2,128,power_law_1.01,0.01970047950744629
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,2,128,power_law_1.01,0.02006527900695801
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,2,128,power_law_1.01,0.019898879528045654
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,2,128,power_law_1.01,0.019838720560073853
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,2,128,power_law_1.01,0.01976959943771362
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,2,128,power_law_1.01,0.02066431999206543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,2,128,power_law_1.01,0.020465919971466066
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,2,128,power_law_1.01,0.020261120796203614
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,2,128,power_law_1.01,0.020262401103973388
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,2,128,power_law_1.01,0.020696320533752442
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,2,128,power_law_1.01,0.020446720123291014
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,2,128,power_law_1.01,0.020583679676055906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,2,128,power_law_1.01,0.0202841591835022
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,2,128,power_law_1.01,0.020997118949890137
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,2,128,power_law_1.01,0.01990272045135498
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,2,128,balanced,0.01987328052520752
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,2,128,balanced,0.024302079677581787
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,2,128,balanced,0.023914239406585693
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,2,128,balanced,0.02723072052001953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,2,128,balanced,0.03239680051803589
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,2,128,power_law_1.01,0.02231679916381836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,2,128,power_law_1.01,0.0230732798576355
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,2,128,power_law_1.01,0.024761600494384767
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,128,power_law_1.2,0.057944321632385255
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,128,power_law_1.2,0.058072319030761724
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,128,power_law_1.2,0.0592409610748291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,128,power_law_1.2,0.10346624374389649
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,128,power_law_1.2,0.10638336181640624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,128,power_law_1.2,0.1165555191040039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,128,power_law_1.2,0.11333248138427734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,128,power_law_1.2,0.11631232261657716
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,128,power_law_1.2,0.11968128204345703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,128,power_law_1.2,0.12352000236511232
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,128,power_law_1.2,0.12613632202148437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,2,128,power_law_1.01,0.027151360511779788
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,128,power_law_1.2,0.1301465606689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,128,power_law_1.2,0.12765439987182617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,128,power_law_1.2,0.14326656341552735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,128,power_law_1.2,0.1474124813079834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,128,power_law_1.2,0.15072511672973632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,128,power_law_1.2,0.16110464096069338
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,128,power_law_1.2,0.17312768936157225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,128,power_law_1.2,0.1460377597808838
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,128,power_law_1.2,0.2079232025146484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,128,power_law_1.2,0.24894720077514648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,128,power_law_1.2,0.355731201171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,128,power_law_1.2,0.4727257537841797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,128,power_law_1.2,0.7746380615234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,128,power_law_1.2,0.9895513916015626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,2,128,power_law_1.01,0.032048640251159666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,128,power_law_1.2,1.4517733764648437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,128,power_law_1.2,0.036744959354400634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,128,power_law_1.2,0.037318398952484136
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,2,128,power_law_1.01,0.05837952136993409
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,2,128,power_law_1.01,0.04626431941986084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,128,power_law_1.2,0.036792318820953365
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,128,power_law_1.2,0.04482816219329834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,128,power_law_1.2,0.04551551818847656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,128,power_law_1.2,0.04708864212036133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,128,power_law_1.2,0.04610047817230224
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,128,power_law_1.2,0.04684415817260742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,128,power_law_1.2,0.04731520175933838
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,128,power_law_1.2,0.04889344215393067
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,128,power_law_1.2,0.049341440200805664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,128,power_law_1.2,2.1828965759277343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,128,power_law_1.2,0.05386240005493163
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,128,power_law_1.2,0.05586944103240967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,128,power_law_1.2,0.06147712230682373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,128,power_law_1.2,0.06207615852355956
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,2,128,power_law_1.01,0.07686399936676025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,128,power_law_1.2,0.07033088207244872
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,128,power_law_1.2,0.07828351974487305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,128,power_law_1.2,0.11077759742736817
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,128,power_law_1.2,0.11996543884277344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,128,power_law_1.2,0.05102848052978516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,128,power_law_1.2,0.19634687423706054
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,128,power_law_1.2,0.21813375473022462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,128,power_law_1.2,0.0510425615310669
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,128,power_law_1.2,0.3745113754272461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,2,128,power_law_1.2,0.020286719799041748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,128,power_law_1.2,0.5462796783447266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,2,128,power_law_1.2,0.019790079593658447
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,2,128,power_law_1.2,0.0198361599445343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,128,power_law_1.2,0.7126758575439454
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,2,128,power_law_1.2,0.019797760248184203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,2,128,power_law_1.2,0.01982079982757568
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,2,128,power_law_1.2,0.020275199413299562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,2,128,power_law_1.2,0.019846400022506712
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,2,128,power_law_1.2,0.019792640209198
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,2,128,power_law_1.2,0.020147199630737307
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,2,128,power_law_1.2,0.020244479179382324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,128,power_law_1.2,1.1097216033935546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,2,128,power_law_1.2,0.020357120037078857
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,2,128,power_law_1.2,0.02036479949951172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,2,128,power_law_1.2,0.020281600952148437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,2,128,power_law_1.2,0.020650238990783693
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,2,128,power_law_1.2,0.020309760570526122
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,2,128,power_law_1.2,0.02026623964309692
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,2,128,power_law_1.2,0.021518080234527587
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,2,128,power_law_1.2,0.027993600368499755
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,2,128,power_law_1.01,0.10765439987182616
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,2,128,power_law_1.01,0.14344703674316406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,2,128,power_law_1.2,0.02283776044845581
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,2,128,power_law_1.2,0.024387838840484618
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,2,128,power_law_1.2,0.08889599800109863
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,2,128,power_law_1.2,0.05372799873352051
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,64,balanced,0.03750272035598755
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,64,balanced,0.032592639923095704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,64,balanced,0.032119040489196775
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,64,balanced,0.03705471992492676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,64,balanced,0.035191040039062496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,64,balanced,0.05750016212463379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,64,balanced,0.11108736038208007
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,64,balanced,0.10575615882873535
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,2,128,power_law_1.2,0.03523967981338501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,64,balanced,0.1080742359161377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,64,balanced,0.11035264015197754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,64,balanced,0.10873215675354005
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,64,balanced,0.11032320022583009
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,64,balanced,0.11719296455383302
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,64,balanced,0.11950079917907715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,64,balanced,0.12102144241333009
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,64,balanced,0.1253606414794922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,64,balanced,0.12668928146362304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,64,balanced,0.1348697566986084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,64,balanced,0.1334182357788086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,64,balanced,0.14479871749877932
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,2,128,power_law_1.2,0.06139264106750488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,64,balanced,0.15166336059570312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,64,balanced,0.16671615600585937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,64,balanced,0.17919872283935548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,64,balanced,0.21475711822509766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,64,balanced,0.24883712768554686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,64,balanced,0.037480959892272944
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,64,balanced,0.03373568058013916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,64,balanced,0.3135974311828613
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,64,balanced,0.03377408027648926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,64,balanced,0.03746432065963745
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,64,balanced,0.383191032409668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,64,balanced,0.0353766393661499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,64,balanced,0.03627392053604126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,64,balanced,0.04566143989562989
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,64,balanced,0.04527743816375732
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,64,balanced,0.045771517753601075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,64,balanced,0.04732800006866455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,64,balanced,0.048564481735229495
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,64,balanced,0.048124160766601566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,64,balanced,0.04936960220336914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,64,balanced,0.04892416000366211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,64,balanced,0.04978047847747803
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,64,balanced,0.04987520217895508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,64,balanced,0.05340415954589843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,64,balanced,0.054274559020996094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,64,balanced,0.06314752101898194
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,64,balanced,0.0692748785018921
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,64,balanced,0.07528192043304444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,64,balanced,0.0842137622833252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,64,balanced,0.10897151947021484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,64,balanced,0.1269158363342285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,64,balanced,0.16893056869506834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,64,balanced,0.21248512268066405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,64,balanced,0.056313600540161136
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,2,128,power_law_1.2,0.12329855918884278
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,2,128,power_law_1.2,0.21000064849853514
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,2,128,power_law_1.2,0.22919168472290039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,4,64,balanced,0.019047679901123046
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,4,64,balanced,0.021821439266204834
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,4,64,balanced,0.019328000545501708
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,4,64,balanced,0.02144000053405762
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,4,64,balanced,0.019440640211105347
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,4,64,balanced,0.02189055919647217
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,4,64,balanced,0.022259199619293214
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,4,64,balanced,0.021916160583496096
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,4,64,balanced,0.022293760776519775
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,4,64,balanced,0.022269439697265626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,4,64,balanced,0.022152960300445557
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,4,64,balanced,0.022273280620574952
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,4,64,balanced,0.019760639667510987
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,4,64,balanced,0.022332160472869872
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,4,64,balanced,0.02233599901199341
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,4,64,balanced,0.022315518856048582
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,4,64,balanced,0.023975679874420168
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,4,64,balanced,0.019445760250091555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,4,64,balanced,0.02390399932861328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,4,64,balanced,0.02235647916793823
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,64,power_law_1.01,0.06304768085479737
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,64,power_law_1.01,0.06015232086181641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,64,power_law_1.01,0.06024831771850586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,64,power_law_1.01,0.06040832042694092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,64,power_law_1.01,0.1087820816040039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,64,power_law_1.01,0.11236991882324218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,64,power_law_1.01,0.11041024208068848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,64,power_law_1.01,0.11286911964416504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,64,power_law_1.01,0.11436415672302247
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,64,power_law_1.01,0.11853695869445802
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,64,power_law_1.01,0.12131711959838867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,64,power_law_1.01,0.1275584030151367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,64,power_law_1.01,0.1319603157043457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,64,power_law_1.01,0.13261568069458007
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,64,power_law_1.01,0.14414719581604002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,64,power_law_1.01,0.14003711700439453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,64,power_law_1.01,0.1431385612487793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,64,power_law_1.01,0.15132287979125975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,64,power_law_1.01,0.16160768508911133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,64,power_law_1.01,0.2006412887573242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,64,power_law_1.01,0.21769344329833987
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,64,power_law_1.01,0.27630592346191407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,64,power_law_1.01,0.314467830657959
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,4,64,balanced,0.021510400772094727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,64,power_law_1.01,0.44550014495849605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,64,power_law_1.01,0.7291763305664063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,4,64,balanced,0.021498880386352538
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,64,power_law_1.01,0.9315187072753905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,64,power_law_1.01,0.0366208004951477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,64,power_law_1.01,0.03638400077819824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,64,power_law_1.01,0.03746432065963745
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,64,power_law_1.01,0.036665599346160885
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,64,power_law_1.01,0.04605696201324463
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,64,power_law_1.01,1.2536768341064453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,64,power_law_1.01,0.04562816143035889
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,64,power_law_1.01,0.04934400081634521
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,64,power_law_1.01,0.05268735885620117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,64,power_law_1.01,0.04812032222747803
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,64,power_law_1.01,0.054613761901855476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,64,power_law_1.01,0.04894336223602295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,64,power_law_1.01,0.05057024002075196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,64,power_law_1.01,0.05226880073547363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,64,power_law_1.01,0.05566720008850098
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,64,power_law_1.01,0.05792384147644043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,64,power_law_1.01,0.05836671829223633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,64,power_law_1.01,0.06400383949279785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,64,power_law_1.01,0.06849023818969727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,64,power_law_1.01,0.07747200012207031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,64,power_law_1.01,0.09110272407531739
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,64,power_law_1.01,0.11967488288879395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,64,power_law_1.01,0.1623232078552246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,64,power_law_1.01,0.18063615798950194
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,64,power_law_1.01,0.2906227111816406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,64,power_law_1.01,0.400970230102539
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,4,64,power_law_1.01,0.0215500807762146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,64,power_law_1.01,0.6066790390014648
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,4,64,power_law_1.01,0.02149120092391968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,4,64,power_law_1.01,0.021429760456085207
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,4,64,power_law_1.01,0.021891839504241943
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,4,64,power_law_1.01,0.021495039463043215
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,4,64,power_law_1.01,0.021454079151153566
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,4,64,power_law_1.01,0.02144511938095093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,4,64,power_law_1.01,0.021931519508361814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,64,power_law_1.01,0.5148633575439453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,4,64,power_law_1.01,0.021573119163513184
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,4,64,power_law_1.01,0.021635839939117434
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,4,64,power_law_1.01,0.021804800033569334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,4,64,power_law_1.01,0.021982719898223875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,4,64,power_law_1.01,0.022319359779357912
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,4,64,power_law_1.01,0.022320640087127686
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,4,64,power_law_1.01,0.022213120460510254
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,4,64,power_law_1.01,0.022115840911865234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,4,64,power_law_1.01,0.022339839935302735
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,4,64,balanced,0.021509120464324953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,4,64,balanced,0.02814847946166992
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,4,64,balanced,0.035577600002288816
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,4,64,balanced,0.04369919776916504
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,4,64,balanced,0.029132800102233885
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,4,64,power_law_1.01,0.02353408098220825
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,4,64,power_law_1.01,0.02522111892700195
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,4,64,power_law_1.01,0.026501119136810303
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,64,power_law_1.2,0.05809152126312256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,64,power_law_1.2,0.056675839424133304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,64,power_law_1.2,0.05770880222320557
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,64,power_law_1.2,0.05763072013854981
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,64,power_law_1.2,0.08528256416320801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,64,power_law_1.2,0.1074995231628418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,64,power_law_1.2,0.11038335800170898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,64,power_law_1.2,0.11117695808410644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,64,power_law_1.2,0.11445759773254394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,64,power_law_1.2,0.11434495925903321
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,64,power_law_1.2,0.12297984123229981
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,64,power_law_1.2,0.13008128166198732
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,4,64,power_law_1.01,0.032960000038146975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,64,power_law_1.2,0.14222847938537597
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,64,power_law_1.2,0.13837696075439454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,64,power_law_1.2,0.14060799598693846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,64,power_law_1.2,0.14446080207824707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,64,power_law_1.2,0.13303936004638672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,64,power_law_1.2,0.1595967960357666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,64,power_law_1.2,0.18438144683837893
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,64,power_law_1.2,0.2244326400756836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,64,power_law_1.2,0.2606118392944336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,64,power_law_1.2,0.36453632354736326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,64,power_law_1.2,0.4711334228515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,64,power_law_1.2,0.7489958190917969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,64,power_law_1.2,0.9174143981933593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,64,power_law_1.2,1.2507289886474608
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,4,64,power_law_1.01,0.038988800048828126
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,4,64,power_law_1.01,0.04727935791015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,64,power_law_1.2,0.035348479747772214
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,64,power_law_1.2,0.03627007961273193
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,4,64,power_law_1.01,0.06746240139007569
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,64,power_law_1.2,0.036915199756622316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,64,power_law_1.2,0.04447103977203369
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,64,power_law_1.2,0.04563583850860596
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,64,power_law_1.2,0.047673602104187016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,64,power_law_1.2,0.04609792232513428
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,64,power_law_1.2,0.04756735801696778
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,64,power_law_1.2,0.051423997879028314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,64,power_law_1.2,0.05264639854431152
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,64,power_law_1.2,0.04978176116943359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,64,power_law_1.2,0.05461503982543945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,64,power_law_1.2,0.05266304016113281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,64,power_law_1.2,0.06516863822937011
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,64,power_law_1.2,0.06369408130645751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,64,power_law_1.2,0.06350719928741455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,64,power_law_1.2,0.0651916790008545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,64,power_law_1.2,0.07321599960327149
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,64,power_law_1.2,0.08895615577697755
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,64,power_law_1.2,0.10934911727905275
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,64,power_law_1.2,0.12602751731872558
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,64,power_law_1.2,0.18815744400024415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,64,power_law_1.2,0.22354047775268554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,64,power_law_1.2,1.8323110961914062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,64,power_law_1.2,0.3524403381347656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,4,64,power_law_1.01,0.08393600463867187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,4,64,power_law_1.2,0.02148736000061035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,64,power_law_1.2,0.5176307296752929
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,4,64,power_law_1.2,0.021542398929595946
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,4,64,power_law_1.2,0.021516799926757812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,4,64,power_law_1.2,0.02148224115371704
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,4,64,power_law_1.2,0.021056001186370847
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,4,64,power_law_1.2,0.021402881145477295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,64,power_law_1.2,0.7667097473144532
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,4,64,power_law_1.2,0.021751039028167725
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,4,64,power_law_1.2,0.021519360542297365
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,4,64,power_law_1.2,0.0214899206161499
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,4,64,power_law_1.2,0.02176383972167969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,4,64,power_law_1.2,0.021451520919799804
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,4,64,power_law_1.2,0.021903359889984132
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,64,power_law_1.2,1.0688563537597655
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,4,64,power_law_1.2,0.021856000423431395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,4,64,power_law_1.2,0.021944320201873778
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,4,64,power_law_1.2,0.022259199619293214
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,4,64,power_law_1.2,0.02308864116668701
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,4,64,power_law_1.2,0.02184448003768921
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,4,64,power_law_1.01,0.1038963222503662
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,4,64,power_law_1.01,0.1274304008483887
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,4,64,power_law_1.2,0.02388351917266846
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,4,64,power_law_1.2,0.028501760959625245
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,4,64,power_law_1.2,0.03625855922698974
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,8,32,balanced,0.03216768026351929
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,8,32,balanced,0.03175551891326904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,8,32,balanced,0.031692800521850584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,8,32,balanced,0.03338624000549316
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,4,64,power_law_1.2,0.05902719974517823
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,4,64,power_law_1.2,0.04440320014953613
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,8,32,balanced,0.06246016025543213
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,4,64,power_law_1.2,0.0329804801940918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,8,32,balanced,0.11761792182922363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,8,32,balanced,0.11811072349548339
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,8,32,balanced,0.11775744438171387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,8,32,balanced,0.12033151626586913
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,8,32,balanced,0.12124671936035156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,8,32,balanced,0.1178048038482666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,8,32,balanced,0.12394751548767088
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,8,32,balanced,0.12968704223632813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,8,32,balanced,0.13350655555725097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,8,32,balanced,0.13397888183593748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,8,32,balanced,0.13081343650817873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,8,32,balanced,0.03545344114303589
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,8,32,balanced,0.15031807899475097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,8,32,balanced,0.16245759963989256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,8,32,balanced,0.17345279693603516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,8,32,balanced,0.20830848693847653
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,8,32,balanced,0.2363596725463867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,8,32,balanced,0.29633407592773436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,8,32,balanced,0.3590796661376953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,8,32,balanced,0.1420582389831543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,8,32,balanced,0.4789977645874023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,8,32,balanced,0.03386240005493164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,8,32,balanced,0.03301248073577881
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,8,32,balanced,0.03379584074020386
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,8,32,balanced,0.5999513626098633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,8,32,balanced,0.03412863969802856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,8,32,balanced,0.03667327880859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,8,32,balanced,0.040320000648498534
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,8,32,balanced,0.05013887882232666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,8,32,balanced,0.05468287944793702
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,8,32,balanced,0.05180799961090088
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,8,32,balanced,0.0514521598815918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,8,32,balanced,0.06040832042694092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,8,32,balanced,0.053052158355712885
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,8,32,balanced,0.05404672145843505
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,8,32,balanced,0.05640704154968261
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,8,32,balanced,0.057840638160705574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,8,32,balanced,0.0591487979888916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,8,32,balanced,0.057715201377868654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,8,32,balanced,0.06575359821319579
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,8,32,balanced,0.07721343994140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,8,32,balanced,0.0851033592224121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,8,32,balanced,0.10414208412170409
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,8,32,balanced,0.12151807785034179
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,8,32,balanced,0.16830080032348632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,8,32,balanced,0.20102399826049805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,8,32,balanced,0.27969024658203123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,8,32,balanced,0.3599347305297852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,8,32,balanced,0.0708569622039795
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,4,64,power_law_1.2,0.09351167678833008
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,4,64,power_law_1.2,0.12076928138732909
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,4,64,power_law_1.2,0.18913536071777343
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,4,64,power_law_1.2,0.21185535430908203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,8,32,balanced,0.01941248059272766
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,8,32,balanced,0.023491840362548828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,8,32,balanced,0.02387455940246582
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,8,32,balanced,0.023880960941314696
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,8,32,balanced,0.019564800262451172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,8,32,balanced,0.024202239513397217
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,8,32,balanced,0.02405503988265991
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,8,32,balanced,0.02388864040374756
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,8,32,balanced,0.024375040531158448
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,8,32,balanced,0.01981824040412903
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,8,32,balanced,0.02428800106048584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,8,32,balanced,0.024128000736236572
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,8,32,balanced,0.0194214403629303
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,8,32,balanced,0.024323840141296387
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,8,32,balanced,0.024289278984069823
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,8,32,balanced,0.02600575923919678
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,8,32,balanced,0.02601344108581543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,8,32,balanced,0.0275161600112915
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,8,32,balanced,0.02920959949493408
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,8,32,balanced,0.021497600078582764
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,8,32,balanced,0.023544321060180666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,8,32,power_law_1.01,0.03745919942855835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,8,32,power_law_1.01,0.04813439846038818
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,8,32,power_law_1.01,0.05074687957763672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,8,32,power_law_1.01,0.05794943809509277
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,8,32,power_law_1.01,0.10655743598937988
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,8,32,power_law_1.01,0.10331775665283202
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,8,32,power_law_1.01,0.11766783714294433
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,8,32,power_law_1.01,0.1135923194885254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,8,32,power_law_1.01,0.11769856452941893
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,8,32,power_law_1.01,0.12097791671752929
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,8,32,power_law_1.01,0.12090880393981933
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,8,32,power_law_1.01,0.13026432037353514
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,8,32,power_law_1.01,0.1343564796447754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,8,32,power_law_1.01,0.13459327697753906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,8,32,power_law_1.01,0.13558143615722656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,8,32,power_law_1.01,0.14031999588012695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,8,32,power_law_1.01,0.14787967681884767
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,8,32,power_law_1.01,0.15964287757873535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,8,32,power_law_1.01,0.18035200119018555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,8,32,power_law_1.01,0.20317951202392578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,8,32,power_law_1.01,0.23041151046752928
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,8,32,power_law_1.01,0.2808281517028809
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,8,32,power_law_1.01,0.3660723114013672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,8,32,power_law_1.01,0.5356172943115235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,8,32,power_law_1.01,0.6356633758544922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,8,32,power_law_1.01,1.0086873626708983
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,8,32,power_law_1.01,0.042859520912170414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,8,32,power_law_1.01,0.03990911960601807
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,8,32,power_law_1.01,1.50785400390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,8,32,power_law_1.01,0.03991296052932739
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,8,32,power_law_1.01,0.04234367847442627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,8,32,power_law_1.01,0.04934656143188477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,8,32,power_law_1.01,0.05339903831481933
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,8,32,power_law_1.01,0.05097087860107422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,8,32,power_law_1.01,0.05793024063110351
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,8,32,power_law_1.01,0.05392000198364257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,8,32,power_law_1.01,0.05342591762542724
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,8,32,power_law_1.01,0.05642752170562744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,8,32,power_law_1.01,0.05227263927459717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,8,32,power_law_1.01,0.05794176101684571
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,8,32,power_law_1.01,0.06122111797332763
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,8,32,power_law_1.01,0.06046847820281982
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,8,32,power_law_1.01,0.06327936172485352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,8,32,power_law_1.01,0.06913407802581786
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,8,32,power_law_1.01,0.0783897590637207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,8,32,power_law_1.01,0.0855078411102295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,8,32,power_law_1.01,0.11022591590881348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,8,32,power_law_1.01,0.13899776458740234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,8,32,power_law_1.01,0.16226560592651368
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,8,32,power_law_1.01,0.19959039688110353
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,8,32,power_law_1.01,0.2745254325866699
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,8,32,power_law_1.01,0.3404966354370117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,8,32,power_law_1.01,0.48776065826416015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,8,32,power_law_1.01,0.02393728017807007
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,8,32,power_law_1.01,0.02370176076889038
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,8,32,power_law_1.01,0.7337586975097656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,8,32,balanced,0.023408639430999755
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,8,32,power_law_1.01,0.02396159887313843
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,8,32,power_law_1.01,0.023562240600585937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,8,32,power_law_1.01,0.023476479053497316
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,8,32,power_law_1.01,0.022740480899810792
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,8,32,power_law_1.01,0.023092479705810548
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,8,32,power_law_1.01,0.023028481006622314
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,8,32,power_law_1.01,0.02306432008743286
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,8,32,power_law_1.01,0.02311295986175537
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,8,32,power_law_1.01,0.02348927974700928
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,8,32,power_law_1.01,0.023889920711517333
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,8,32,power_law_1.01,0.023598079681396482
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,8,32,power_law_1.01,0.02348543882369995
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,8,32,power_law_1.01,0.023956480026245116
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,8,32,power_law_1.01,0.024005119800567624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,8,32,power_law_1.01,0.02476288080215454
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,8,32,balanced,0.02349695920944214
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,8,32,balanced,0.06539648056030274
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,8,32,balanced,0.038794240951538085
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,8,32,balanced,0.05591296195983887
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,8,32,balanced,0.042374401092529296
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,8,32,power_law_1.01,0.033137919902801516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,8,32,power_law_1.01,0.027559680938720705
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,8,32,power_law_1.01,0.030266880989074707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,8,32,power_law_1.2,0.06227200031280518
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,8,32,power_law_1.2,0.05963776111602783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,8,32,power_law_1.2,0.061550078392028806
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,8,32,power_law_1.01,0.04097152233123779
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,8,32,power_law_1.2,0.061941761970520016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,8,32,power_law_1.2,0.10029312133789063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,8,32,power_law_1.2,0.11558527946472168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,8,32,power_law_1.2,0.1144320011138916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,8,32,power_law_1.2,0.11560319900512694
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,8,32,power_law_1.2,0.11959808349609376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,8,32,power_law_1.2,0.12138751983642578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,8,32,power_law_1.2,0.12903039932250976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,8,32,power_law_1.2,0.13409279823303222
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,8,32,power_law_1.2,0.1381606388092041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,8,32,power_law_1.2,0.1381939220428467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,8,32,power_law_1.2,0.13822208404541017
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,8,32,power_law_1.2,0.1400486373901367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,8,32,power_law_1.2,0.14777983665466307
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,8,32,power_law_1.2,0.16809728622436523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,8,32,power_law_1.2,0.19714176177978515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,8,32,power_law_1.2,0.24951040267944338
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,8,32,power_law_1.2,0.2702592086791992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,8,32,power_law_1.2,0.3667046356201172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,8,32,power_law_1.2,0.43466751098632816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,8,32,power_law_1.2,0.6375667190551757
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,8,32,power_law_1.2,0.9157772827148438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,8,32,power_law_1.01,0.05074048042297363
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,8,32,power_law_1.01,0.06254464149475097
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,8,32,power_law_1.01,0.0895359992980957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,8,32,power_law_1.2,0.03910784006118774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,8,32,power_law_1.2,1.2020108795166016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,8,32,power_law_1.2,0.04008319854736328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,8,32,power_law_1.2,0.03953792095184326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,8,32,power_law_1.2,0.049180159568786616
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,8,32,power_law_1.2,0.05059455871582032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,8,32,power_law_1.2,0.05392000198364257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,8,32,power_law_1.2,0.05219967842102051
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,8,32,power_law_1.2,0.05256319999694824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,8,32,power_law_1.2,0.05216127872467041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,8,32,power_law_1.2,0.05427840232849122
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,8,32,power_law_1.2,1.6437709045410156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,8,32,power_law_1.2,0.06191359996795655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,8,32,power_law_1.2,0.05780608177185058
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,8,32,power_law_1.2,0.061655039787292476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,8,32,power_law_1.2,0.06377471923828125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,8,32,power_law_1.2,0.06680319786071778
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,8,32,power_law_1.2,0.06944640159606934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,8,32,power_law_1.2,0.04028800010681152
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,8,32,power_law_1.2,0.08166144371032716
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,8,32,power_law_1.2,0.09220479965209961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,8,32,power_law_1.2,0.11283455848693849
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,8,32,power_law_1.01,0.09818367958068848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,8,32,power_law_1.2,0.1493235206604004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,8,32,power_law_1.2,0.19748992919921876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,8,32,power_law_1.2,0.2802009582519531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,8,32,power_law_1.2,0.3682099151611328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,8,32,power_law_1.2,0.023162879943847657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,8,32,power_law_1.2,0.023712000846862792
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,8,32,power_law_1.2,0.02354687929153442
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,8,32,power_law_1.2,0.023146240711212157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,8,32,power_law_1.2,0.534073600769043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,8,32,power_law_1.2,0.02278656005859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,8,32,power_law_1.2,0.022649600505828857
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,8,32,power_law_1.2,0.023393280506134033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,8,32,power_law_1.2,0.8339507293701172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,8,32,power_law_1.2,0.023074560165405274
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,8,32,power_law_1.2,0.02305919885635376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,8,32,power_law_1.2,1.0031961822509765
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,8,32,power_law_1.2,0.023119359016418456
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,8,32,power_law_1.2,0.023110399246215822
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,8,32,power_law_1.2,0.023636479377746582
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,8,32,power_law_1.2,0.023641600608825683
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,8,32,power_law_1.2,0.024376320838928222
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,8,32,power_law_1.2,0.025917439460754393
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,8,32,power_law_1.2,0.028323841094970704
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,8,32,power_law_1.2,0.023388159275054932
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,8,32,power_law_1.2,0.023559679985046385
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,8,32,power_law_1.01,0.1491865634918213
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,8,32,power_law_1.01,0.18272895812988282
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,8,32,power_law_1.2,0.033077759742736815
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,8,32,power_law_1.2,0.05023231983184815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,16,16,balanced,0.03429888010025024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,16,16,balanced,0.034206719398498536
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,16,16,balanced,0.03300479888916016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,8,32,power_law_1.2,0.06263040065765381
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,16,16,balanced,0.03541376113891602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,16,16,balanced,0.038502399921417234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,8,32,power_law_1.2,0.07932415962219239
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,16,16,balanced,0.06934656143188477
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,8,32,power_law_1.2,0.041637120246887205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,16,16,balanced,0.12749695777893066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,16,16,balanced,0.12288895606994629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,16,16,balanced,0.12625151634216308
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,16,16,balanced,0.12408576011657715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,16,16,balanced,0.13307135581970214
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,16,16,balanced,0.13855744361877442
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,8,32,power_law_1.2,0.11872639656066894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,16,16,balanced,0.14180607795715333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,16,16,balanced,0.13500672340393066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,16,16,balanced,0.1410905647277832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,16,16,balanced,0.14495231628417968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,16,16,balanced,0.15352959632873536
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,16,16,balanced,0.16417407989501953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,16,16,balanced,0.17658880233764648
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,16,16,balanced,0.20641408920288087
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,16,16,balanced,0.23766912460327147
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,16,16,balanced,0.2943552017211914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,16,16,balanced,0.3539161682128906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,16,16,balanced,0.034200320243835444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,16,16,balanced,0.4710579299926758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,16,16,balanced,0.5837939071655274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,16,16,balanced,0.03404671907424926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,16,16,balanced,0.03665152072906494
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,16,16,balanced,0.03865087985992431
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,16,16,balanced,0.04330751895904541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,16,16,balanced,0.04849408149719238
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,16,16,balanced,0.8103526306152343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,16,16,balanced,0.06577023983001709
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,16,16,balanced,0.06122879981994629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,16,16,balanced,0.06946303844451904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,16,16,balanced,0.06134655952453614
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,16,16,balanced,0.06263936042785645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,16,16,balanced,1.0397862243652345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,16,16,balanced,0.06449664115905762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,16,16,balanced,0.06647039890289307
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,16,16,balanced,0.06372608184814453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,16,16,balanced,0.06754176139831543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,16,16,balanced,0.07038976192474365
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,16,16,balanced,0.07683072090148926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,16,16,balanced,0.0830361557006836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,16,16,balanced,0.09113216400146484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,16,16,balanced,0.10953472137451173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,16,16,balanced,0.1283289623260498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,16,16,balanced,0.16436479568481446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,16,16,balanced,0.1955392074584961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,16,16,balanced,0.2817318344116211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,16,16,balanced,0.3464947128295898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,16,16,balanced,0.49513599395751956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,16,16,balanced,0.6434854125976562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,8,32,power_law_1.2,0.14456704139709473
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,8,32,power_law_1.2,0.2263180732727051
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,8,32,power_law_1.2,0.28551807403564455
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,16,16,balanced,0.020636160373687745
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,16,16,balanced,0.020776960849761963
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,16,16,balanced,0.02879744052886963
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,16,16,balanced,0.02880000114440918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,16,16,balanced,0.02964992046356201
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,16,16,balanced,0.021064960956573488
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,16,16,balanced,0.029053440093994142
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,16,16,balanced,0.02973695993423462
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,16,16,balanced,0.030120959281921388
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,16,16,balanced,0.023071999549865725
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,16,16,balanced,0.029694719314575197
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,16,16,balanced,0.030042879581451416
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,16,16,balanced,0.029621760845184326
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,16,16,balanced,0.03191296100616455
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,16,16,balanced,0.03175551891326904
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,16,16,balanced,0.03398144006729126
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,16,16,balanced,0.03737087965011597
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,16,16,balanced,0.039470078945159914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,16,16,balanced,0.04173312187194824
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,16,16,balanced,0.024422399997711182
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,16,16,power_law_1.01,0.034561278820037844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,16,16,power_law_1.01,0.04658304214477539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,16,16,power_law_1.01,0.0537446403503418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,16,16,power_law_1.01,0.06015615940093995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,16,16,power_law_1.01,0.06421376228332519
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,16,16,power_law_1.01,0.11542911529541015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,16,16,power_law_1.01,0.11723008155822753
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,16,16,power_law_1.01,0.11812352180480956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,16,16,power_law_1.01,0.12340991973876954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,16,16,power_law_1.01,0.12944640159606932
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,16,16,power_law_1.01,0.13315327644348146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,16,16,power_law_1.01,0.1373030376434326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,16,16,power_law_1.01,0.14057600021362304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,16,16,power_law_1.01,0.13616127967834474
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,16,16,power_law_1.01,0.14319231986999512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,16,16,power_law_1.01,0.1464857578277588
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,16,16,power_law_1.01,0.15400575637817382
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,16,16,power_law_1.01,0.17288576126098634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,16,16,power_law_1.01,0.2006015968322754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,16,16,power_law_1.01,0.2502566337585449
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,16,16,power_law_1.01,0.26887552261352543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,16,16,power_law_1.01,0.34863361358642575
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,16,16,balanced,0.02961663961410522
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,16,16,power_law_1.01,0.4314726257324219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,16,16,power_law_1.01,0.596049919128418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,16,16,power_law_1.01,0.7419123077392579
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,16,16,power_law_1.01,1.0656230163574218
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,16,16,balanced,0.028866560459136964
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,16,16,power_law_1.01,0.03873280048370361
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,16,16,power_law_1.01,0.046813440322875974
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,16,16,power_law_1.01,0.04619264125823975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,16,16,power_law_1.01,0.04860032081604004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,16,16,power_law_1.01,0.048037118911743164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,16,16,power_law_1.01,0.061631999015808105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,16,16,power_law_1.01,0.05814655780792236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,16,16,power_law_1.01,0.06056447982788086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,16,16,power_law_1.01,0.05989120006561279
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,16,16,power_law_1.01,0.059275522232055664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,16,16,power_law_1.01,0.06254079818725586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,16,16,power_law_1.01,0.06390399932861328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,16,16,power_law_1.01,0.07523071765899658
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,16,16,power_law_1.01,0.06688640117645264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,16,16,power_law_1.01,0.07102975845336915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,16,16,power_law_1.01,0.07308671951293946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,16,16,power_law_1.01,0.07558144092559814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,16,16,power_law_1.01,0.09262463569641113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,16,16,power_law_1.01,0.09772416114807128
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,16,16,power_law_1.01,0.12941568374633788
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,16,16,power_law_1.01,0.16542207717895507
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,16,16,power_law_1.01,1.4336306762695312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,16,16,power_law_1.01,0.2060032081604004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,16,16,power_law_1.01,0.25256959915161137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,16,16,power_law_1.01,0.3349132919311523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,16,16,power_law_1.01,0.4784140777587891
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,16,16,power_law_1.01,0.024351999759674073
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,16,16,power_law_1.01,0.6535628509521485
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,16,16,power_law_1.01,0.029269759654998777
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,16,16,power_law_1.01,0.029163520336151123
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,16,16,power_law_1.01,0.0288102388381958
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,16,16,power_law_1.01,0.9573632049560548
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,16,16,power_law_1.01,0.029528319835662842
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,16,16,power_law_1.01,0.029285120964050292
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,16,16,power_law_1.01,0.02742271900177002
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,16,16,power_law_1.01,0.028552958965301512
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,16,16,power_law_1.01,0.028474879264831544
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,16,16,power_law_1.01,0.027969279289245606
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,16,16,power_law_1.01,0.02925184011459351
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,16,16,power_law_1.01,0.028770558834075928
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,16,16,power_law_1.01,0.029291520118713377
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,16,16,power_law_1.01,0.029196801185607912
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,16,16,power_law_1.01,0.03044735908508301
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,16,16,power_law_1.01,0.031366400718688965
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,16,16,power_law_1.01,0.03278464078903198
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,16,16,balanced,0.028720641136169435
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,16,16,balanced,0.060026879310607906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,16,16,balanced,0.0655731201171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,16,16,balanced,0.08305024147033692
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,16,16,balanced,0.10117376327514649
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,16,16,power_law_1.01,0.03739903926849365
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,16,16,power_law_1.01,0.042154240608215335
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,16,16,power_law_1.01,0.04977536201477051
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,16,16,power_law_1.2,0.035791358947753905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,16,16,power_law_1.2,0.053757438659667964
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,16,16,power_law_1.2,0.055801601409912104
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,16,16,power_law_1.2,0.059726080894470214
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,16,16,power_law_1.2,0.06417791843414307
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,16,16,power_law_1.2,0.11672831535339356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,16,16,power_law_1.2,0.12646016120910644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,16,16,power_law_1.2,0.12623744010925292
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,16,16,power_law_1.2,0.12532735824584962
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,16,16,power_law_1.2,0.12850815773010255
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,16,16,power_law_1.2,0.13359616279602052
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,16,16,power_law_1.2,0.14203392028808595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,16,16,power_law_1.2,0.1467955207824707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,16,16,power_law_1.2,0.13975808143615723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,16,16,power_law_1.2,0.1459712028503418
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,16,16,power_law_1.01,0.061975040435791016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,16,16,power_law_1.2,0.15026687622070312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,16,16,power_law_1.2,0.15828736305236818
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,16,16,power_law_1.2,0.1859903907775879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,16,16,power_law_1.2,0.19789184570312499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,16,16,power_law_1.2,0.2501043128967285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,16,16,power_law_1.2,0.28436479568481443
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,16,16,power_law_1.2,0.3730316925048828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,16,16,power_law_1.01,0.08411775588989258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,16,16,power_law_1.01,0.09951359748840331
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,16,16,power_law_1.2,0.4789542388916016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,16,16,power_law_1.01,0.12996479988098145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,16,16,power_law_1.2,0.5952550506591796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,16,16,power_law_1.2,0.039263999462127684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,16,16,power_law_1.2,0.04559743881225586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,16,16,power_law_1.2,0.8928793334960938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,16,16,power_law_1.2,0.045519361495971675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,16,16,power_law_1.2,0.047319040298461915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,16,16,power_law_1.2,0.047667198181152344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,16,16,power_law_1.2,0.06166783809661865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,16,16,power_law_1.2,0.062048001289367674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,16,16,power_law_1.2,0.05874944210052491
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,16,16,power_law_1.2,1.5034675598144531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,16,16,power_law_1.2,0.06082943916320801
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,16,16,power_law_1.2,0.0592793607711792
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,16,16,power_law_1.2,0.0612659215927124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,16,16,power_law_1.2,0.06638847827911377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,16,16,power_law_1.2,1.6640229797363282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,16,16,power_law_1.2,0.06947968006134034
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,16,16,power_law_1.2,0.07286528110504151
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,16,16,power_law_1.2,0.07600384235382081
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,16,16,power_law_1.2,0.07968895912170411
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,16,16,power_law_1.2,0.09415936470031738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,16,16,power_law_1.2,0.11554944038391113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,16,16,power_law_1.2,0.14188672065734864
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,16,16,power_law_1.2,0.16091520309448243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,16,16,power_law_1.2,0.22596864700317382
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,16,16,power_law_1.2,0.2858393669128418
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,16,16,power_law_1.01,0.1628544044494629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,16,16,power_law_1.2,0.06822656154632568
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,16,16,power_law_1.2,0.4630464172363281
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,16,16,power_law_1.2,0.024364800453186036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,16,16,power_law_1.2,0.5704742431640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,16,16,power_law_1.2,0.0288703989982605
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,16,16,power_law_1.2,0.028806400299072266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,16,16,power_law_1.2,0.02884864091873169
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,16,16,power_law_1.2,0.029216001033782957
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,16,16,power_law_1.2,0.027285759449005124
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,16,16,power_law_1.2,0.028491520881652833
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,16,16,power_law_1.2,0.028862719535827634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,16,16,power_law_1.2,0.9192947387695313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,16,16,power_law_1.2,0.028107519149780276
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,16,16,power_law_1.2,0.028195838928222656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,16,16,power_law_1.2,0.02824320077896118
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,16,16,power_law_1.2,0.028737280368804928
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,16,16,power_law_1.2,1.0834291076660156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,16,16,power_law_1.2,0.029762558937072754
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,16,16,power_law_1.2,0.03218048095703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,16,16,power_law_1.2,0.031777279376983644
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,16,16,power_law_1.2,0.041762561798095704
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,16,16,power_law_1.2,0.04647295951843262
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,16,16,power_law_1.2,0.02891648054122925
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,16,16,power_law_1.01,0.24409727096557615
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,16,16,power_law_1.01,0.3181670379638672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,16,16,power_law_1.2,0.036034560203552245
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,16,16,power_law_1.2,0.06162176132202148
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,16,16,power_law_1.2,0.07573760032653809
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,16,16,power_law_1.2,0.10242815971374512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,32,8,balanced,0.03867392063140869
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,16,16,power_law_1.2,0.13468671798706056
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,32,8,balanced,0.033775360584259034
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,32,8,balanced,0.03703167915344238
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,32,8,balanced,0.0399014401435852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,32,8,balanced,0.055311360359191895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,32,8,balanced,0.08372480392456054
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,32,8,balanced,0.13961471557617186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,32,8,balanced,0.13826047897338867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,32,8,balanced,0.14381440162658693
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,32,8,balanced,0.15152000427246093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,32,8,balanced,0.14478464126586915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,32,8,balanced,0.15140735626220703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,32,8,balanced,0.15515263557434084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,32,8,balanced,0.1630988883972168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,32,8,balanced,0.1656051254272461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,32,8,balanced,0.17311744689941405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,32,8,balanced,0.1853606414794922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,32,8,balanced,0.21640064239501955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,32,8,balanced,0.2443008041381836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,32,8,balanced,0.301136646270752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,32,8,balanced,0.35854209899902345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,32,8,balanced,0.4745075225830078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,32,8,balanced,0.5852236938476563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,16,16,power_law_1.2,0.18113407135009765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,32,8,balanced,0.8055321502685547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,32,8,balanced,0.034204161167144774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,32,8,balanced,1.027110366821289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,32,8,balanced,0.042849278450012206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,32,8,balanced,0.048555521965026854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,32,8,balanced,0.051837439537048335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,32,8,balanced,0.07103360176086426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,32,8,balanced,0.07719039916992188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,32,8,balanced,0.07881216049194337
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,32,8,balanced,1.4802149963378906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,32,8,balanced,0.08080512046813965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,32,8,balanced,0.08305279731750488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,32,8,balanced,0.037914879322052
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,32,8,balanced,0.07868544101715089
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,32,8,balanced,0.082543363571167
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,32,8,balanced,1.9265945434570313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,32,8,balanced,0.0850227165222168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,32,8,balanced,0.09124735832214356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,32,8,balanced,0.09217408180236816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,32,8,balanced,0.0978444766998291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,32,8,balanced,0.10542976379394531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,32,8,balanced,0.12440832138061524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,32,8,balanced,0.14198911666870118
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,32,8,balanced,0.17720575332641603
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,32,8,balanced,0.21292415618896485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,32,8,balanced,0.283000316619873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,32,8,balanced,0.34732288360595703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,32,8,balanced,0.5077350234985352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,32,8,balanced,0.6414514923095703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,32,8,balanced,0.9338227081298829
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,32,8,balanced,1.223631362915039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,16,16,power_law_1.2,0.2231155204772949
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,16,16,power_law_1.2,0.35189888000488284
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,16,16,power_law_1.2,0.4546956634521484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,32,8,balanced,0.023589119911193848
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,32,8,balanced,0.024261119365692137
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,32,8,balanced,0.0415718412399292
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,32,8,balanced,0.025582079887390134
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,32,8,balanced,0.04196479797363281
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,32,8,balanced,0.04223231792449951
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,32,8,balanced,0.04306560039520264
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,32,8,balanced,0.04251008033752442
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,32,8,balanced,0.04290944099426269
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,32,8,balanced,0.042768640518188475
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,32,8,balanced,0.04343039989471435
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,32,8,balanced,0.04542079925537109
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,32,8,balanced,0.04704127788543701
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,32,8,balanced,0.04865920066833496
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,32,8,balanced,0.053208317756652836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,32,8,balanced,0.03050368070602417
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,32,8,balanced,0.055888638496398926
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,32,8,balanced,0.06236415863037109
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,32,8,balanced,0.0680947208404541
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,32,8,balanced,0.027258880138397217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,32,8,power_law_1.01,0.037876479625701905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,32,8,power_law_1.01,0.039508481025695805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,32,8,power_law_1.01,0.056212477684021
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,32,8,power_law_1.01,0.0623308801651001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,32,8,power_law_1.01,0.06980607986450196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,32,8,power_law_1.01,0.07113088130950927
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,32,8,power_law_1.01,0.1358899211883545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,32,8,power_law_1.01,0.13799936294555665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,32,8,power_law_1.01,0.13965184211730958
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,32,8,power_law_1.01,0.14237055778503419
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,32,8,power_law_1.01,0.1520345592498779
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,32,8,power_law_1.01,0.14940799713134764
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,32,8,power_law_1.01,0.14948224067687987
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,32,8,power_law_1.01,0.15374336242675782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,32,8,power_law_1.01,0.15859328269958498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,32,8,power_law_1.01,0.16373376846313475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,32,8,power_law_1.01,0.17420671463012696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,32,8,power_law_1.01,0.20329216003417966
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,32,8,power_law_1.01,0.22711936950683595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,32,8,power_law_1.01,0.26105728149414065
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,32,8,balanced,0.04134016036987305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,32,8,balanced,0.041140480041503905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,32,8,power_law_1.01,0.3282099151611328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,32,8,power_law_1.01,0.41343742370605463
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,32,8,power_law_1.01,0.5205785751342773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,32,8,power_law_1.01,0.7291149139404297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,32,8,power_law_1.01,0.9612032318115235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,32,8,power_law_1.01,0.039061760902404784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,32,8,power_law_1.01,1.3689573669433595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,32,8,power_law_1.01,0.04635647773742676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,32,8,power_law_1.01,0.05174784183502197
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,32,8,power_law_1.01,1.6588697814941404
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,32,8,power_law_1.01,0.0530406379699707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,32,8,power_law_1.01,0.05674111843109131
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,32,8,power_law_1.01,0.05673855781555176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,32,8,power_law_1.01,0.07484543800354004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,32,8,power_law_1.01,0.0743616008758545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,32,8,power_law_1.01,0.07512959957122803
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,32,8,power_law_1.01,0.07680255889892577
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,32,8,power_law_1.01,0.08531968116760254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,32,8,power_law_1.01,0.08307456016540528
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,32,8,power_law_1.01,0.07879168033599854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,32,8,power_law_1.01,0.08479743957519531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,32,8,power_law_1.01,0.08608639717102051
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,32,8,power_law_1.01,0.08832639694213867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,32,8,power_law_1.01,0.09538047790527344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,32,8,power_law_1.01,0.11137663841247558
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,32,8,power_law_1.01,0.12729855537414553
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,32,8,power_law_1.01,0.16608640670776367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,32,8,power_law_1.01,0.20132352828979494
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,32,8,power_law_1.01,0.2604915237426758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,32,8,power_law_1.01,0.3460812759399414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,32,8,power_law_1.01,0.4070924758911133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,32,8,power_law_1.01,0.6248332977294921
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,32,8,power_law_1.01,0.8569548797607421
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,32,8,power_law_1.01,0.026853120326995848
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,32,8,power_law_1.01,0.030051839351654053
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,32,8,power_law_1.01,1.0980172729492188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,32,8,power_law_1.01,0.034288640022277835
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,32,8,power_law_1.01,0.03544064044952393
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,32,8,power_law_1.01,0.03838335990905762
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,32,8,power_law_1.01,0.03901695966720581
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,32,8,power_law_1.01,0.04101759910583496
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,32,8,power_law_1.01,0.038869760036468505
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,32,8,power_law_1.01,0.03891968011856079
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,32,8,power_law_1.01,0.039028480052948
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,32,8,power_law_1.01,0.039453439712524414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,32,8,power_law_1.01,0.04028672218322754
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,32,8,power_law_1.01,0.042000641822814946
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,32,8,power_law_1.01,0.04383615970611572
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,32,8,power_law_1.01,0.04529280185699463
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,32,8,power_law_1.01,0.049443840980529785
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,32,8,power_law_1.01,0.046268157958984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,32,8,balanced,0.04117119789123535
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,32,8,balanced,0.10828800201416015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,32,8,balanced,0.15032447814941405
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,32,8,balanced,0.09579008102416993
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,32,8,balanced,0.19268352508544923
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,32,8,power_law_1.01,0.06860928058624267
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,32,8,power_law_1.01,0.08720255851745605
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,32,8,power_law_1.01,0.05638912200927735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,32,8,power_law_1.2,0.038440959453582765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,32,8,power_law_1.2,0.038297600746154785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,32,8,power_law_1.2,0.04816895961761475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,32,8,power_law_1.2,0.05932415962219238
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,32,8,power_law_1.2,0.06806399822235107
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,32,8,power_law_1.2,0.07127552032470703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,32,8,power_law_1.2,0.13501824378967284
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,32,8,power_law_1.2,0.13505791664123534
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,32,8,power_law_1.2,0.13704575538635252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,32,8,power_law_1.2,0.14327168464660645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,32,8,power_law_1.2,0.1514572811126709
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,32,8,power_law_1.2,0.14703231811523437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,32,8,power_law_1.2,0.147390718460083
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,32,8,power_law_1.2,0.15447039604187013
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,32,8,power_law_1.2,0.16039552688598632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,32,8,power_law_1.2,0.16391935348510742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,32,8,power_law_1.2,0.17937152862548827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,32,8,power_law_1.2,0.20170751571655274
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,32,8,power_law_1.01,0.1076863956451416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,32,8,power_law_1.2,0.22867200851440428
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,32,8,power_law_1.2,0.29155712127685546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,32,8,power_law_1.2,0.33909759521484373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,32,8,power_law_1.2,0.4339302444458008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,32,8,power_law_1.2,0.5657932662963867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,32,8,power_law_1.2,0.7542156982421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,32,8,power_law_1.01,0.13494784355163575
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,32,8,power_law_1.01,0.15991040229797365
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,32,8,power_law_1.2,0.9720038604736327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,32,8,power_law_1.2,0.04239232063293457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,32,8,power_law_1.2,0.04623616218566894
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,32,8,power_law_1.01,0.29036415100097657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,32,8,power_law_1.2,0.04895359992980957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,32,8,power_law_1.2,0.05244800090789795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,32,8,power_law_1.2,1.3407334899902343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,32,8,power_law_1.2,0.05715968132019043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,32,8,power_law_1.2,0.06061952114105225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,32,8,power_law_1.2,0.07439743995666503
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,32,8,power_law_1.01,0.23440895080566407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,32,8,power_law_1.2,0.07476223945617676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,32,8,power_law_1.2,0.07545087814331056
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,32,8,power_law_1.2,0.07889920234680176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,32,8,power_law_1.2,0.08018303871154785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,32,8,power_law_1.2,0.0832153606414795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,32,8,power_law_1.2,0.08528384208679199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,32,8,power_law_1.2,0.08750335693359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,32,8,power_law_1.2,0.08881664276123047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,32,8,power_law_1.2,0.09705984115600585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,32,8,power_law_1.2,0.1166438388824463
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,32,8,power_law_1.2,1.8579443359374999
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,32,8,power_law_1.2,0.13185791969299315
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,32,8,power_law_1.2,0.1750271987915039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,32,8,power_law_1.2,0.2133900833129883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,32,8,power_law_1.2,0.29209983825683594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,32,8,power_law_1.2,0.07062528133392335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,32,8,power_law_1.2,0.36703487396240236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,32,8,power_law_1.2,0.5235942459106445
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,32,8,power_law_1.2,0.026762239933013916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,32,8,power_law_1.2,0.030088961124420166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,32,8,power_law_1.2,0.6102566528320312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,32,8,power_law_1.2,0.0318067193031311
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,32,8,power_law_1.2,0.035075840950012205
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,32,8,power_law_1.2,0.038638079166412355
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,32,8,power_law_1.2,0.040376319885253906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,32,8,power_law_1.2,0.039749119281768795
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,32,8,power_law_1.2,0.0376140809059143
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,32,8,power_law_1.2,0.03751168012619018
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,32,8,power_law_1.2,0.994909439086914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,32,8,power_law_1.2,0.03790976047515869
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,32,8,power_law_1.2,0.04147327899932861
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,32,8,power_law_1.2,0.039750399589538576
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,32,8,power_law_1.2,0.042133760452270505
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,32,8,power_law_1.2,0.043276801109313964
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,32,8,power_law_1.2,0.045610241889953614
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,32,8,power_law_1.2,0.04785535812377929
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,32,8,power_law_1.2,1.3694157409667969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,32,8,power_law_1.2,0.053835520744323725
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,32,8,power_law_1.2,0.06173056125640869
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,32,8,power_law_1.01,0.40648574829101564
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,32,8,power_law_1.01,0.5501657485961914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,32,8,power_law_1.2,0.13613951683044434
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,32,8,power_law_1.2,0.21124479293823245
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,256,balanced,0.03739007949829102
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,32,8,power_law_1.2,0.10667391777038575
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,256,balanced,0.03623680114746094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,256,balanced,0.03546495914459229
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,256,balanced,0.037039361000061034
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,256,balanced,0.035846400260925296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,256,balanced,0.037080318927764894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,256,balanced,0.0651033592224121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,256,balanced,0.06653823852539062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,256,balanced,0.06998400211334228
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,32,8,power_law_1.2,0.07730559825897217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,256,balanced,0.06822656154632568
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,256,balanced,0.11117183685302734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,256,balanced,0.0859596824645996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,256,balanced,0.11348863601684571
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,256,balanced,0.08495488166809081
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,256,balanced,0.11521535873413087
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,256,balanced,0.11051136016845704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,256,balanced,0.1199884796142578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,256,balanced,0.11707391738891601
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,256,balanced,0.1225126361846924
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,32,8,power_law_1.2,0.2997875213623047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,32,8,power_law_1.2,0.17214975357055665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,256,balanced,0.07501567840576172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,256,balanced,0.1220633602142334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,256,balanced,0.1344204807281494
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,256,balanced,0.13481856346130372
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,256,balanced,0.14923775672912598
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,256,balanced,0.17768064498901368
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,256,balanced,0.19799039840698243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,256,balanced,0.03587455987930298
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,256,balanced,0.03593600034713745
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,256,balanced,0.035357439517974855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,256,balanced,0.035715839862823485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,256,balanced,0.03625087976455689
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,256,balanced,0.03549439907073974
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,256,balanced,0.04486656188964844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,256,balanced,0.0448089599609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,256,balanced,0.0454963207244873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,256,balanced,0.04509696006774903
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,256,balanced,0.04523392200469971
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,256,balanced,0.04568192005157471
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,256,balanced,0.046589441299438476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,256,balanced,0.045537281036376956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,256,balanced,0.046846718788146974
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,256,balanced,0.04732416152954101
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,256,balanced,0.05060480117797852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,256,balanced,0.04770944118499756
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,256,balanced,0.05264895915985107
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,256,balanced,0.04849279880523681
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,256,balanced,0.1591423988342285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,256,balanced,0.055463681221008296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,256,balanced,0.05420671939849854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,256,balanced,0.06501760005950927
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,256,balanced,0.06947711944580079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,256,balanced,0.08218496322631837
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,256,balanced,0.09669247627258301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,256,balanced,0.05071743965148926
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,32,8,power_law_1.2,0.4334668731689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,32,8,power_law_1.2,0.7394854736328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,32,8,power_law_1.2,0.8309811401367189
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,1,256,balanced,0.018581759929656983
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,1,256,balanced,0.018177280426025393
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,1,256,balanced,0.018150399923324585
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,1,256,balanced,0.018978559970855714
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,1,256,balanced,0.01857151985168457
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,1,256,balanced,0.01855487942695618
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,1,256,balanced,0.018570239543914797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,1,256,balanced,0.01819007992744446
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,1,256,balanced,0.0189683198928833
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,1,256,balanced,0.018575359582901
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,1,256,balanced,0.018636800050735473
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,1,256,balanced,0.01853440046310425
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,1,256,balanced,0.01897984027862549
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,1,256,balanced,0.01899392008781433
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,1,256,balanced,0.018873599767684938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,1,256,balanced,0.018714879751205442
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,1,256,balanced,0.018429440259933472
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,1,256,balanced,0.018650879859924314
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,1,256,balanced,0.01817471981048584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,1,256,balanced,0.019686399698257445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,256,power_law_1.01,0.03707007884979248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,256,power_law_1.01,0.037470719814300536
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,256,power_law_1.01,0.036599040031433105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,256,power_law_1.01,0.03700223922729492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,256,power_law_1.01,0.0727564811706543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,256,power_law_1.01,0.08849663734436035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,256,power_law_1.01,0.08825216293334961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,256,power_law_1.01,0.07223552227020263
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,256,power_law_1.01,0.07809919834136962
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,256,power_law_1.01,0.08301568031311035
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,1,256,balanced,0.018414080142974854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,256,power_law_1.01,0.08627840042114257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,256,power_law_1.01,0.10514431953430177
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,256,power_law_1.01,0.09702143669128419
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,256,power_law_1.01,0.11513855934143066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,256,power_law_1.01,0.1059507179260254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,256,power_law_1.01,0.12089471817016602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,256,power_law_1.01,0.13742719650268556
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,256,power_law_1.01,0.1415347194671631
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,256,power_law_1.01,0.15136256217956542
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,1,256,balanced,0.018570239543914797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,256,power_law_1.01,0.16793472290039063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,256,power_law_1.01,0.1972991943359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,256,power_law_1.01,0.2538163185119629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,256,power_law_1.01,0.330511360168457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,256,power_law_1.01,0.502999038696289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,256,power_law_1.01,0.6607756805419922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,256,power_law_1.01,0.03544064044952393
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,256,power_law_1.01,0.03551232099533081
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,256,power_law_1.01,1.0731597137451172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,256,power_law_1.01,0.03543040037155151
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,256,power_law_1.01,0.035777280330657954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,256,power_law_1.01,1.3047628784179688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,256,power_law_1.01,0.04483839988708496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,256,power_law_1.01,0.04568192005157471
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,256,power_law_1.01,0.046845440864562986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,256,power_law_1.01,0.0456115198135376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,256,power_law_1.01,0.046049280166625975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,256,power_law_1.01,0.04893824100494385
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,256,power_law_1.01,0.046973438262939454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,256,power_law_1.01,0.047740159034729
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,256,power_law_1.01,0.05428864002227783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,256,power_law_1.01,0.0481382417678833
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,256,power_law_1.01,0.04958079814910889
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,256,power_law_1.01,0.051143679618835446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,256,power_law_1.01,0.05507328033447265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,256,power_law_1.01,0.06100096225738525
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,256,power_law_1.01,0.06838528156280518
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,256,power_law_1.01,0.07813248157501221
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,256,power_law_1.01,0.09064448356628418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,256,power_law_1.01,0.1157926368713379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,256,power_law_1.01,0.1546009635925293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,256,power_law_1.01,0.21745920181274414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,256,power_law_1.01,0.30286464691162107
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,256,power_law_1.01,0.460423698425293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,256,power_law_1.01,0.5893964767456055
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,1,256,power_law_1.01,0.018165760040283203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,1,256,power_law_1.01,0.01817855954170227
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,1,256,power_law_1.01,0.01853440046310425
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,1,256,power_law_1.01,0.018579200506210328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,1,256,power_law_1.01,0.018353919982910156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,1,256,power_law_1.01,0.01818112015724182
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,1,256,power_law_1.01,0.018570239543914797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,1,256,power_law_1.01,0.01860352039337158
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,1,256,power_law_1.01,0.01856768012046814
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,1,256,power_law_1.01,0.01821568012237549
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,1,256,power_law_1.01,0.01894528031349182
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,1,256,power_law_1.01,0.01855487942695618
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,1,256,power_law_1.01,0.01863039970397949
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,1,256,power_law_1.01,0.01888383984565735
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,1,256,power_law_1.01,0.018983680009841918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,1,256,power_law_1.01,0.018640639781951903
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,1,256,power_law_1.01,0.019133440256118774
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,1,256,balanced,0.018380800485610964
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,1,256,balanced,0.022329599857330323
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,1,256,balanced,0.02070784091949463
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,1,256,balanced,0.025606400966644287
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,1,256,balanced,0.02099967956542969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,1,256,power_law_1.01,0.02069888114929199
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,1,256,power_law_1.01,0.022238719463348388
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,1,256,power_law_1.01,0.019845119714736938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,1,256,power_law_1.2,0.037419519424438476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,1,256,power_law_1.2,0.036692481040954586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,1,256,power_law_1.2,0.03665791988372803
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,1,256,power_law_1.2,0.06542719841003418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,1,256,power_law_1.2,0.06739583969116211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,1,256,power_law_1.2,0.08017408370971679
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,1,256,power_law_1.2,0.0773311996459961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,1,256,power_law_1.2,0.07886591911315918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,1,256,power_law_1.2,0.08603263854980468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,1,256,power_law_1.2,0.09250304222106934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,1,256,power_law_1.2,0.09314432144165039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,1,256,power_law_1.2,0.1151039981842041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,1,256,power_law_1.2,0.11519488334655761
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,1,256,power_law_1.2,0.1158950424194336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,1,256,power_law_1.2,0.13041024208068847
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,1,256,power_law_1.2,0.1376204776763916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,1,256,power_law_1.2,0.1427443218231201
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,1,256,power_law_1.2,0.15647104263305664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,1,256,power_law_1.2,0.1698636817932129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,1,256,power_law_1.2,0.2145894432067871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,1,256,power_law_1.2,0.2601087951660156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,1,256,power_law_1.01,0.025189120769500733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,1,256,power_law_1.2,0.4232896041870117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,1,256,power_law_1.2,0.5348019027709962
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,1,256,power_law_1.2,0.8650688171386719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,1,256,power_law_1.2,1.1811558532714845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,1,256,power_law_1.2,2.1051315307617187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,1,256,power_law_1.2,0.03596159934997559
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,1,256,power_law_1.2,0.03575551986694336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,1,256,power_law_1.01,0.03395711898803711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,1,256,power_law_1.2,3.2284286499023436
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,1,256,power_law_1.01,0.040650238990783696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,1,256,power_law_1.2,0.03632128000259399
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,1,256,power_law_1.2,0.045053439140319826
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,1,256,power_law_1.01,0.053958401679992676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,1,256,power_law_1.2,0.04503039836883545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,1,256,power_law_1.2,0.04730879783630371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,1,256,power_law_1.2,0.04571135997772217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,1,256,power_law_1.2,0.04724224090576172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,1,256,power_law_1.2,0.04790527820587158
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,1,256,power_law_1.2,0.05026048183441162
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,1,256,power_law_1.2,0.05102591991424561
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,1,256,power_law_1.2,0.047735037803649905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,1,256,power_law_1.2,0.049835519790649416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,1,256,power_law_1.2,0.0519052791595459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,1,256,power_law_1.2,0.05443327903747559
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,1,256,power_law_1.2,0.058348798751831056
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,1,256,power_law_1.2,0.07033728122711182
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,1,256,power_law_1.2,0.07886079788208009
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,1,256,power_law_1.2,0.09608192443847656
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,1,256,power_law_1.01,0.07127039909362794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,1,256,power_law_1.2,0.13033856391906737
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,1,256,power_law_1.2,0.19936895370483398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,1,256,power_law_1.2,0.2448217582702637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,1,256,power_law_1.2,0.04772863864898681
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,1,256,power_law_1.2,0.40013694763183594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,1,256,power_law_1.2,0.01823232054710388
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,1,256,power_law_1.2,0.018638080358505248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,1,256,power_law_1.2,0.58115966796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,1,256,power_law_1.2,0.018350080251693726
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,1,256,power_law_1.2,0.01860607981681824
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,1,256,power_law_1.2,0.01834879994392395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,1,256,power_law_1.2,0.018638080358505248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,1,256,power_law_1.2,0.9496832275390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,1,256,power_law_1.2,0.01861376047134399
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,1,256,power_law_1.2,0.01860095977783203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,1,256,power_law_1.2,0.01856256008148193
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,1,256,power_law_1.2,0.019041279554367067
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,1,256,power_law_1.2,1.227667236328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,1,256,power_law_1.2,0.018705919981002808
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,1,256,power_law_1.2,0.018592000007629395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,1,256,power_law_1.2,0.0187827205657959
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,1,256,power_law_1.2,0.01856384038925171
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,1,256,power_law_1.2,0.018247679471969605
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,1,256,power_law_1.2,0.01831040024757385
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,1,256,power_law_1.01,0.104518404006958
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,1,256,power_law_1.01,0.13281023979187012
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,1,256,power_law_1.2,0.0288972806930542
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,1,256,power_law_1.2,0.0198361599445343
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,1,256,power_law_1.2,0.021077759265899658
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,1,256,power_law_1.2,0.03258368015289307
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,2,128,balanced,0.031816959381103516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,1,256,power_law_1.2,0.048382720947265624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,2,128,balanced,0.03214207887649536
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,2,128,balanced,0.03410943984985352
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,2,128,balanced,0.032170240879058835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,2,128,balanced,0.034588160514831545
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,1,256,power_law_1.2,0.0223091197013855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,2,128,balanced,0.04304128170013428
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,2,128,balanced,0.060834560394287106
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,2,128,balanced,0.07667840003967286
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,2,128,balanced,0.08017536163330077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,2,128,balanced,0.06332159996032714
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,2,128,balanced,0.0784550380706787
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,2,128,balanced,0.08695167541503906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,2,128,balanced,0.08585599899291993
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,2,128,balanced,0.07922815799713134
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,2,128,balanced,0.10643072128295898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,2,128,balanced,0.0926796817779541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,2,128,balanced,0.08575488090515136
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,2,128,balanced,0.1070963191986084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,2,128,balanced,0.11266688346862792
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,2,128,balanced,0.12760576248168945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,2,128,balanced,0.12019455909729004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,2,128,balanced,0.13207807540893554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,2,128,balanced,0.14179455757141113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,2,128,balanced,0.1533580780029297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,2,128,balanced,0.17061376571655273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,2,128,balanced,0.2063667106628418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,2,128,balanced,0.24347776412963867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,2,128,balanced,0.032951040267944334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,2,128,balanced,0.03289344072341919
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,2,128,balanced,0.03300352096557617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,2,128,balanced,0.03258752107620239
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,2,128,balanced,0.03357311964035034
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,2,128,balanced,0.04230783939361572
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,2,128,balanced,0.041739521026611326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,2,128,balanced,0.04278143882751465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,2,128,balanced,0.04235648155212403
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,2,128,balanced,0.043114237785339356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,2,128,balanced,0.04278656005859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,2,128,balanced,0.044049921035766604
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,2,128,balanced,0.03299328088760376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,2,128,balanced,0.043600640296936034
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,2,128,balanced,0.04608128070831299
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,2,128,balanced,0.04953343868255615
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,2,128,balanced,0.04565248012542725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,2,128,balanced,0.045583357810974126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,2,128,balanced,0.04858751773834229
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,2,128,balanced,0.05220863819122314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,2,128,balanced,0.04894336223602295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,2,128,balanced,0.05555967807769775
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,2,128,balanced,0.06211328029632568
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,2,128,balanced,0.07274367809295654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,2,128,balanced,0.08235136032104493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,2,128,balanced,0.1029043197631836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,2,128,balanced,0.12500224113464356
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,1,256,power_law_1.2,0.06567296028137207
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,1,256,power_law_1.2,0.09419391632080079
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,1,256,power_law_1.2,0.12048128128051758
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,1,256,power_law_1.2,0.18649471282958985
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,1,256,power_law_1.2,0.2662668800354004
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,2,128,balanced,0.018170880079269407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,2,128,balanced,0.01863935947418213
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,2,128,balanced,0.018713599443435668
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,2,128,balanced,0.01856256008148193
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,2,128,balanced,0.018668800592422485
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,2,128,balanced,0.018778879642486573
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,2,128,balanced,0.018617600202560425
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,2,128,balanced,0.018621439933776858
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,2,128,balanced,0.018931200504302977
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,2,128,balanced,0.018650879859924314
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,2,128,balanced,0.018955520391464233
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,2,128,balanced,0.018228479623794556
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,2,128,balanced,0.019007999897003174
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,2,128,balanced,0.019489279985427856
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,2,128,balanced,0.0194214403629303
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,2,128,balanced,0.01977471947669983
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,2,128,balanced,0.01871999979019165
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,2,128,balanced,0.019834879636764526
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,2,128,balanced,0.018236160278320312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,2,128,balanced,0.018652160167694092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,2,128,power_law_1.01,0.033642239570617676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,2,128,power_law_1.01,0.034131200313568116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,2,128,power_law_1.01,0.03452415943145752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,2,128,power_law_1.01,0.03377919912338257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,2,128,power_law_1.01,0.0651635217666626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,2,128,power_law_1.01,0.06777984142303467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,2,128,power_law_1.01,0.08528512001037598
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,2,128,power_law_1.01,0.09135999679565429
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,2,128,power_law_1.01,0.09469440460205078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,2,128,power_law_1.01,0.10411904335021974
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,2,128,power_law_1.01,0.10829440116882325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,2,128,power_law_1.01,0.1041215991973877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,2,128,power_law_1.01,0.1060531234741211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,2,128,power_law_1.01,0.10895872116088867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,2,128,power_law_1.01,0.11911935806274414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,2,128,power_law_1.01,0.12532735824584962
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,2,128,power_law_1.01,0.12759807586669922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,2,128,power_law_1.01,0.13801983833312986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,2,128,power_law_1.01,0.1459315204620361
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,2,128,power_law_1.01,0.1756096076965332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,2,128,power_law_1.01,0.18840063095092774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,2,128,power_law_1.01,0.24022911071777342
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,2,128,power_law_1.01,0.2844659233093262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,2,128,power_law_1.01,0.436071662902832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,2,128,power_law_1.01,0.5446847915649414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,2,128,power_law_1.01,0.8469593811035157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,2,128,power_law_1.01,1.272336654663086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,2,128,power_law_1.01,0.0337932801246643
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,2,128,power_law_1.01,0.03461375951766968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,2,128,power_law_1.01,0.03460864067077637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,2,128,power_law_1.01,0.03452672004699707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,2,128,power_law_1.01,0.04314112186431885
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,2,128,power_law_1.01,0.04413311958312988
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,2,128,power_law_1.01,0.044860157966613766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,2,128,power_law_1.01,0.04605696201324463
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,2,128,power_law_1.01,0.046900482177734376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,2,128,power_law_1.01,0.04837376117706299
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,2,128,power_law_1.01,0.049329919815063475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,2,128,power_law_1.01,0.04564991950988769
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,2,128,power_law_1.01,0.04812543869018555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,2,128,power_law_1.01,0.04818816184997558
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,2,128,balanced,0.01866752028465271
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,2,128,power_law_1.01,0.051151361465454105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,2,128,power_law_1.01,0.05239295959472656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,2,128,power_law_1.01,0.06020351886749268
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,2,128,power_law_1.01,0.06614655971527099
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,2,128,balanced,0.018688000440597534
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,2,128,power_law_1.01,0.07801343917846679
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,2,128,power_law_1.01,0.0867187213897705
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,2,128,balanced,0.01857792019844055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,2,128,power_law_1.01,0.10742015838623047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,2,128,power_law_1.01,0.15218048095703124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,2,128,power_law_1.01,0.05182208061218262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,2,128,power_law_1.01,0.22076416015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,2,128,power_law_1.01,0.2858022308349609
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,2,128,power_law_1.01,0.018403840065002442
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,2,128,power_law_1.01,0.4816447830200195
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,2,128,power_law_1.01,0.01874815940856934
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,2,128,power_law_1.01,0.018586879968643187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,2,128,power_law_1.01,0.6182156753540039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,2,128,power_law_1.01,0.01877760052680969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,2,128,power_law_1.01,0.018632960319519044
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,2,128,power_law_1.01,0.01866752028465271
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,2,128,power_law_1.01,0.019031039476394652
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,2,128,power_law_1.01,0.018622720241546632
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,2,128,power_law_1.01,0.018780159950256347
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,2,128,power_law_1.01,0.018906879425048827
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,2,128,power_law_1.01,0.01894144058227539
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,2,128,power_law_1.01,0.01899135947227478
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,2,128,power_law_1.01,0.019031039476394652
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,2,128,power_law_1.01,0.01936511993408203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,2,128,power_law_1.01,0.019438079595565795
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,2,128,power_law_1.01,0.019631359577178955
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,2,128,power_law_1.01,0.018983680009841918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,2,128,balanced,0.022385919094085695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,2,128,balanced,0.02477695941925049
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,2,128,balanced,0.029624319076538085
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,2,128,balanced,0.02317312002182007
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,2,128,power_law_1.01,0.021062400341033936
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,2,128,power_law_1.01,0.021812479496002197
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,2,128,power_law_1.01,0.023508479595184324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,2,128,power_law_1.2,0.033388800621032715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,2,128,power_law_1.2,0.034391040802001956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,2,128,power_law_1.2,0.03381632089614868
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,2,128,power_law_1.01,0.025047039985656737
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,2,128,power_law_1.2,0.05756800174713135
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,2,128,power_law_1.2,0.07665023803710938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,2,128,power_law_1.2,0.09712767601013184
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,2,128,power_law_1.2,0.0793446397781372
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,2,128,power_law_1.2,0.0988479995727539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,2,128,power_law_1.2,0.09898624420166016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,2,128,power_law_1.2,0.10535552024841308
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,2,128,power_law_1.2,0.109039363861084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,2,128,power_law_1.2,0.1088038444519043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,2,128,power_law_1.2,0.11291775703430176
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,2,128,power_law_1.01,0.032627201080322264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,2,128,power_law_1.2,0.12184576034545898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,2,128,power_law_1.2,0.1274623966217041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,2,128,power_law_1.2,0.13077887535095215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,2,128,power_law_1.2,0.13122431755065916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,2,128,power_law_1.2,0.1476518440246582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,2,128,power_law_1.2,0.15845376014709472
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,2,128,power_law_1.2,0.190118408203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,2,128,power_law_1.2,0.22726144790649414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,2,128,power_law_1.01,0.05575551986694336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,2,128,power_law_1.01,0.041898241043090825
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,2,128,power_law_1.2,0.3769510269165039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,2,128,power_law_1.2,0.45931262969970704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,2,128,power_law_1.2,0.03340800046920776
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,2,128,power_law_1.2,0.8056050872802734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,2,128,power_law_1.2,0.03333631992340088
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,2,128,power_law_1.2,0.8878131103515624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,2,128,power_law_1.2,0.03373823881149292
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,2,128,power_law_1.2,0.041972479820251464
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,2,128,power_law_1.2,0.04200319766998291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,2,128,power_law_1.2,0.04362624168395996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,2,128,power_law_1.2,0.04258815765380859
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,2,128,power_law_1.01,0.06702464103698731
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,2,128,power_law_1.2,0.04342912197113037
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,2,128,power_law_1.2,1.4867674255371095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,2,128,power_law_1.2,0.04569983959197998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,2,128,power_law_1.2,0.04477312088012696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,2,128,power_law_1.2,0.04526976108551025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,2,128,power_law_1.2,0.04482687950134277
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,2,128,power_law_1.2,0.047331838607788085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,2,128,power_law_1.2,0.04979584217071533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,2,128,power_law_1.2,0.051165437698364256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,2,128,power_law_1.2,0.053859839439392085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,2,128,power_law_1.2,0.05656832218170166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,2,128,power_law_1.2,0.06493184089660645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,2,128,power_law_1.2,0.0732915210723877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,2,128,power_law_1.2,0.09365887641906738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,2,128,power_law_1.2,0.11504128456115723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,2,128,power_law_1.2,2.1141990661621093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,2,128,power_law_1.2,0.17746559143066407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,2,128,power_law_1.2,0.22802688598632814
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,2,128,power_law_1.2,0.0186188805103302
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,2,128,power_law_1.2,0.3071513557434082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,2,128,power_law_1.2,0.01856511950492859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,2,128,power_law_1.2,0.4776012802124024
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,2,128,power_law_1.2,0.018754559755325317
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,2,128,power_law_1.2,0.6745152282714844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,2,128,power_law_1.2,0.01854591965675354
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,2,128,power_law_1.2,0.018634239435195925
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,2,128,power_law_1.2,0.018659839630126955
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,2,128,power_law_1.2,0.01894271969795227
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,2,128,power_law_1.2,0.018638080358505248
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,2,128,power_law_1.2,0.018655999898910522
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,2,128,power_law_1.2,0.018965760469436647
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,2,128,power_law_1.2,1.0346816253662108
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,2,128,power_law_1.2,0.018653440475463866
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,2,128,power_law_1.2,0.019020800590515134
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,2,128,power_law_1.2,0.018952959775924684
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,2,128,power_law_1.2,0.019137279987335203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,2,128,power_law_1.2,0.01944192051887512
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,2,128,power_law_1.2,0.0230732798576355
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,2,128,power_law_1.2,0.018568960428237916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,2,128,power_law_1.01,0.09747712135314941
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,2,128,power_law_1.01,0.11978624343872071
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,2,128,power_law_1.2,0.02519295930862427
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,2,128,power_law_1.2,0.021095681190490722
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,2,128,power_law_1.2,0.020367360115051268
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,4,64,balanced,0.030021119117736816
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,2,128,power_law_1.2,0.03356031894683838
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,4,64,balanced,0.030086400508880617
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,2,128,power_law_1.2,0.044650239944458006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,4,64,balanced,0.030959360599517823
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,4,64,balanced,0.03012864112854004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,4,64,balanced,0.031685121059417724
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,4,64,balanced,0.03419519901275635
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,4,64,balanced,0.07493631839752198
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,4,64,balanced,0.07259647846221924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,4,64,balanced,0.09350912094116211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,4,64,balanced,0.06327807903289795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,4,64,balanced,0.09636223793029786
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,4,64,balanced,0.07639552116394042
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,4,64,balanced,0.09963264465332031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,4,64,balanced,0.08240256309509278
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,4,64,balanced,0.0984102439880371
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,2,128,power_law_1.2,0.05336575984954835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,4,64,balanced,0.09265407562255859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,4,64,balanced,0.10814335823059082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,4,64,balanced,0.112542724609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,4,64,balanced,0.11404416084289551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,4,64,balanced,0.1254259204864502
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,4,64,balanced,0.13252863883972169
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,4,64,balanced,0.14467328071594238
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,4,64,balanced,0.1568716812133789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,4,64,balanced,0.1865945625305176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,4,64,balanced,0.21826303482055662
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,4,64,balanced,0.0313100790977478
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,4,64,balanced,0.33455486297607423
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,4,64,balanced,0.03132416009902954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,4,64,balanced,0.0317248010635376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,4,64,balanced,0.03133055925369262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,4,64,balanced,0.032099840641021726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,4,64,balanced,0.03462271928787232
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,4,64,balanced,0.04074368000030518
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,4,64,balanced,0.04402304172515869
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,4,64,balanced,0.041532158851623535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,4,64,balanced,0.045666561126708985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,4,64,balanced,0.043208961486816404
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,4,64,balanced,0.04521984100341797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,4,64,balanced,0.045561599731445315
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,4,64,balanced,0.046060800552368164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,4,64,balanced,0.047700481414794924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,4,64,balanced,0.041520638465881346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,4,64,balanced,0.05221375942230224
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,4,64,balanced,0.04936063766479492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,4,64,balanced,0.055403518676757815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,4,64,balanced,0.0607257604598999
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,4,64,balanced,0.0668838405609131
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,4,64,balanced,0.07690752029418946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,4,64,balanced,0.048944640159606936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,4,64,balanced,0.09728128433227538
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,4,64,balanced,0.1141964817047119
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,4,64,balanced,0.1513932800292969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,4,64,balanced,0.1899839973449707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,4,64,balanced,0.27640960693359373
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,2,128,power_law_1.2,0.08060671806335448
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,2,128,power_law_1.2,0.10496767997741699
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,2,128,power_law_1.2,0.15124480247497557
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,2,128,power_law_1.2,0.21225599288940428
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,4,64,balanced,0.0186188805103302
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,4,64,balanced,0.02069375991821289
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,4,64,balanced,0.018643200397491455
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,4,64,balanced,0.02108288049697876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,4,64,balanced,0.021091840267181396
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,4,64,balanced,0.021079039573669432
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,4,64,balanced,0.02110975980758667
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,4,64,balanced,0.021418240070343018
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,4,64,balanced,0.021512959003448486
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,4,64,balanced,0.021192960739135742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,4,64,balanced,0.021483519077301026
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,4,64,balanced,0.021486079692840575
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,4,64,balanced,0.021542398929595946
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,4,64,balanced,0.021574399471282958
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,4,64,balanced,0.021464319229125978
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,4,64,balanced,0.02149631977081299
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,4,64,balanced,0.023119359016418456
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,4,64,balanced,0.018632960319519044
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,4,64,balanced,0.018585599660873413
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,4,64,balanced,0.019041279554367067
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,4,64,power_law_1.01,0.03828864097595215
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,4,64,balanced,0.020590078830718995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,4,64,power_law_1.01,0.03419647932052612
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,4,64,power_law_1.01,0.05043712139129639
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,4,64,power_law_1.01,0.0345689606666565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,4,64,power_law_1.01,0.08214912414550782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,4,64,power_law_1.01,0.0829081630706787
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,4,64,power_law_1.01,0.0840435218811035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,4,64,power_law_1.01,0.076976637840271
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,4,64,power_law_1.01,0.0949721622467041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,4,64,power_law_1.01,0.08144000053405762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,4,64,power_law_1.01,0.09339903831481934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,4,64,power_law_1.01,0.10189184188842773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,4,64,power_law_1.01,0.11125632286071778
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,4,64,power_law_1.01,0.11775103569030762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,4,64,power_law_1.01,0.12155008316040039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,4,64,power_law_1.01,0.12337920188903809
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,4,64,power_law_1.01,0.12351231575012207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,4,64,power_law_1.01,0.13310336112976073
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,4,64,power_law_1.01,0.140829439163208
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,4,64,power_law_1.01,0.16619520187377929
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,4,64,power_law_1.01,0.1978995132446289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,4,64,power_law_1.01,0.24642559051513674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,4,64,power_law_1.01,0.31011840820312503
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,4,64,power_law_1.01,0.44045566558837895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,4,64,power_law_1.01,0.49904769897460943
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,4,64,power_law_1.01,0.7675801849365234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,4,64,power_law_1.01,0.033657600879669194
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,4,64,balanced,0.021035521030426024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,4,64,power_law_1.01,0.034174718856811524
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,4,64,balanced,0.02062975883483887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,4,64,power_law_1.01,0.03501568078994751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,4,64,power_law_1.01,1.1054975891113281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,4,64,power_law_1.01,0.04046207904815674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,4,64,power_law_1.01,0.043191041946411136
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,4,64,power_law_1.01,0.041635842323303224
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,4,64,power_law_1.01,0.042758398056030274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,4,64,power_law_1.01,0.048520960807800294
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,4,64,power_law_1.01,0.04443903923034668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,4,64,power_law_1.01,0.04526847839355469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,4,64,power_law_1.01,0.046579198837280275
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,4,64,power_law_1.01,0.047760639190673825
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,4,64,power_law_1.01,0.05101439952850342
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,4,64,power_law_1.01,0.054737920761108394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,4,64,power_law_1.01,0.03340928077697754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,4,64,power_law_1.01,0.053489918708801265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,4,64,power_law_1.01,0.057372159957885735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,4,64,power_law_1.01,0.06388224124908447
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,4,64,power_law_1.01,0.06782464027404786
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,4,64,power_law_1.01,0.07904895782470703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,4,64,power_law_1.01,0.0979327964782715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,4,64,power_law_1.01,0.12509440422058105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,4,64,power_law_1.01,0.16256511688232422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,4,64,power_law_1.01,0.2337958335876465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,4,64,power_law_1.01,0.26764799118041993
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,4,64,power_law_1.01,0.020692479610443116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,4,64,power_law_1.01,0.43681152343749996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,4,64,power_law_1.01,0.02069119930267334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,4,64,power_law_1.01,0.021013760566711427
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,4,64,power_law_1.01,0.6044812774658203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,4,64,power_law_1.01,0.020686080455780027
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,4,64,power_law_1.01,0.020281600952148437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,4,64,power_law_1.01,0.02027008056640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,4,64,power_law_1.01,0.020687360763549805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,4,64,power_law_1.01,0.02023168087005615
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,4,64,power_law_1.01,0.0202675199508667
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,4,64,power_law_1.01,0.020432639122009277
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,4,64,power_law_1.01,0.02064768075942993
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,4,64,power_law_1.01,0.020673279762268064
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,4,64,power_law_1.01,0.020673279762268064
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,4,64,power_law_1.01,0.020753920078277588
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,4,64,power_law_1.01,0.02105855941772461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,4,64,power_law_1.01,0.020651519298553467
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,4,64,power_law_1.01,0.020770559310913088
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,4,64,balanced,0.026426880359649657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,4,64,balanced,0.025017600059509277
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,4,64,balanced,0.038581759929656984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,4,64,balanced,0.03177983999252319
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,4,64,power_law_1.01,0.022735359668731688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,4,64,power_law_1.01,0.025235199928283693
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,4,64,power_law_1.01,0.022291200160980223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,4,64,power_law_1.2,0.03831295967102051
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,4,64,power_law_1.2,0.033766400814056394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,4,64,power_law_1.2,0.034777600765228275
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,4,64,power_law_1.2,0.04114431858062744
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,4,64,power_law_1.01,0.027616000175476073
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,4,64,power_law_1.2,0.07243135929107666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,4,64,power_law_1.2,0.09277312278747558
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,4,64,power_law_1.2,0.07329279899597167
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,4,64,power_law_1.2,0.09316351890563965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,4,64,power_law_1.2,0.07934080123901367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,4,64,power_law_1.2,0.10441727638244629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,4,64,power_law_1.2,0.09919487953186035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,4,64,power_law_1.2,0.11387136459350586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,4,64,power_law_1.01,0.03507967948913574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,4,64,power_law_1.2,0.1090112018585205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,4,64,power_law_1.2,0.12355456352233887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,4,64,power_law_1.2,0.12385279655456542
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,4,64,power_law_1.01,0.04090112209320068
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,4,64,power_law_1.2,0.1205465602874756
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,4,64,power_law_1.2,0.12877823829650878
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,4,64,power_law_1.2,0.14310400009155272
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,4,64,power_law_1.2,0.1543462371826172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,4,64,power_law_1.2,0.20813823699951173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,4,64,power_law_1.2,0.21747968673706053
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,4,64,power_law_1.2,0.27412096023559573
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,4,64,power_law_1.2,0.4008563232421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,4,64,power_law_1.2,0.567644157409668
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,4,64,power_law_1.01,0.06000768184661866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,4,64,power_law_1.2,0.03626751899719238
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,4,64,power_law_1.2,0.7712127685546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,4,64,power_law_1.2,0.033775360584259034
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,4,64,power_law_1.2,0.034237439632415774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,4,64,power_law_1.2,0.040767998695373536
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,4,64,power_law_1.2,0.044372482299804686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,4,64,power_law_1.2,1.2700825500488282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,4,64,power_law_1.2,0.04196479797363281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,4,64,power_law_1.2,0.0472870397567749
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,4,64,power_law_1.01,0.0721779203414917
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,4,64,power_law_1.2,0.049838080406188964
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,4,64,power_law_1.2,0.04491136074066162
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,4,64,power_law_1.2,0.033551359176635744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,4,64,power_law_1.2,0.04729343891143799
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,4,64,power_law_1.2,0.047697920799255375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,4,64,power_law_1.2,0.04896128177642822
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,4,64,power_law_1.2,0.050728960037231444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,4,64,power_law_1.2,0.05389311790466309
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,4,64,power_law_1.2,0.05386879920959473
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,4,64,power_law_1.2,2.001148223876953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,4,64,power_law_1.2,0.060890879631042484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,4,64,power_law_1.2,0.06871808052062989
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,4,64,power_law_1.2,0.07689599990844727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,4,64,power_law_1.2,0.10150527954101562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,4,64,power_law_1.2,0.11612544059753419
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,4,64,power_law_1.2,0.16597503662109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,4,64,power_law_1.2,0.25262975692749023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,4,64,power_law_1.2,0.3090265655517578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,4,64,power_law_1.2,0.41866497039794925
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,4,64,power_law_1.2,0.020977919101715085
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,4,64,power_law_1.2,0.02102783918380737
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,4,64,power_law_1.2,0.020718080997467043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,4,64,power_law_1.2,0.019865599870681764
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,4,64,power_law_1.2,0.7170111846923828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,4,64,power_law_1.2,0.020280320644378663
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,4,64,power_law_1.2,0.02019711971282959
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,4,64,power_law_1.2,0.020273919105529788
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,4,64,power_law_1.2,0.02064255952835083
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,4,64,power_law_1.2,0.020538880825042724
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,4,64,power_law_1.2,0.02068095922470093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,4,64,power_law_1.2,1.1005299377441407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,4,64,power_law_1.2,0.020678400993347168
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,4,64,power_law_1.2,0.02069119930267334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,4,64,power_law_1.2,0.020736000537872314
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,4,64,power_law_1.2,0.021089279651641847
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,4,64,power_law_1.2,0.02063744068145752
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,4,64,power_law_1.2,0.023174400329589846
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,4,64,power_law_1.2,0.02071295976638794
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,4,64,power_law_1.01,0.09520895957946778
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,4,64,power_law_1.01,0.12989055633544921
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,4,64,power_law_1.2,0.022101759910583496
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,4,64,power_law_1.2,0.030392320156097413
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,4,64,power_law_1.2,0.03678719997406006
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,4,64,power_law_1.2,0.025512959957122806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,8,32,balanced,0.03334784030914307
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,8,32,balanced,0.03092992067337036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,8,32,balanced,0.03254656076431274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,8,32,balanced,0.03049344062805176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,8,32,balanced,0.032533760070800784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,8,32,balanced,0.047759361267089843
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,4,64,power_law_1.2,0.064236798286438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,8,32,balanced,0.07210879802703858
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,8,32,balanced,0.07058944225311278
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,8,32,balanced,0.08228351593017577
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,8,32,balanced,0.07352960109710693
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,8,32,balanced,0.0835865592956543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,8,32,balanced,0.08644351959228516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,8,32,balanced,0.10332159996032715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,8,32,balanced,0.1092249584197998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,8,32,balanced,0.11303808212280272
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,8,32,balanced,0.11224063873291015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,4,64,power_law_1.2,0.04600192070007324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,8,32,balanced,0.12254464149475097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,8,32,balanced,0.130513916015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,8,32,balanced,0.14073216438293457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,8,32,balanced,0.15114496231079103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,8,32,balanced,0.18142976760864257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,8,32,balanced,0.20605823516845706
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,8,32,balanced,0.25969024658203127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,8,32,balanced,0.0909721565246582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,8,32,balanced,0.3142425537109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,8,32,balanced,0.034190080165863036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,8,32,balanced,0.4214937591552735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,8,32,balanced,0.03082880020141602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,8,32,balanced,0.03408384084701538
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,8,32,balanced,0.0313100790977478
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,8,32,balanced,0.0333951997756958
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,8,32,balanced,0.528927993774414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,8,32,balanced,0.0374566388130188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,8,32,balanced,0.04688255786895752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,8,32,balanced,0.04853248119354248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,8,32,balanced,0.053109760284423824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,8,32,balanced,0.04770559787750244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,8,32,balanced,0.04806655883789063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,8,32,balanced,0.048951039314270015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,8,32,balanced,0.048949761390686033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,8,32,balanced,0.05053311824798584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,8,32,balanced,0.052629761695861824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,8,32,balanced,0.0551526403427124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,8,32,balanced,0.057105917930603024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,8,32,balanced,0.06208640098571777
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,8,32,balanced,0.06770175933837891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,8,32,balanced,0.07659008026123046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,8,32,balanced,0.09285120010375977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,8,32,balanced,0.109104642868042
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,8,32,balanced,0.1503923225402832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,8,32,balanced,0.1795212745666504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,8,32,balanced,0.2501286315917969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,8,32,balanced,0.3220211029052734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,8,32,balanced,0.05183104038238525
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,4,64,power_law_1.2,0.0754918384552002
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,4,64,power_law_1.2,0.10816127777099609
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,4,64,power_law_1.2,0.1577023983001709
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,4,64,power_law_1.2,0.17922687530517578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,8,32,balanced,0.01874559998512268
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,8,32,balanced,0.019095040559768676
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,8,32,balanced,0.022307839393615723
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,8,32,balanced,0.02232192039489746
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,8,32,balanced,0.019032319784164427
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,8,32,balanced,0.022150399684906005
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,8,32,balanced,0.022735359668731688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,8,32,balanced,0.022750720977783204
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,8,32,balanced,0.022547199726104736
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,8,32,balanced,0.022663679122924805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,8,32,balanced,0.022885119915008544
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,8,32,balanced,0.02270335912704468
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,8,32,balanced,0.018676480054855345
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,8,32,balanced,0.022707200050354003
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,8,32,balanced,0.022708480358123777
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,8,32,balanced,0.02443903923034668
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,8,32,balanced,0.024775679111480715
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,8,32,balanced,0.026004478931427
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,8,32,balanced,0.022704639434814454
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,8,32,balanced,0.021063680648803713
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,8,32,balanced,0.02190592050552368
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,8,32,power_law_1.01,0.033771519660949705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,8,32,power_law_1.01,0.03381887912750244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,8,32,power_law_1.01,0.03545471906661987
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,8,32,power_law_1.01,0.03704960107803344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,8,32,power_law_1.01,0.08335871696472168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,8,32,power_law_1.01,0.0768998384475708
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,8,32,power_law_1.01,0.09430527687072754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,8,32,power_law_1.01,0.08691712379455566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,8,32,power_law_1.01,0.09511167526245116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,8,32,power_law_1.01,0.08947967529296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,8,32,power_law_1.01,0.1038976001739502
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,8,32,power_law_1.01,0.10708352088928223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,8,32,power_law_1.01,0.11105536460876464
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,8,32,power_law_1.01,0.11777279853820802
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,8,32,power_law_1.01,0.11908991813659668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,8,32,power_law_1.01,0.12021247863769531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,8,32,power_law_1.01,0.1293017578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,8,32,power_law_1.01,0.13860608100891114
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,8,32,power_law_1.01,0.15597439765930177
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,8,32,power_law_1.01,0.1872345542907715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,8,32,power_law_1.01,0.19437440872192382
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,8,32,power_law_1.01,0.2822233581542969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,8,32,power_law_1.01,0.3130547142028809
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,8,32,power_law_1.01,0.4417267227172852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,8,32,power_law_1.01,0.5747660827636719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,8,32,power_law_1.01,0.7684429168701172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,8,32,power_law_1.01,0.0366156792640686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,8,32,power_law_1.01,1.0694950103759766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,8,32,power_law_1.01,0.036590080261230465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,8,32,power_law_1.01,0.03705728054046631
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,8,32,power_law_1.01,0.03660928010940552
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,8,32,power_law_1.01,0.04590335845947265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,8,32,power_law_1.01,0.0458406400680542
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,8,32,power_law_1.01,0.04708352088928223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,8,32,power_law_1.01,0.047735037803649905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,8,32,power_law_1.01,0.04814720153808594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,8,32,power_law_1.01,0.04807424068450928
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,8,32,power_law_1.01,0.049269762039184574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,8,32,power_law_1.01,0.05181056022644043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,8,32,power_law_1.01,0.05214335918426514
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,8,32,power_law_1.01,0.05549439907073974
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,8,32,power_law_1.01,0.057731838226318354
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,8,32,power_law_1.01,0.05761536121368408
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,8,32,power_law_1.01,0.06256896018981933
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,8,32,power_law_1.01,0.06924416065216064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,8,32,power_law_1.01,0.07581823825836181
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,8,32,power_law_1.01,0.09891072273254395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,8,32,power_law_1.01,0.11404288291931153
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,8,32,balanced,0.02224128007888794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,8,32,power_law_1.01,0.15270912170410156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,8,32,balanced,0.022238719463348388
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,8,32,power_law_1.01,0.1800985527038574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,8,32,power_law_1.01,0.25202816009521484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,8,32,power_law_1.01,0.33830272674560546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,8,32,power_law_1.01,0.022215681076049803
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,8,32,power_law_1.01,0.02191744089126587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,8,32,power_law_1.01,0.5313689422607422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,8,32,power_law_1.01,0.022305281162261964
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,8,32,power_law_1.01,0.6453759765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,8,32,power_law_1.01,0.022099199295043944
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,8,32,power_law_1.01,0.021857280731201172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,8,32,power_law_1.01,0.021418240070343018
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,8,32,power_law_1.01,0.021839361190795898
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,8,32,power_law_1.01,0.02182015895843506
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,8,32,power_law_1.01,0.021913599967956544
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,8,32,power_law_1.01,0.022158079147338867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,8,32,power_law_1.01,0.02185472011566162
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,8,32,power_law_1.01,0.022286078929901122
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,8,32,power_law_1.01,0.022261760234832763
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,8,32,power_law_1.01,0.022305281162261964
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,8,32,power_law_1.01,0.022475519180297852
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,8,32,power_law_1.01,0.02273024082183838
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,8,32,power_law_1.01,0.02326911926269531
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,8,32,balanced,0.038384640216827394
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,8,32,balanced,0.048046078681945804
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,8,32,balanced,0.03337088108062744
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,8,32,balanced,0.05916287899017334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,8,32,power_law_1.01,0.03214848041534424
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,8,32,power_law_1.01,0.025882880687713623
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,8,32,power_law_1.01,0.03537408113479614
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,8,32,power_law_1.01,0.026374399662017822
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,8,32,power_law_1.2,0.03536256074905396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,8,32,power_law_1.2,0.03569792032241821
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,8,32,power_law_1.2,0.03519999980926514
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,8,32,power_law_1.2,0.051351041793823236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,8,32,power_law_1.2,0.07698048114776611
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,8,32,power_law_1.2,0.0774284791946411
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,8,32,power_law_1.2,0.08997504234313965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,8,32,power_law_1.2,0.08507391929626465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,8,32,power_law_1.2,0.08768896102905274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,8,32,power_law_1.2,0.09447808265686035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,8,32,power_law_1.2,0.10435839653015137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,8,32,power_law_1.2,0.10850815773010254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,8,32,power_law_1.2,0.10853376388549804
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,8,32,power_law_1.01,0.04556928157806396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,8,32,power_law_1.2,0.11684096336364747
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,8,32,power_law_1.2,0.11787648200988769
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,8,32,power_law_1.2,0.12433792114257813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,8,32,power_law_1.2,0.1287168025970459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,8,32,power_law_1.2,0.13898112297058104
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,8,32,power_law_1.01,0.05467391967773437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,8,32,power_law_1.2,0.16673408508300783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,8,32,power_law_1.2,0.2175385665893555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,8,32,power_law_1.2,0.2405721664428711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,8,32,power_law_1.2,0.32768127441406253
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,8,32,power_law_1.2,0.42939521789550783
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,8,32,power_law_1.01,0.0984166431427002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,8,32,power_law_1.2,0.5847398376464843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,8,32,power_law_1.2,0.03667327880859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,8,32,power_law_1.2,0.03700992107391358
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,8,32,power_law_1.2,0.7813926696777344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,8,32,power_law_1.2,0.037480959892272944
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,8,32,power_law_1.01,0.07569024085998535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,8,32,power_law_1.2,0.03749248027801514
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,8,32,power_law_1.2,0.04487679958343506
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,8,32,power_law_1.2,1.083947525024414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,8,32,power_law_1.2,0.04937727928161621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,8,32,power_law_1.2,0.04710527896881104
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,8,32,power_law_1.2,0.0509555196762085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,8,32,power_law_1.2,0.048229122161865236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,8,32,power_law_1.2,0.04770559787750244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,8,32,power_law_1.2,0.052046079635620114
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,8,32,power_law_1.2,1.3573504638671874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,8,32,power_law_1.2,0.05527935981750488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,8,32,power_law_1.2,0.0572160005569458
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,8,32,power_law_1.2,0.056986880302429196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,8,32,power_law_1.2,0.060833277702331545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,8,32,power_law_1.2,0.07105023860931396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,8,32,power_law_1.2,0.0810854434967041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,8,32,power_law_1.2,0.11297920227050781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,8,32,power_law_1.2,0.1367807960510254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,8,32,power_law_1.2,0.19037824630737304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,8,32,power_law_1.2,0.23599615097045898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,8,32,power_law_1.2,0.30632448196411133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,8,32,power_law_1.2,0.049492478370666504
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,8,32,power_law_1.2,0.02217087984085083
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,8,32,power_law_1.2,0.5210099029541015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,8,32,power_law_1.2,0.022312960624694823
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,8,32,power_law_1.2,0.02184448003768921
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,8,32,power_law_1.2,0.6492684936523437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,8,32,power_law_1.2,0.061363201141357425
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,8,32,power_law_1.2,0.021610240936279296
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,8,32,power_law_1.2,0.02139904022216797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,8,32,power_law_1.2,0.021857280731201172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,8,32,power_law_1.2,0.021899518966674806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,8,32,power_law_1.2,0.9957644653320312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,8,32,power_law_1.2,0.02188416004180908
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,8,32,power_law_1.2,0.021945600509643552
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,8,32,power_law_1.2,0.022315518856048582
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,8,32,power_law_1.2,0.022653439044952393
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,8,32,power_law_1.2,0.021850879192352297
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,8,32,power_law_1.2,0.02306688070297241
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,8,32,power_law_1.2,0.02311295986175537
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,8,32,power_law_1.2,0.027084798812866212
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,8,32,power_law_1.2,0.03728895902633667
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,8,32,power_law_1.2,0.021914880275726318
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,8,32,power_law_1.2,0.021913599967956544
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,8,32,power_law_1.01,0.11879039764404298
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,8,32,power_law_1.01,0.1521343994140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,8,32,power_law_1.2,0.024140799045562746
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,8,32,power_law_1.2,0.031137280464172364
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,8,32,power_law_1.2,0.045692157745361325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,16,16,balanced,0.028424959182739257
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,8,32,power_law_1.2,0.06156544208526611
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,16,16,balanced,0.030024960041046145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,16,16,balanced,0.03169663906097412
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,16,16,balanced,0.03453824043273926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,16,16,balanced,0.04070144176483154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,16,16,balanced,0.08948351860046386
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,16,16,balanced,0.08421504020690918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,8,32,power_law_1.2,0.09772671699523926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,16,16,balanced,0.09054207801818848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,16,16,balanced,0.10092415809631347
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,16,16,balanced,0.10500096321105956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,16,16,balanced,0.116428804397583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,16,16,balanced,0.11517824172973631
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,16,16,balanced,0.028494079113006592
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,16,16,balanced,0.12004223823547364
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,8,32,power_law_1.2,0.07855743885040282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,16,16,balanced,0.12340096473693847
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,16,16,balanced,0.13216383934020998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,16,16,balanced,0.14221952438354493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,16,16,balanced,0.15388031959533693
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,16,16,balanced,0.17900928497314453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,16,16,balanced,0.2067967987060547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,16,16,balanced,0.2573196792602539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,16,16,balanced,0.3095923233032226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,16,16,balanced,0.11573760032653808
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,16,16,balanced,0.41208961486816403
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,16,16,balanced,0.5147750473022461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,16,16,balanced,0.034335999488830565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,16,16,balanced,0.03377408027648926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,16,16,balanced,0.7120435333251953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,16,16,balanced,0.03524096012115478
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,16,16,balanced,0.03741568088531494
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,16,16,balanced,0.04394239902496338
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,16,16,balanced,0.9151372528076172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,16,16,balanced,0.05348224163055419
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,16,16,balanced,0.05384191989898681
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,16,16,balanced,0.05503359794616699
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,16,16,balanced,0.05502848148345947
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,16,16,balanced,0.057091841697692866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,16,16,balanced,0.05822976112365723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,16,16,balanced,0.06017792224884033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,16,16,balanced,0.05720960140228272
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,16,16,balanced,0.0608460807800293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,16,16,balanced,0.06345600128173828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,16,16,balanced,0.06653696060180664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,16,16,balanced,0.07270400047302246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,16,16,balanced,0.03133568048477173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,16,16,balanced,0.09767295837402343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,16,16,balanced,0.1138969612121582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,16,16,balanced,0.14659839630126953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,16,16,balanced,0.17418624877929687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,16,16,balanced,0.25065471649169924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,16,16,balanced,0.3104217529296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,16,16,balanced,0.440423698425293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,16,16,balanced,0.5726784133911134
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,16,16,balanced,0.08211199760437012
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,8,32,power_law_1.2,0.1367679977416992
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,8,32,power_law_1.2,0.20743040084838865
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,8,32,power_law_1.2,0.25741952896118164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,16,16,balanced,0.01989375948905945
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,16,16,balanced,0.026680319309234618
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,16,16,balanced,0.02021888017654419
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,16,16,balanced,0.026847999095916748
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,16,16,balanced,0.02064255952835083
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,16,16,balanced,0.02681472063064575
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,16,16,balanced,0.026759679317474368
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,16,16,balanced,0.0268339204788208
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,16,16,balanced,0.02689663887023926
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,16,16,balanced,0.027173120975494385
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,16,16,balanced,0.026823680400848388
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,16,16,balanced,0.026830079555511477
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,16,16,balanced,0.027308800220489503
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,16,16,balanced,0.02886399984359741
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,16,16,balanced,0.023566079139709473
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,16,16,balanced,0.029730560779571535
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,16,16,balanced,0.031239678859710695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,16,16,balanced,0.03433727979660034
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,16,16,balanced,0.03788928031921387
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,16,16,balanced,0.02243839979171753
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,16,16,power_law_1.01,0.03424511909484863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,16,16,power_law_1.01,0.035856640338897704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,16,16,power_law_1.01,0.04239615917205811
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,16,16,power_law_1.01,0.04622975826263428
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,16,16,power_law_1.01,0.05074048042297363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,16,16,power_law_1.01,0.08497664451599121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,16,16,power_law_1.01,0.0937177562713623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,16,16,power_law_1.01,0.1014899158477783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,16,16,power_law_1.01,0.10194815635681151
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,16,16,power_law_1.01,0.10292736053466797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,16,16,power_law_1.01,0.11131520271301269
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,16,16,power_law_1.01,0.11770367622375488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,16,16,power_law_1.01,0.12393471717834473
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,16,16,power_law_1.01,0.12139391899108887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,16,16,power_law_1.01,0.12636159896850585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,16,16,power_law_1.01,0.13128576278686524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,16,16,power_law_1.01,0.1365401554107666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,16,16,power_law_1.01,0.15163264274597169
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,16,16,power_law_1.01,0.16751104354858398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,16,16,power_law_1.01,0.21196928024291992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,16,16,power_law_1.01,0.22402175903320312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,16,16,power_law_1.01,0.29835136413574215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,16,16,power_law_1.01,0.37577472686767577
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,16,16,power_law_1.01,0.5362124633789063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,16,16,balanced,0.026819839477539065
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,16,16,balanced,0.026776320934295654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,16,16,power_law_1.01,0.7035916900634765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,16,16,power_law_1.01,0.035436799526214605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,16,16,power_law_1.01,0.989159698486328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,16,16,power_law_1.01,0.041136641502380375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,16,16,power_law_1.01,0.04148096084594727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,16,16,power_law_1.01,0.04239488124847412
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,16,16,power_law_1.01,0.04322303771972656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,16,16,power_law_1.01,1.32674560546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,16,16,power_law_1.01,0.052547841072082525
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,16,16,power_law_1.01,0.054261760711669924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,16,16,power_law_1.01,0.05435904026031494
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,16,16,power_law_1.01,0.056733441352844236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,16,16,power_law_1.01,0.05095424175262451
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,16,16,power_law_1.01,0.05988351821899414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,16,16,power_law_1.01,0.052634878158569334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,16,16,power_law_1.01,0.06495872020721435
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,16,16,power_law_1.01,0.05951871871948242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,16,16,power_law_1.01,0.06487040042877197
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,16,16,power_law_1.01,0.06817152023315429
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,16,16,power_law_1.01,0.06998655796051026
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,16,16,power_law_1.01,0.07966464042663575
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,16,16,power_law_1.01,0.0883903980255127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,16,16,power_law_1.01,0.11472767829895018
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,16,16,power_law_1.01,0.1387507152557373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,16,16,power_law_1.01,0.19147903442382813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,16,16,power_law_1.01,0.2377663993835449
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,16,16,power_law_1.01,0.3382233428955078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,16,16,power_law_1.01,0.42598911285400387
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,16,16,power_law_1.01,0.02270335912704468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,16,16,power_law_1.01,0.6166361618041992
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,16,16,power_law_1.01,0.02696063995361328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,16,16,power_law_1.01,0.02720639944076538
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,16,16,power_law_1.01,0.7435903930664063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,16,16,power_law_1.01,0.02643712043762207
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,16,16,power_law_1.01,0.02673536062240601
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,16,16,power_law_1.01,0.025626881122589113
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,16,16,power_law_1.01,0.026027519702911377
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,16,16,power_law_1.01,0.026190080642700196
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,16,16,power_law_1.01,0.02593280076980591
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,16,16,power_law_1.01,0.02634239912033081
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,16,16,power_law_1.01,0.026236159801483156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,16,16,power_law_1.01,0.026337280273437503
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,16,16,power_law_1.01,0.026836481094360355
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,16,16,power_law_1.01,0.027246079444885253
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,16,16,power_law_1.01,0.02777343988418579
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,16,16,power_law_1.01,0.02811392068862915
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,16,16,power_law_1.01,0.029205760955810546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,16,16,balanced,0.026824960708618163
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,16,16,balanced,0.05903744220733642
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,16,16,balanced,0.0537011194229126
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,16,16,balanced,0.0740121603012085
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,16,16,balanced,0.08958975791931152
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,16,16,power_law_1.01,0.03455359935760498
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,16,16,power_law_1.01,0.04614272117614746
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,16,16,power_law_1.01,0.03804800033569336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,16,16,power_law_1.01,0.05354239940643311
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,16,16,power_law_1.2,0.034593279361724856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,16,16,power_law_1.2,0.04404736042022705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,16,16,power_law_1.2,0.0364902400970459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,16,16,power_law_1.2,0.04058879852294922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,16,16,power_law_1.2,0.053519358634948735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,16,16,power_law_1.2,0.09378944396972656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,16,16,power_law_1.2,0.09497983932495117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,16,16,power_law_1.2,0.09847295761108399
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,16,16,power_law_1.2,0.09567359924316407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,16,16,power_law_1.2,0.11064703941345215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,16,16,power_law_1.2,0.1119603157043457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,16,16,power_law_1.2,0.12425984382629393
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,16,16,power_law_1.2,0.12424063682556152
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,16,16,power_law_1.2,0.1221235179901123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,16,16,power_law_1.2,0.1277888011932373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,16,16,power_law_1.2,0.1329587173461914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,16,16,power_law_1.2,0.13633152008056643
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,16,16,power_law_1.2,0.1561843204498291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,16,16,power_law_1.2,0.18006528854370116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,16,16,power_law_1.2,0.21204864501953127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,16,16,power_law_1.2,0.256048641204834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,16,16,power_law_1.2,0.3700915145874023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,16,16,power_law_1.2,0.4156851196289063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,16,16,power_law_1.2,0.6013875198364258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,16,16,power_law_1.2,0.73804931640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,16,16,power_law_1.01,0.0697753620147705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,16,16,power_law_1.2,1.2121727752685547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,16,16,power_law_1.2,0.035795199871063235
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,16,16,power_law_1.01,0.09093503952026367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,16,16,power_law_1.2,0.04076672077178955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,16,16,power_law_1.2,0.04115583896636963
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,16,16,power_law_1.01,0.10968064308166503
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,16,16,power_law_1.01,0.14819711685180664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,16,16,power_law_1.2,0.043656959533691406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,16,16,power_law_1.2,0.04322175979614258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,16,16,power_law_1.2,0.05355008125305176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,16,16,power_law_1.2,0.052019200325012206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,16,16,power_law_1.2,0.052707839012146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,16,16,power_law_1.2,1.6073394775390626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,16,16,power_law_1.2,0.055904002189636225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,16,16,power_law_1.2,0.056908798217773435
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,16,16,power_law_1.2,0.0587059211730957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,16,16,power_law_1.2,0.05910655975341796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,16,16,power_law_1.2,0.06656000137329102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,16,16,power_law_1.2,0.06726912021636963
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,16,16,power_law_1.2,0.06885888099670409
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,16,16,power_law_1.2,0.08223615646362305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,16,16,power_law_1.2,0.09464703559875488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,16,16,power_law_1.2,0.11933183670043945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,16,16,power_law_1.2,0.15151616096496584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,16,16,power_law_1.2,0.21348352432250978
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,16,16,power_law_1.2,0.24760063171386718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,16,16,power_law_1.2,0.39650177001953124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,16,16,power_law_1.2,0.05426815986633301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,16,16,power_law_1.2,0.47090816497802734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,16,16,power_law_1.2,0.022734079360961914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,16,16,power_law_1.2,0.06203008174896241
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,16,16,power_law_1.2,0.026438400745391843
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,16,16,power_law_1.2,0.026050560474395752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,16,16,power_law_1.2,0.7385945892333984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,16,16,power_law_1.2,0.02642047882080078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,16,16,power_law_1.2,0.026165759563446044
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,16,16,power_law_1.2,0.02609791994094849
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,16,16,power_law_1.2,0.025562880039215086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,16,16,power_law_1.2,0.025589759349823
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,16,16,power_law_1.2,0.025927679538726805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,16,16,power_law_1.2,0.02639616012573242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,16,16,power_law_1.2,0.026287360191345217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,16,16,power_law_1.2,1.0487155151367187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,16,16,power_law_1.2,0.02640768051147461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,16,16,power_law_1.2,0.02682751893997192
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,16,16,power_law_1.2,0.027580161094665528
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,16,16,power_law_1.2,0.028049919605255123
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,16,16,power_law_1.2,0.02839296102523804
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,16,16,power_law_1.2,0.037253119945526124
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,16,16,power_law_1.2,0.04690815925598145
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,16,16,power_law_1.2,0.03212543964385987
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,16,16,power_law_1.01,0.19862911224365234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,16,16,power_law_1.01,0.2612825584411621
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,16,16,power_law_1.2,0.09040255546569824
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,16,16,power_law_1.2,0.06707583904266358
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,32,8,balanced,0.02965248107910156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,16,16,power_law_1.2,0.05652095794677734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,32,8,balanced,0.03160192012786865
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,16,16,power_law_1.2,0.1231603240966797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,32,8,balanced,0.03300352096557617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,32,8,balanced,0.03667072057723999
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,32,8,balanced,0.043184638023376465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,32,8,balanced,0.11458304405212402
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,32,8,balanced,0.11009152412414551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,32,8,balanced,0.11853952407836914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,32,8,balanced,0.12080127716064452
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,32,8,balanced,0.12204416275024414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,32,8,balanced,0.12921088218688964
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,32,8,balanced,0.13222399711608887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,32,8,balanced,0.14153599739074707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,32,8,balanced,0.1426969623565674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,32,8,balanced,0.14979200363159179
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,32,8,balanced,0.16118528366088866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,32,8,balanced,0.18753280639648437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,16,16,power_law_1.2,0.19295488357543947
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,32,8,balanced,0.2126963233947754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,32,8,balanced,0.262476806640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,32,8,balanced,0.05862912178039551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,32,8,balanced,0.3140979194641113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,32,8,balanced,0.4149760055541992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,32,8,balanced,0.5142758560180665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,32,8,balanced,0.031718399524688724
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,32,8,balanced,0.7066598510742187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,32,8,balanced,0.03383039951324463
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,16,16,power_law_1.2,0.13409664154052733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,32,8,balanced,0.03950079917907715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,32,8,balanced,0.04194303989410401
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,32,8,balanced,0.9048000335693359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,32,8,balanced,0.04711552143096924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,32,8,balanced,0.057955842018127444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,32,8,balanced,0.06895487785339356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,32,8,balanced,0.07053823947906494
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,32,8,balanced,0.07412352085113524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,32,8,balanced,0.07021056175231934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,32,8,balanced,0.07346943855285645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,32,8,balanced,0.0748031997680664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,32,8,balanced,0.07972095966339111
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,32,8,balanced,0.08072447776794434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,32,8,balanced,0.08500351905822753
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,32,8,balanced,0.09372415542602539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,32,8,balanced,0.10892671585083007
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,32,8,balanced,1.6873446655273436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,32,8,balanced,0.1262169647216797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,32,8,balanced,0.1576371192932129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,32,8,balanced,0.1883967971801758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,32,8,balanced,0.25172096252441406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,32,8,balanced,0.3095027160644531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,32,8,balanced,0.07215487957000732
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,32,8,balanced,0.45249790191650396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,32,8,balanced,1.2960742187499998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,32,8,balanced,0.5687475204467773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,32,8,balanced,0.8279436492919923
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,32,8,balanced,1.0856690979003907
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,16,16,power_law_1.2,0.2897587203979492
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,16,16,power_law_1.2,0.4296652984619141
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,32,8,balanced,0.023404800891876222
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,32,8,balanced,0.03706624031066895
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,32,8,balanced,0.03759743928909302
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,32,8,balanced,0.025615360736846927
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,32,8,balanced,0.03746687889099121
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,32,8,balanced,0.03833215951919556
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,32,8,balanced,0.03793920040130615
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,32,8,balanced,0.03871871948242188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,32,8,balanced,0.03745791912078857
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,32,8,balanced,0.03935103893280029
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,32,8,balanced,0.02840447902679443
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,32,8,balanced,0.03944063901901245
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,32,8,balanced,0.04163328170776367
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,32,8,balanced,0.0437286376953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,32,8,balanced,0.03670912027359009
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,32,8,balanced,0.050335998535156246
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,32,8,balanced,0.05686528205871581
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,32,8,balanced,0.06173312187194824
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,32,8,balanced,0.02436608076095581
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,32,8,balanced,0.04616191864013672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,32,8,balanced,0.023144960403442383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,32,8,power_law_1.01,0.0373145604133606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,32,8,power_law_1.01,0.0349785590171814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,32,8,power_law_1.01,0.03941247940063476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,32,8,power_law_1.01,0.050895361900329596
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,32,8,power_law_1.01,0.05115007877349853
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,32,8,power_law_1.01,0.061144318580627445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,32,8,power_law_1.01,0.10402175903320313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,32,8,power_law_1.01,0.11408639907836915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,32,8,power_law_1.01,0.11025024414062501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,32,8,power_law_1.01,0.11800704002380372
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,32,8,power_law_1.01,0.12046976089477539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,32,8,power_law_1.01,0.1263756847381592
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,32,8,power_law_1.01,0.12717568397521972
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,32,8,power_law_1.01,0.13345791816711425
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,32,8,power_law_1.01,0.13883904457092283
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,32,8,power_law_1.01,0.14366975784301758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,32,8,power_law_1.01,0.15088512420654296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,32,8,power_law_1.01,0.17510400772094728
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,32,8,power_law_1.01,0.19155967712402344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,32,8,power_law_1.01,0.23112831115722657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,32,8,power_law_1.01,0.27771007537841796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,32,8,power_law_1.01,0.37424129486083985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,32,8,power_law_1.01,0.4726105499267578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,32,8,power_law_1.01,0.6210956954956055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,32,8,power_law_1.01,0.7345394897460937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,32,8,balanced,0.03618688106536865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,32,8,power_law_1.01,0.03622015953063965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,32,8,power_law_1.01,0.04168447971343994
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,32,8,power_law_1.01,0.04682752132415772
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,32,8,power_law_1.01,0.048483839035034185
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,32,8,balanced,0.03623680114746094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,32,8,power_law_1.01,1.4599781799316407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,32,8,power_law_1.01,0.051490559577941894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,32,8,power_law_1.01,0.05751423835754395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,32,8,power_law_1.01,0.06449535846710205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,32,8,power_law_1.01,1.0823500823974608
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,32,8,power_law_1.01,0.06781311988830566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,32,8,power_law_1.01,0.06908544063568114
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,32,8,power_law_1.01,0.07104383945465088
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,32,8,power_law_1.01,0.07085951805114746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,32,8,power_law_1.01,0.07550335884094238
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,32,8,power_law_1.01,0.07186944007873534
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,32,8,power_law_1.01,0.07828351974487305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,32,8,power_law_1.01,0.07887360095977783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,32,8,power_law_1.01,0.0832703971862793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,32,8,power_law_1.01,0.09677696228027344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,32,8,power_law_1.01,0.11902463912963868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,32,8,power_law_1.01,0.14850943565368652
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,32,8,power_law_1.01,0.17317632675170896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,32,8,power_law_1.01,0.23119232177734378
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,32,8,power_law_1.01,0.298545913696289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,32,8,power_law_1.01,0.42383361816406245
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,32,8,power_law_1.01,0.07998335838317872
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,32,8,power_law_1.01,0.5289791870117188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,32,8,power_law_1.01,0.025352959632873533
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,32,8,power_law_1.01,0.02846848011016846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,32,8,power_law_1.01,0.7533132934570312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,32,8,power_law_1.01,0.030254080295562747
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,32,8,power_law_1.01,0.03351295948028564
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,32,8,power_law_1.01,0.03480319976806641
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,32,8,power_law_1.01,0.03389695882797241
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,32,8,power_law_1.01,0.9749120330810547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,32,8,power_law_1.01,0.03661952018737793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,32,8,power_law_1.01,0.03523711919784546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,32,8,power_law_1.01,0.035828480720520015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,32,8,power_law_1.01,0.03650304079055786
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,32,8,power_law_1.01,0.03611648082733154
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,32,8,power_law_1.01,0.0362227201461792
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,32,8,power_law_1.01,0.0382809591293335
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,32,8,power_law_1.01,0.0403276777267456
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,32,8,power_law_1.01,0.04153984069824219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,32,8,power_law_1.01,0.04326655864715576
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,32,8,power_law_1.01,0.037372159957885745
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,32,8,balanced,0.08522751808166504
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,32,8,balanced,0.09540736198425293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,32,8,balanced,0.13287039756774904
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,32,8,balanced,0.16858495712280272
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,32,8,power_law_1.01,0.05294847965240479
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,32,8,power_law_1.01,0.07928063869476318
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,32,8,power_law_1.01,0.08711039543151855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2048,8,256,32,8,power_law_1.2,0.032156159877777105
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,32,8,power_law_1.01,0.057784318923950195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2048,8,256,32,8,power_law_1.2,0.03581568002700806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2048,8,256,32,8,power_law_1.2,0.03625087976455689
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2048,8,256,32,8,power_law_1.2,0.053059840202331544
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2048,8,256,32,8,power_law_1.2,0.05654272079467774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2048,8,256,32,8,power_law_1.2,0.062368001937866215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2048,8,256,32,8,power_law_1.2,0.11081215858459473
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2048,8,256,32,8,power_law_1.2,0.11862272262573241
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2048,8,256,32,8,power_law_1.2,0.11033472061157226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2048,8,256,32,8,power_law_1.2,0.11944448471069335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2048,8,256,32,8,power_law_1.2,0.1268607997894287
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2048,8,256,32,8,power_law_1.2,0.12919936180114747
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2048,8,256,32,8,power_law_1.2,0.12849920272827148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2048,8,256,32,8,power_law_1.2,0.134268159866333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2048,8,256,32,8,power_law_1.2,0.13971455574035646
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2048,8,256,32,8,power_law_1.2,0.14450559616088868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2048,8,256,32,8,power_law_1.2,0.1533555221557617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2048,8,256,32,8,power_law_1.2,0.1714828872680664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2048,8,256,32,8,power_law_1.2,0.20076160430908202
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2048,8,256,32,8,power_law_1.2,0.23836544036865231
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2048,8,256,32,8,power_law_1.2,0.29705215454101563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2048,8,256,32,8,power_law_1.2,0.35797374725341796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2048,8,256,32,8,power_law_1.2,0.4653235244750976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2048,8,256,32,8,power_law_1.2,0.640216293334961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2048,8,256,32,8,power_law_1.2,0.7694924926757812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2048,8,256,32,8,power_law_1.2,1.207294692993164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2048,8,256,32,8,power_law_1.2,0.03623424053192138
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2048,8,256,32,8,power_law_1.2,0.04162303924560547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2048,8,256,32,8,power_law_1.2,1.5652056884765624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2048,8,256,32,8,power_law_1.2,0.04439295768737793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2048,8,256,32,8,power_law_1.2,0.047248640060424806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2048,8,256,32,8,power_law_1.2,0.05182591915130615
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2048,8,256,32,8,power_law_1.2,0.054200320243835455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2048,8,256,32,8,power_law_1.2,0.06686336040496826
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2048,8,256,32,8,power_law_1.2,0.06699647903442382
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2048,8,256,32,8,power_law_1.2,0.0692748785018921
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2048,8,256,32,8,power_law_1.2,0.06969855785369873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2048,8,256,32,8,power_law_1.2,0.073853440284729
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2048,8,256,32,8,power_law_1.2,0.07146495819091797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2048,8,256,32,8,power_law_1.2,0.07166719913482666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2048,8,256,32,8,power_law_1.2,0.07663871765136718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2048,8,256,32,8,power_law_1.2,0.07893119812011719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2048,8,256,32,8,power_law_1.2,0.08168576240539552
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2048,8,256,32,8,power_law_1.2,0.09032575607299805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2048,8,256,32,8,power_law_1.2,0.10235136032104493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2048,8,256,32,8,power_law_1.2,0.11966208457946777
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2048,8,256,32,8,power_law_1.2,0.15320704460144044
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,32,8,power_law_1.01,0.12572928428649904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2048,8,256,32,8,power_law_1.2,0.18047615051269533
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,32,8,power_law_1.01,0.15856127738952636
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2048,8,256,32,8,power_law_1.2,0.23508607864379885
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,32,8,power_law_1.01,0.21624448776245114
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2048,8,256,32,8,power_law_1.2,0.3262246322631836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2048,8,256,32,8,power_law_1.2,0.41160831451416013
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2048,8,256,32,8,power_law_1.2,0.025521919727325437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2048,8,256,32,8,power_law_1.2,0.576703987121582
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,32,8,power_law_1.01,0.24760704040527343
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2048,8,256,32,8,power_law_1.2,0.02844543933868408
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2048,8,256,32,8,power_law_1.2,0.028866560459136964
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2048,8,256,32,8,power_law_1.2,0.031252479553222655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2048,8,256,32,8,power_law_1.2,0.8849651336669921
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2048,8,256,32,8,power_law_1.2,0.032927999496459956
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2048,8,256,32,8,power_law_1.2,0.033890559673309326
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2048,8,256,32,8,power_law_1.2,0.0366425609588623
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2048,8,256,32,8,power_law_1.2,0.03536256074905396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2048,8,256,32,8,power_law_1.2,1.0898432159423828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2048,8,256,32,8,power_law_1.2,0.03608959913253784
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2048,8,256,32,8,power_law_1.2,0.03543168067932129
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2048,8,256,32,8,power_law_1.2,0.03466111898422241
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2048,8,256,32,8,power_law_1.2,0.036883199214935304
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2048,8,256,32,8,power_law_1.2,0.03809792041778565
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2048,8,256,32,8,power_law_1.2,0.03966720104217529
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2048,8,256,32,8,power_law_1.2,0.04153984069824219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2048,8,256,32,8,power_law_1.2,0.04355584144592285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2048,8,256,32,8,power_law_1.2,0.04877312183380127
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2048,8,256,32,8,power_law_1.2,0.05606400012969971
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,32,8,power_law_1.01,0.39393024444580077
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,32,8,power_law_1.01,0.44998527526855464
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2048,8,256,32,8,power_law_1.2,0.14746879577636718
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2048,8,256,32,8,power_law_1.2,0.11158656120300292
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2048,8,256,32,8,power_law_1.2,0.07523583889007568
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,128,balanced,0.017834240198135377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,128,balanced,0.01774719953536987
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,128,balanced,0.017766400575637817
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2048,8,256,32,8,power_law_1.2,0.30344192504882816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,128,balanced,0.01734272003173828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,128,balanced,0.018173439502716066
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2048,8,256,32,8,power_law_1.2,0.09882752418518066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,128,balanced,0.028494079113006592
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,128,balanced,0.02765183925628662
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,128,balanced,0.03001215934753418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,128,balanced,0.030275840759277344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,128,balanced,0.03007999897003174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,128,balanced,0.030854399204254153
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,128,balanced,0.030918400287628173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,128,balanced,0.031366400718688965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,128,balanced,0.03256319999694825
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,128,balanced,0.03293823957443237
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,128,balanced,0.03499392032623291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,128,balanced,0.034714879989624026
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,128,balanced,0.03660032033920288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,128,balanced,0.04028672218322754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,128,balanced,0.035964159965515136
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,128,balanced,0.040320000648498534
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,128,balanced,0.043136000633239746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,128,balanced,0.04734848022460937
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2048,8,256,32,8,power_law_1.2,0.21876607894897462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,128,balanced,0.05354239940643311
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,128,balanced,0.06509183883666993
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,128,balanced,0.08206848144531251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,128,balanced,0.023110399246215822
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,128,balanced,0.022725119590759277
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,128,balanced,0.022691841125488284
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,128,balanced,0.02269695997238159
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,128,balanced,0.026835200786590574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,128,balanced,0.027641599178314206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,128,balanced,0.03264512062072754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,128,balanced,0.028049919605255123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,128,balanced,0.028522241115570068
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,128,balanced,0.02803584098815918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,128,balanced,0.029141759872436522
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,128,balanced,0.028506879806518552
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,128,balanced,0.029679360389709475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,128,balanced,0.02928895950317383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,128,balanced,0.022649600505828857
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,128,balanced,0.031383039951324465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,128,balanced,0.030612480640411378
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,128,balanced,0.03274240016937256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,128,balanced,0.03494784116744995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,128,balanced,0.03214591979980469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,128,balanced,0.02760960102081299
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,128,balanced,0.0337779188156128
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,128,balanced,0.03465856075286865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,128,balanced,0.03783423900604248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,128,balanced,0.0407206392288208
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,128,balanced,0.04651904106140137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,128,balanced,0.05589632034301758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,128,balanced,0.030099198818206784
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2048,8,256,32,8,power_law_1.2,0.42328319549560545
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2048,8,256,32,8,power_law_1.2,0.5420940780639649
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,2,128,balanced,0.01372032046318054
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,2,128,balanced,0.014067200422286989
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,2,128,balanced,0.013790719509124754
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,2,128,balanced,0.01371392011642456
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,2,128,balanced,0.013701119422912597
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,2,128,balanced,0.014100480079650878
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,2,128,balanced,0.013679360151290893
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,2,128,balanced,0.014172159433364867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,2,128,balanced,0.01405951976776123
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,2,128,balanced,0.013674240112304687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,2,128,balanced,0.014067200422286989
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,2,128,balanced,0.014142719507217408
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,2,128,balanced,0.014095360040664673
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,2,128,balanced,0.013940479755401611
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,2,128,balanced,0.014131200313568116
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,2,128,balanced,0.014553600549697876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,2,128,balanced,0.014520319700241088
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,2,128,balanced,0.013968640565872192
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,2,128,balanced,0.014095360040664673
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2048,8,256,32,8,power_law_1.2,0.7692479705810547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,2,128,balanced,0.0142195200920105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,128,power_law_1.01,0.01732223987579346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,128,power_law_1.01,0.017392640113830564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,128,power_law_1.01,0.02763391971588135
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,128,power_law_1.01,0.02811903953552246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,128,power_law_1.01,0.030109438896179196
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,2,128,balanced,0.014092799425125122
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,128,power_law_1.01,0.03084800004959106
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,128,power_law_1.01,0.03254400014877319
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,128,power_law_1.01,0.03337215900421143
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,128,power_law_1.01,0.03461503982543945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,128,power_law_1.01,0.03457407951354981
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,128,power_law_1.01,0.03475840091705322
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,128,power_law_1.01,0.03623552083969116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,128,power_law_1.01,0.031354880332946776
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,128,power_law_1.01,0.038691840171813964
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,128,power_law_1.01,0.04048895835876465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,128,power_law_1.01,0.04077568054199219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,128,power_law_1.01,0.0404095983505249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,128,power_law_1.01,0.044766721725463865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,128,power_law_1.01,0.050392317771911624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,128,power_law_1.01,0.06701568126678467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,128,power_law_1.01,0.0841420841217041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,128,power_law_1.01,0.11304191589355468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,128,power_law_1.01,0.1481663990020752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,128,power_law_1.01,0.21216896057128903
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,128,power_law_1.01,0.2740006446838379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,128,power_law_1.01,0.39078910827636715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,128,power_law_1.01,0.023136000633239746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,128,power_law_1.01,0.5706329727172852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,128,power_law_1.01,0.02346112012863159
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,128,power_law_1.01,0.02802432060241699
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,128,power_law_1.01,0.02797055959701538
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,128,power_law_1.01,0.028088319301605224
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,128,power_law_1.01,0.028039679527282712
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,128,power_law_1.01,0.028471040725708007
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,128,power_law_1.01,0.02967423915863037
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,128,power_law_1.01,0.0293503999710083
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,128,power_law_1.01,0.03134207963943482
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,128,power_law_1.01,0.030069758892059328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,128,power_law_1.01,0.0309171199798584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,128,power_law_1.01,0.033720319271087644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,128,power_law_1.01,0.031769599914550785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,128,power_law_1.01,0.035767040252685546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,128,power_law_1.01,0.035549440383911134
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,128,power_law_1.01,0.03329792022705078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,128,power_law_1.01,0.03620991945266724
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,128,power_law_1.01,0.038609919548034666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,128,power_law_1.01,0.04812928199768066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,128,power_law_1.01,0.05591807842254639
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,128,power_law_1.01,0.06737279891967773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,128,power_law_1.01,0.08864128112792968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,128,power_law_1.01,0.12722816467285156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,128,power_law_1.01,0.16540416717529297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,128,power_law_1.01,0.24754688262939456
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,128,power_law_1.01,0.342663688659668
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,2,128,power_law_1.01,0.0140774405002594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,2,128,power_law_1.01,0.014104319810867311
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,2,128,power_law_1.01,0.014254080057144163
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,2,128,power_law_1.01,0.014083839654922487
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,2,128,power_law_1.01,0.014117120504379271
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,2,128,power_law_1.01,0.01443071961402893
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,2,128,power_law_1.01,0.014508800506591795
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,2,128,power_law_1.01,0.014118399620056152
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,2,128,balanced,0.014565119743347168
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,2,128,power_law_1.01,0.014172159433364867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,2,128,balanced,0.014940160512924194
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,2,128,power_law_1.01,0.014453760385513307
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,2,128,power_law_1.01,0.014183679819107056
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,2,128,power_law_1.01,0.014540799856185912
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,2,128,power_law_1.01,0.01437183976173401
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,2,128,power_law_1.01,0.014073599576950074
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,2,128,balanced,0.0156876802444458
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,2,128,power_law_1.01,0.015281920433044433
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,2,128,balanced,0.014911999702453613
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,2,128,power_law_1.01,0.014168319702148439
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,2,128,balanced,0.016605440378189087
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,2,128,balanced,0.017099519968032835
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,2,128,power_law_1.01,0.014583040475845337
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,2,128,power_law_1.01,0.014571520090103148
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,2,128,power_law_1.01,0.016605440378189087
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,2,128,power_law_1.01,0.01776255965232849
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,2,128,power_law_1.01,0.01941632032394409
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,128,power_law_1.2,0.01742848038673401
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,2,128,power_law_1.01,0.014899200201034546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,128,power_law_1.2,0.01734272003173828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,128,power_law_1.2,0.027564799785614012
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,128,power_law_1.2,0.028816640377044678
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,128,power_law_1.2,0.030903038978576658
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,128,power_law_1.2,0.030846719741821287
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,128,power_law_1.2,0.03174911975860596
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,128,power_law_1.2,0.03335295915603638
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,128,power_law_1.2,0.03415296077728271
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,128,power_law_1.2,0.03458048105239868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,128,power_law_1.2,0.03617919921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,128,power_law_1.2,0.038613760471343996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,128,power_law_1.2,0.04069632053375244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,128,power_law_1.2,0.04036736011505127
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,2,128,power_law_1.01,0.025681920051574707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,128,power_law_1.2,0.040656638145446775
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,128,power_law_1.2,0.04403200149536133
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,2,128,power_law_1.01,0.020657920837402345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,128,power_law_1.2,0.05364480018615723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,128,power_law_1.2,0.06664447784423828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,128,power_law_1.2,0.0930726432800293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,128,power_law_1.2,0.09935615539550782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,128,power_law_1.2,0.03337599992752075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,128,power_law_1.2,0.16588415145874022
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,128,power_law_1.2,0.2316991996765137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,128,power_law_1.2,0.02242687940597534
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,128,power_law_1.2,0.02265984058380127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,128,power_law_1.2,0.027180800437927245
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,128,power_law_1.2,0.02727807998657227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,128,power_law_1.2,0.02798464059829712
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,128,power_law_1.2,0.028017919063568115
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,128,power_law_1.2,0.028409600257873535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,128,power_law_1.2,0.029231359958648683
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,128,power_law_1.2,0.030122239589691163
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,128,power_law_1.2,0.030520319938659668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,128,power_law_1.2,0.03044991970062256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,128,power_law_1.2,0.03174911975860596
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,128,power_law_1.2,0.3559692764282226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,128,power_law_1.2,0.03337471961975098
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,128,power_law_1.2,0.03533951997756958
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,128,power_law_1.2,0.03460864067077637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,128,power_law_1.2,0.033369600772857666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,128,power_law_1.2,0.03538304090499878
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,128,power_law_1.2,0.03903743982315063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,128,power_law_1.2,0.04463232040405273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,128,power_law_1.2,0.06108160018920898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,128,power_law_1.2,0.06772096157073974
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,128,power_law_1.2,0.11010432243347168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,128,power_law_1.2,0.14492927551269533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,128,power_law_1.2,0.220263671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,128,power_law_1.2,0.6590246582031251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,128,power_law_1.2,0.4259545516967774
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,2,128,power_law_1.2,0.014072320461273194
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,2,128,power_law_1.2,0.014191360473632814
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,2,128,power_law_1.2,0.014128639698028564
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,2,128,power_law_1.2,0.014118399620056152
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,2,128,power_law_1.2,0.014119679927825928
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,128,power_law_1.2,0.4713523101806641
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,2,128,power_law_1.2,0.014124799966812134
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,2,128,power_law_1.2,0.014092799425125122
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,2,128,power_law_1.01,0.030739200115203858
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,2,128,power_law_1.01,0.0572057580947876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,2,128,power_law_1.2,0.014444799423217775
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,2,128,power_law_1.2,0.01415295958518982
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,2,128,power_law_1.2,0.014202879667282103
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,2,128,power_law_1.01,0.03974656105041504
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,2,128,power_law_1.2,0.014536960124969484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,2,128,power_law_1.2,0.01454208016395569
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,2,128,power_law_1.2,0.01423359990119934
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,2,128,power_law_1.2,0.014521600008010865
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,2,128,power_law_1.2,0.014511359930038454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,128,power_law_1.2,0.3001932716369629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,128,power_law_1.2,0.8848703765869141
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,2,128,power_law_1.2,0.014892799854278566
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,2,128,power_law_1.2,0.014551039934158325
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,2,128,power_law_1.2,0.015411200523376464
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,2,128,power_law_1.2,0.016839679479599
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,2,128,power_law_1.2,0.01783936023712158
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,2,128,power_law_1.2,0.01980288028717041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,4,64,balanced,0.018293759822845458
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,4,64,balanced,0.018615039587020873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,64,balanced,0.018205440044403075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,64,balanced,0.018612480163574217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,64,balanced,0.018145279884338377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,64,balanced,0.028464639186859132
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,64,balanced,0.028460800647735596
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,64,balanced,0.02892159938812256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,64,balanced,0.029720320701599124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,64,balanced,0.030229759216308594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,64,balanced,0.03047679901123047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,64,balanced,0.030967040061950685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,64,balanced,0.03133055925369262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,64,balanced,0.03216255903244018
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,64,balanced,0.032545280456542966
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,64,balanced,0.03293823957443237
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,64,balanced,0.036119039058685305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,64,balanced,0.038618879318237306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,64,balanced,0.03539455890655517
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,64,balanced,0.039007999897003175
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,64,balanced,0.041196799278259276
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,64,balanced,0.04456448078155518
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,64,balanced,0.050655999183654786
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,128,power_law_1.2,0.587347183227539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,64,balanced,0.060392961502075196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,64,balanced,0.07319295883178711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,64,balanced,0.10466048240661621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,64,balanced,0.12854399681091308
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,64,balanced,0.022338559627532957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,64,balanced,0.02188800096511841
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,64,balanced,0.022389760017395018
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,64,balanced,0.022307839393615723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,64,balanced,0.02236288070678711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,64,balanced,0.02754816055297852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,64,balanced,0.02811392068862915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,64,balanced,0.02800512075424194
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,64,balanced,0.028206079006195067
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,64,balanced,0.028456959724426266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,64,balanced,0.028504319190979004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,64,balanced,0.029283840656280518
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,64,balanced,0.02968575954437256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,64,balanced,0.031307520866394045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,64,balanced,0.030085120201110836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,64,balanced,0.030653440952301027
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,64,balanced,0.032099840641021726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,64,balanced,0.034595839977264405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,64,balanced,0.030920960903167725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,64,balanced,0.032133119106292726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,64,balanced,0.03542655944824219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,64,balanced,0.03772799968719483
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,64,balanced,0.04024831771850586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,64,balanced,0.04560128211975097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,64,balanced,0.052206082344055174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,64,balanced,0.07294591903686523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,64,balanced,0.0891596794128418
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,2,128,power_law_1.2,0.023089919090270996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,2,128,power_law_1.2,0.027536640167236327
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,2,128,power_law_1.2,0.036487679481506344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,2,128,power_law_1.2,0.07923711776733398
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,2,128,power_law_1.2,0.04648320198059082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,4,64,balanced,0.014110720157623291
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,4,64,balanced,0.014536960124969484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,4,64,balanced,0.014533120393753051
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,4,64,balanced,0.01454208016395569
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,4,64,balanced,0.014517760276794434
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,4,64,balanced,0.01451647996902466
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,4,64,balanced,0.01452415943145752
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,4,64,balanced,0.01410815954208374
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,4,64,balanced,0.014792959690093994
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,4,64,balanced,0.014949120283126833
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,4,64,balanced,0.0145305597782135
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,4,64,balanced,0.014528000354766845
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,4,64,balanced,0.014855680465698243
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,4,64,balanced,0.014992640018463135
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,4,64,balanced,0.014853119850158691
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,4,64,balanced,0.014941439628601075
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,4,64,balanced,0.014062080383300781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,4,64,balanced,0.014519040584564208
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,4,64,balanced,0.014469120502471924
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,4,64,balanced,0.014641920328140259
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,2,128,power_law_1.2,0.0640883207321167
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,4,64,power_law_1.01,0.01894912004470825
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,4,64,power_law_1.01,0.018236160278320312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,64,power_law_1.01,0.018234879970550538
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,64,power_law_1.01,0.027654399871826173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,64,power_law_1.01,0.02922624111175537
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,64,power_law_1.01,0.029922559261322024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,64,power_law_1.01,0.03133055925369262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,64,power_law_1.01,0.03378943920135498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,64,power_law_1.01,0.033351678848266605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,64,power_law_1.01,0.033614079952239986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,64,power_law_1.01,0.035777280330657954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,64,power_law_1.01,0.03706624031066895
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,4,64,balanced,0.014945280551910401
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,64,power_law_1.01,0.03790591955184937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,64,power_law_1.01,0.0391590404510498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,64,power_law_1.01,0.03748863935470581
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,64,power_law_1.01,0.03940351963043213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,64,power_law_1.01,0.04037375926971436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,64,power_law_1.01,0.045139198303222654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,64,power_law_1.01,0.05065216064453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,64,power_law_1.01,0.0680844783782959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,64,power_law_1.01,0.07563519954681397
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,64,power_law_1.01,0.11378560066223145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,64,power_law_1.01,0.1507571220397949
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,64,power_law_1.01,0.20114303588867188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,64,power_law_1.01,0.2595302391052246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,64,power_law_1.01,0.38572032928466793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,4,64,balanced,0.015595519542694091
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,64,power_law_1.01,0.023184640407562258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,64,power_law_1.01,0.541454086303711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,64,power_law_1.01,0.022734079360961914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,64,power_law_1.01,0.02310784101486206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,64,power_law_1.01,0.02765183925628662
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,64,power_law_1.01,0.027998719215393063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,64,power_law_1.01,0.028456959724426266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,64,power_law_1.01,0.03011840105056763
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,64,power_law_1.01,0.029765119552612306
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,4,64,balanced,0.015994880199432373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,64,power_law_1.01,0.030507519245147708
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,64,power_law_1.01,0.0309171199798584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,64,power_law_1.01,0.03215487957000732
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,64,power_law_1.01,0.0326963210105896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,64,power_law_1.01,0.03503871917724609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,64,power_law_1.01,0.033405439853668215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,64,power_law_1.01,0.03218559980392456
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,64,power_law_1.01,0.03342848062515259
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,64,power_law_1.01,0.03622015953063965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,64,power_law_1.01,0.03922944068908692
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,64,power_law_1.01,0.04951168060302734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,64,power_law_1.01,0.06226943969726563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,4,64,balanced,0.01660416007041931
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,64,power_law_1.01,0.030103039741516114
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,64,power_law_1.01,0.07769087791442872
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,64,power_law_1.01,0.09958016395568847
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,64,power_law_1.01,0.13609600067138672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,64,power_law_1.01,0.1712985610961914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,4,64,power_law_1.01,0.014666240215301513
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,64,power_law_1.01,0.2629017639160156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,4,64,power_law_1.01,0.014515199661254882
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,4,64,power_law_1.01,0.014049279689788818
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,4,64,power_law_1.01,0.014594559669494628
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,64,power_law_1.01,0.40387710571289065
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,4,64,power_law_1.01,0.014519040584564208
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,4,64,power_law_1.01,0.014545919895172118
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,4,64,power_law_1.01,0.014803199768066405
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,4,64,power_law_1.01,0.014918400049209597
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,4,64,power_law_1.01,0.014612480401992797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,4,64,power_law_1.01,0.014923520088195802
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,4,64,power_law_1.01,0.014488320350646972
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,4,64,power_law_1.01,0.014913280010223389
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,4,64,power_law_1.01,0.014895360469818115
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,4,64,power_law_1.01,0.015027199983596801
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,4,64,balanced,0.017388800382614138
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,4,64,balanced,0.019411200284957887
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,4,64,balanced,0.020290560722351074
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,4,64,power_law_1.01,0.015215359926223753
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,4,64,power_law_1.01,0.014929920434951782
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,4,64,power_law_1.01,0.015397119522094726
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,4,64,power_law_1.01,0.015792640447616576
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,4,64,power_law_1.01,0.016391680240631104
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,4,64,power_law_1.01,0.016999679803848266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,4,64,power_law_1.2,0.017858560085296633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,4,64,power_law_1.2,0.018202879428863526
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,4,64,power_law_1.01,0.019004160165786745
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,64,power_law_1.2,0.01822335958480835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,64,power_law_1.2,0.028055040836334227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,64,power_law_1.2,0.030086400508880617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,64,power_law_1.2,0.03011199951171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,64,power_law_1.2,0.032133119106292726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,64,power_law_1.2,0.032020480632781984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,64,power_law_1.2,0.0333401608467102
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,64,power_law_1.2,0.03291264057159424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,64,power_law_1.2,0.03585792064666748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,64,power_law_1.2,0.037067520618438723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,64,power_law_1.2,0.0374617600440979
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,64,power_law_1.2,0.03761919975280762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,64,power_law_1.2,0.03943808078765869
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,64,power_law_1.2,0.039882240295410154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,64,power_law_1.2,0.04314367771148682
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,64,power_law_1.2,0.05018879890441894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,64,power_law_1.2,0.06692607879638672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,64,power_law_1.2,0.08825599670410156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,64,power_law_1.2,0.11716480255126953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,64,power_law_1.2,0.15706239700317384
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,4,64,power_law_1.01,0.02065664052963257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,64,power_law_1.2,0.2077785682678223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,64,power_law_1.2,0.3871257781982422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,64,power_law_1.2,0.31100543975830075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,64,power_law_1.2,0.022305281162261964
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,64,power_law_1.2,0.022736639976501466
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,64,power_law_1.2,0.027232000827789305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,64,power_law_1.2,0.027644159793853762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,64,power_law_1.2,0.02797312021255493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,64,power_law_1.2,0.028401920795440672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,64,power_law_1.2,0.0288319993019104
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,64,power_law_1.2,0.028974080085754396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,64,power_law_1.2,0.029665279388427734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,64,power_law_1.2,0.030507519245147708
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,64,power_law_1.2,0.031315200328826905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,64,power_law_1.2,0.032143359184265134
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,64,power_law_1.2,0.03371776103973388
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,64,power_law_1.2,0.034190080165863036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,64,power_law_1.2,0.03250175952911377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,64,power_law_1.2,0.033496320247650146
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,4,64,power_law_1.01,0.024006400108337402
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,64,power_law_1.2,0.03579391956329346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,64,power_law_1.2,0.03947135925292969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,64,power_law_1.2,0.044381442070007326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,64,power_law_1.2,0.059932160377502444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,64,power_law_1.2,0.07132031917572021
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,4,64,power_law_1.01,0.029943039417266847
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,64,power_law_1.2,0.10620927810668945
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,4,64,power_law_1.01,0.03329407930374145
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,4,64,power_law_1.01,0.049684481620788576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,64,power_law_1.2,0.17877248764038084
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,4,64,power_law_1.2,0.014565119743347168
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,4,64,power_law_1.2,0.01456768035888672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,64,power_law_1.2,0.5681280136108399
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,4,64,power_law_1.2,0.01467519998550415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,64,power_law_1.2,0.12929535865783692
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,4,64,power_law_1.2,0.014476799964904785
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,4,64,power_law_1.2,0.014553600549697876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,4,64,power_law_1.2,0.014511359930038454
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,4,64,power_law_1.2,0.014531840085983277
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,4,64,power_law_1.2,0.014544639587402344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,4,64,power_law_1.2,0.014854400157928466
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,4,64,power_law_1.2,0.014862079620361329
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,4,64,power_law_1.2,0.014536960124969484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,4,64,power_law_1.2,0.01496832013130188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,4,64,power_law_1.2,0.014763519763946534
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,4,64,power_law_1.2,0.015054080486297607
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,4,64,power_law_1.01,0.05498367786407471
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,64,power_law_1.2,0.253757438659668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,64,power_law_1.2,0.7894976043701172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,64,power_law_1.2,0.43759616851806643
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,64,power_law_1.2,0.5671667098999024
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,4,64,power_law_1.2,0.015345920324325562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,4,64,power_law_1.2,0.015859199762344362
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,4,64,power_law_1.2,0.014915839433670045
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,4,64,power_law_1.2,0.016558079719543456
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,4,64,power_law_1.2,0.017329920530319214
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,4,64,power_law_1.2,0.019057919979095457
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,4,64,power_law_1.2,0.02023808002471924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,128,balanced,0.024316160678863524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,128,balanced,0.024479999542236328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,128,balanced,0.02413055896759033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,128,balanced,0.024729599952697752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,128,balanced,0.04271743774414062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,128,balanced,0.04439167976379395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,128,balanced,0.04562560081481933
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,128,balanced,0.028107519149780276
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,128,balanced,0.04658432006835937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,128,balanced,0.04773248195648193
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,128,balanced,0.04821248054504394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,128,balanced,0.04851327896118164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,128,balanced,0.04935296058654785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,128,balanced,0.049730558395385746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,128,balanced,0.05185152053833007
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,128,balanced,0.05167871952056885
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,128,balanced,0.05584512233734131
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,4,64,power_law_1.2,0.024924159049987793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,128,balanced,0.05465472221374512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,128,balanced,0.057253122329711914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,128,balanced,0.06290431976318359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,128,balanced,0.0662553596496582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,128,balanced,0.07262335777282715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,128,balanced,0.09033472061157226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,128,balanced,0.10595583915710449
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,128,balanced,0.13339648246765137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,128,balanced,0.15671680450439454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,128,balanced,0.02807552099227905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,128,balanced,0.02797312021255493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,128,balanced,0.027699201107025145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,128,balanced,0.028066558837890627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,128,balanced,0.03459455966949463
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,128,balanced,0.05982079982757569
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,128,balanced,0.03501568078994751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,128,balanced,0.03518079996109009
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,128,balanced,0.03544064044952393
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,128,balanced,0.03540607929229737
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,128,balanced,0.035795199871063235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,128,balanced,0.03594624042510987
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,128,balanced,0.03664383888244629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,128,balanced,0.036239359378814694
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,128,balanced,0.03788671970367431
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,128,balanced,0.03743616104125976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,128,balanced,0.028023040294647215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,128,balanced,0.03827712059020996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,128,balanced,0.040675840377807616
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,128,balanced,0.039622399806976315
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,128,balanced,0.042803201675415035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,128,balanced,0.040433921813964845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,128,balanced,0.044904961585998535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,128,balanced,0.04725632190704346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,128,balanced,0.051799039840698245
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,128,balanced,0.058373122215271
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,128,balanced,0.07432703971862793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,128,balanced,0.09079296112060546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,4,64,power_law_1.2,0.029591040611267088
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,4,64,power_law_1.2,0.03926016092300415
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,4,64,power_law_1.2,0.053704957962036136
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,2,128,balanced,0.015368959903717042
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,2,128,balanced,0.015384320020675659
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,2,128,balanced,0.01574015974998474
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,2,128,balanced,0.01536128044128418
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,2,128,balanced,0.0153766405582428
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,2,128,balanced,0.015717120170593263
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,2,128,balanced,0.01536512017250061
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,2,128,balanced,0.015717120170593263
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,2,128,balanced,0.015802880525588987
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,2,128,balanced,0.015773439407348634
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,2,128,balanced,0.015299839973449709
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,2,128,balanced,0.01617535948753357
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,2,128,balanced,0.016153600215911865
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,2,128,balanced,0.015801600217819213
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,2,128,balanced,0.015777280330657957
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,2,128,balanced,0.016157439947128295
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,2,128,balanced,0.016148480176925658
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,2,128,balanced,0.01553536057472229
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,2,128,balanced,0.01592319965362549
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,2,128,balanced,0.015402239561080933
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,4,64,power_law_1.2,0.06359295845031739
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,2,128,balanced,0.015773439407348634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,128,power_law_1.01,0.026403839588165286
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,128,power_law_1.01,0.02661247968673706
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,128,power_law_1.01,0.026431999206542968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,128,power_law_1.01,0.04293375968933105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,128,power_law_1.01,0.04692351818084717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,128,power_law_1.01,0.04834432125091553
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,128,power_law_1.01,0.050367999076843264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,128,power_law_1.01,0.051490559577941894
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,4,64,power_law_1.2,0.08577280044555664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,128,power_law_1.01,0.05348351955413818
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,128,power_law_1.01,0.05200128078460693
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,128,power_law_1.01,0.05512191772460937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,128,power_law_1.01,0.054384641647338874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,128,power_law_1.01,0.05606143951416016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,128,power_law_1.01,0.060526080131530755
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,128,power_law_1.01,0.0643174409866333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,128,power_law_1.01,0.06534399986267089
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,128,power_law_1.01,0.06669695854187012
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,128,power_law_1.01,0.0883187198638916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,128,power_law_1.01,0.10026111602783203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,128,power_law_1.01,0.12091903686523438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,128,power_law_1.01,0.1515660762786865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,128,power_law_1.01,0.19567232131958007
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,128,power_law_1.01,0.24526336669921872
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,128,power_law_1.01,0.40308864593505855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,128,power_law_1.01,0.5615475082397461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,128,power_law_1.01,0.7831436920166015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,128,power_law_1.01,0.028006401062011722
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,128,power_law_1.01,0.027621119022369384
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,128,power_law_1.01,0.9638579559326171
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,128,power_law_1.01,0.02768383979797363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,128,power_law_1.01,0.03502079963684082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,128,power_law_1.01,0.03502336025238037
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,128,power_law_1.01,0.03586303949356079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,128,power_law_1.01,0.03559808015823364
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,128,power_law_1.01,0.038754560947418214
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,128,power_law_1.01,0.036587519645690916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,128,power_law_1.01,0.03742079973220826
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,128,power_law_1.01,0.03790848016738892
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,128,power_law_1.01,0.039047679901123046
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,2,128,balanced,0.016618239879608154
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,2,128,balanced,0.01690623998641968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,128,power_law_1.01,0.04117248058319092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,128,power_law_1.01,0.04436863899230957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,128,power_law_1.01,0.04437376022338867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,128,power_law_1.01,0.04817920207977295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,128,power_law_1.01,0.05648255825042725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,128,power_law_1.01,0.04084352016448974
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,2,128,balanced,0.017907199859619142
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,128,power_law_1.01,0.06775551795959472
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,128,power_law_1.01,0.08392064094543457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,128,power_law_1.01,0.11764479637145997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,128,power_law_1.01,0.1623423957824707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,128,power_law_1.01,0.22787967681884766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,128,power_law_1.01,0.3171033668518066
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,2,128,power_law_1.01,0.01580415964126587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,128,power_law_1.01,0.044431362152099606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,128,power_law_1.01,0.43219585418701173
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,2,128,power_law_1.01,0.015742720365524293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,2,128,power_law_1.01,0.01566848039627075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,128,power_law_1.01,0.6169484710693359
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,2,128,power_law_1.01,0.015763200521469116
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,2,128,power_law_1.01,0.016136959791183472
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,2,128,balanced,0.020689918994903564
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,2,128,power_law_1.01,0.015523840188980103
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,2,128,power_law_1.01,0.015615999698638916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,2,128,power_law_1.01,0.0158297598361969
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,2,128,balanced,0.018911999464035035
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,2,128,power_law_1.01,0.015758080482482912
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,2,128,power_law_1.01,0.01576575994491577
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,2,128,power_law_1.01,0.016095999479293823
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,2,128,power_law_1.01,0.016179200410842896
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,2,128,power_law_1.01,0.016176639795303344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,2,128,power_law_1.01,0.016595200300216675
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,2,128,power_law_1.01,0.018549760580062864
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,2,128,power_law_1.01,0.01615231990814209
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,2,128,balanced,0.02222208023071289
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,2,128,power_law_1.01,0.015408639907836913
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,2,128,power_law_1.01,0.016523519754409792
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,2,128,power_law_1.01,0.017799680233001706
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,2,128,power_law_1.01,0.02025984048843384
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,2,128,power_law_1.01,0.0214899206161499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,128,power_law_1.2,0.024892160892486574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,128,power_law_1.2,0.0247654390335083
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,128,power_law_1.2,0.04200960159301757
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,128,power_law_1.2,0.04523647785186767
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,128,power_law_1.2,0.048522238731384275
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,128,power_law_1.2,0.050557441711425774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,128,power_law_1.2,0.05222271919250489
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,2,128,power_law_1.01,0.02448512077331543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,128,power_law_1.2,0.054657278060913084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,128,power_law_1.2,0.05333504199981689
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,128,power_law_1.2,0.05377024173736572
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,128,power_law_1.2,0.056352000236511234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,2,128,power_law_1.01,0.03113215923309326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,128,power_law_1.2,0.06033279895782471
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,128,power_law_1.2,0.064170241355896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,128,power_law_1.2,0.048060159683227535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,128,power_law_1.2,0.06641407966613769
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,128,power_law_1.2,0.06845439910888672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,128,power_law_1.2,0.07657472133636475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,128,power_law_1.2,0.1037491226196289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,128,power_law_1.2,0.12151807785034179
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,128,power_law_1.2,0.15033984184265137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,128,power_law_1.2,0.20976255416870115
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,128,power_law_1.2,0.3325388717651367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,128,power_law_1.2,0.45102977752685547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,2,128,power_law_1.01,0.043380479812622066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,128,power_law_1.2,0.02808703899383545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,128,power_law_1.2,0.027655680179595948
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,128,power_law_1.2,0.034639360904693606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,128,power_law_1.2,0.03482239961624146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,128,power_law_1.2,0.035411200523376464
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,128,power_law_1.2,0.035101439952850345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,128,power_law_1.2,0.03583103895187378
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,128,power_law_1.2,0.03647104024887085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,128,power_law_1.2,0.03713023900985718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,128,power_law_1.2,0.7143334197998047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,128,power_law_1.2,0.037546238899230956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,128,power_law_1.2,1.3890188598632813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,128,power_law_1.2,0.037871360778808594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,128,power_law_1.2,0.03949951887130737
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,128,power_law_1.2,0.04033535957336426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,128,power_law_1.2,0.04440959930419922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,128,power_law_1.2,0.045181441307067874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,128,power_law_1.2,0.04746880054473877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,128,power_law_1.2,0.057955842018127444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,128,power_law_1.2,0.06709504127502441
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,128,power_law_1.2,0.09282688140869141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,128,power_law_1.2,0.12332672119140624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,128,power_law_1.2,0.16344959259033204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,128,power_law_1.2,0.04483071804046631
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,128,power_law_1.2,0.2546739196777344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,128,power_law_1.2,0.3783795166015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,128,power_law_1.2,0.9443532562255859
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,2,128,power_law_1.2,0.015763200521469116
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,2,128,power_law_1.2,0.015793919563293457
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,2,128,power_law_1.2,0.015742720365524293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,2,128,power_law_1.2,0.015837440490722658
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,2,128,power_law_1.2,0.015767040252685546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,2,128,power_law_1.2,0.01618559956550598
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,2,128,power_law_1.2,0.016161279678344725
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,2,128,power_law_1.2,0.015778559446334838
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,2,128,power_law_1.2,0.01616768002510071
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,2,128,power_law_1.2,0.01577087998390198
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,2,128,power_law_1.2,0.01613312005996704
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,2,128,power_law_1.2,0.015846400260925292
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,2,128,power_law_1.2,0.01615231990814209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,128,power_law_1.2,0.7505152130126953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,2,128,power_law_1.2,0.016144640445709228
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,2,128,power_law_1.2,0.016994559764862062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,128,power_law_1.2,0.4980467224121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,2,128,power_law_1.2,0.018648320436477663
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,2,128,power_law_1.01,0.058776321411132815
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,2,128,power_law_1.01,0.07563648223876954
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,2,128,power_law_1.01,0.09686400413513183
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,128,power_law_1.2,1.0245645141601563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,128,power_law_1.2,1.859342041015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,2,128,power_law_1.2,0.017633279561996458
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,2,128,power_law_1.2,0.016596479415893556
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,2,128,power_law_1.2,0.02004607915878296
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,2,128,power_law_1.2,0.02258687973022461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,64,balanced,0.023071999549865725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,64,balanced,0.022677760124206543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,64,balanced,0.023151359558105468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,64,balanced,0.0235647988319397
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,2,128,power_law_1.2,0.028637440204620362
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,64,balanced,0.025149440765380858
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,64,balanced,0.04112256050109863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,64,balanced,0.0439731216430664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,64,balanced,0.044353280067443844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,64,balanced,0.04606719970703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,64,balanced,0.04608640193939209
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,2,128,power_law_1.2,0.03803904056549072
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,64,balanced,0.0468671989440918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,64,balanced,0.05100800037384033
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,2,128,power_law_1.2,0.047283201217651366
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,64,balanced,0.048888320922851565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,64,balanced,0.05466752052307129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,64,balanced,0.05023615837097169
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,64,balanced,0.05139584064483642
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,64,balanced,0.05509119987487793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,64,balanced,0.05883776187896729
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,64,balanced,0.056314878463745124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,64,balanced,0.06369919776916504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,64,balanced,0.06788352012634277
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,64,balanced,0.0766476821899414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,64,balanced,0.09541888236999511
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,64,balanced,0.12524415969848632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,64,balanced,0.14525823593139647
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,64,balanced,0.1850214385986328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,64,balanced,0.026837759017944336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,64,balanced,0.22854272842407228
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,64,balanced,0.026800639629364013
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,64,balanced,0.02723711967468262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,64,balanced,0.027002880573272704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,64,balanced,0.027252480983734128
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,64,balanced,0.033393919467926025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,64,balanced,0.03459712028503418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,64,balanced,0.03387648105621338
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,64,balanced,0.03534976005554199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,64,balanced,0.03420543909072876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,64,balanced,0.03707648038864136
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,64,balanced,0.035345919132232666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,64,balanced,0.037000958919525144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,64,balanced,0.036674559116363525
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,64,balanced,0.0362611198425293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,64,balanced,0.03809920072555542
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,64,balanced,0.040687360763549806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,64,balanced,0.03886080026626587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,64,balanced,0.04238463878631592
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,64,balanced,0.04600831985473633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,64,balanced,0.04976384162902832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,64,balanced,0.05296256065368652
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,64,balanced,0.0686131191253662
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,64,balanced,0.08437888145446777
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,64,balanced,0.11125120162963867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,64,balanced,0.13985535621643067
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,64,balanced,0.036803839206695554
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,2,128,power_law_1.2,0.08613759994506835
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,2,128,power_law_1.2,0.0695359992980957
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,4,64,balanced,0.016129280328750613
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,4,64,balanced,0.016153600215911865
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,4,64,balanced,0.01645311951637268
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,4,64,balanced,0.0168614399433136
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,4,64,balanced,0.016103680133819583
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,4,64,balanced,0.016590080261230468
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,4,64,balanced,0.016162559986114503
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,4,64,balanced,0.016573439836502075
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,4,64,balanced,0.01689471960067749
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,4,64,balanced,0.01654528021812439
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,4,64,balanced,0.016590080261230468
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,4,64,balanced,0.016940799951553346
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,4,64,balanced,0.01657855987548828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,4,64,balanced,0.01708415985107422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,4,64,balanced,0.017186559438705444
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,4,64,balanced,0.017007360458374025
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,4,64,balanced,0.017311999797821047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,4,64,balanced,0.01696128010749817
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,4,64,balanced,0.017328640222549437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,4,64,balanced,0.016573439836502075
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,4,64,balanced,0.016517119407653807
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,2,128,power_law_1.2,0.16145919799804687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,2,128,power_law_1.2,0.12426624298095704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,64,power_law_1.01,0.024311039447784424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,64,power_law_1.01,0.02397952079772949
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,64,power_law_1.01,0.024742400646209715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,64,power_law_1.01,0.041561598777771
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,64,power_law_1.01,0.04449408054351807
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,64,power_law_1.01,0.04552063941955566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,64,power_law_1.01,0.04880640029907227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,64,power_law_1.01,0.048975358009338374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,64,power_law_1.01,0.05106944084167481
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,64,power_law_1.01,0.05058815956115723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,64,power_law_1.01,0.05407872200012207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,64,power_law_1.01,0.05467135906219482
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,64,power_law_1.01,0.05825407981872559
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,64,power_law_1.01,0.05957632064819336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,64,power_law_1.01,0.06047359943389893
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,64,power_law_1.01,0.06251264095306397
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,64,power_law_1.01,0.07160704135894776
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,64,power_law_1.01,0.0907686424255371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,64,power_law_1.01,0.10289536476135255
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,64,power_law_1.01,0.12370047569274903
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,64,power_law_1.01,0.15669504165649414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,64,power_law_1.01,0.2048819160461426
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,4,64,balanced,0.02189824104309082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,4,64,balanced,0.02104448080062866
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,4,64,balanced,0.01860479950904846
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,4,64,balanced,0.01897088050842285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,64,power_law_1.01,0.33932289123535153
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,64,power_law_1.01,0.5239436721801758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,64,power_law_1.01,0.026824960708618163
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,64,power_law_1.01,0.027243518829345705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,64,power_law_1.01,0.027234559059143064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,64,power_law_1.01,0.709428482055664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,64,power_law_1.01,0.24799871444702148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,64,power_law_1.01,0.033085439205169674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,64,power_law_1.01,0.034209280014038085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,64,power_law_1.01,0.0358784008026123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,64,power_law_1.01,0.9726783752441406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,64,power_law_1.01,0.036990718841552736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,64,power_law_1.01,0.035842559337615966
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,64,power_law_1.01,0.03588351964950561
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,64,power_law_1.01,0.03747584104537964
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,64,power_law_1.01,0.03830143928527832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,64,power_law_1.01,0.038302719593048096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,64,power_law_1.01,0.039585280418396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,64,power_law_1.01,0.04240128040313721
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,64,power_law_1.01,0.04197887897491455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,64,power_law_1.01,0.042388482093811033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,64,power_law_1.01,0.04485375881195068
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,64,power_law_1.01,0.0578931188583374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,64,power_law_1.01,0.07457791805267334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,64,power_law_1.01,0.09419391632080079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,64,power_law_1.01,0.12019455909729004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,64,power_law_1.01,0.15378047943115233
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,64,power_law_1.01,0.20910720825195311
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,64,power_law_1.01,0.2657510375976563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,4,64,power_law_1.01,0.01667199969291687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,64,power_law_1.01,0.4274687957763672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,4,64,power_law_1.01,0.016567039489746093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,4,64,power_law_1.01,0.01649791955947876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,64,power_law_1.01,0.5844876861572266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,64,power_law_1.01,0.052523522377014166
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,4,64,power_law_1.01,0.01656191945075989
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,4,64,power_law_1.01,0.01690240025520325
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,4,64,power_law_1.01,0.016925439834594727
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,4,64,power_law_1.01,0.01691264033317566
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,4,64,power_law_1.01,0.01699839949607849
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,4,64,power_law_1.01,0.016915199756622316
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,4,64,power_law_1.01,0.016989439725875854
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,4,64,power_law_1.01,0.016951040029525757
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,4,64,power_law_1.01,0.017050880193710326
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,4,64,power_law_1.01,0.01731711983680725
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,4,64,balanced,0.025282559394836424
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,4,64,power_law_1.01,0.02274303913116455
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,4,64,balanced,0.027761919498443605
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,4,64,power_law_1.01,0.016531200408935548
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,4,64,power_law_1.01,0.018169599771499633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,4,64,power_law_1.01,0.016563199758529663
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,4,64,power_law_1.01,0.018172160387039185
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,4,64,power_law_1.01,0.0201907205581665
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,4,64,power_law_1.01,0.019453439712524414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,64,power_law_1.2,0.025199360847473144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,64,power_law_1.2,0.024396800994873048
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,64,power_law_1.2,0.024367361068725585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,64,power_law_1.2,0.04319744110107422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,64,power_law_1.2,0.0460748815536499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,64,power_law_1.2,0.0476364803314209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,64,power_law_1.2,0.05134592056274414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,64,power_law_1.2,0.05387392044067383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,64,power_law_1.2,0.052252159118652344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,64,power_law_1.2,0.05234303951263428
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,64,power_law_1.2,0.05470719814300538
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,64,power_law_1.2,0.05499392032623292
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,64,power_law_1.2,0.05787136077880859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,64,power_law_1.2,0.059699201583862306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,64,power_law_1.2,0.06436480045318603
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,64,power_law_1.2,0.06842112064361572
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,64,power_law_1.2,0.08400128364562989
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,64,power_law_1.2,0.09912447929382324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,64,power_law_1.2,0.121081600189209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,64,power_law_1.2,0.14351743698120117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,64,power_law_1.2,0.19360000610351563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,64,power_law_1.2,0.2477030372619629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,64,power_law_1.2,0.37736705780029295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,64,power_law_1.2,0.4970111846923828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,4,64,power_law_1.01,0.02472575902938843
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,4,64,power_law_1.01,0.027743999958038328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,64,power_law_1.2,0.7333184051513671
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,64,power_law_1.2,0.02723711967468262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,64,power_law_1.2,0.027656960487365722
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,64,power_law_1.2,0.0337779188156128
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,64,power_law_1.2,0.03462016105651856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,64,power_law_1.2,0.03502079963684082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,64,power_law_1.2,0.03584127902984619
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,64,power_law_1.2,0.03782016038894653
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,64,power_law_1.2,0.03725696086883545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,64,power_law_1.2,0.03868160009384155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,64,power_law_1.2,0.037616639137268065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,64,power_law_1.2,0.03828736066818238
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,64,power_law_1.2,0.040033278465270994
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,4,64,power_law_1.01,0.033278720378875734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,64,power_law_1.2,0.04058879852294922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,64,power_law_1.2,0.04235519886016846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,64,power_law_1.2,0.04239615917205811
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,64,power_law_1.2,0.04442240238189697
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,4,64,power_law_1.01,0.04815487861633301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,64,power_law_1.2,0.047779841423034666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,64,power_law_1.2,0.0590169620513916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,64,power_law_1.2,0.06669695854187012
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,64,power_law_1.2,0.09013504028320313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,4,64,power_law_1.01,0.08031744003295899
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,64,power_law_1.2,0.11551103591918946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,64,power_law_1.2,0.15692543983459473
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,4,64,power_law_1.01,0.1070310401916504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,64,power_law_1.2,0.22819711685180666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,4,64,power_law_1.01,0.07112703800201417
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,4,64,power_law_1.2,0.016714240312576296
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,4,64,power_law_1.2,0.016633599996566772
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,4,64,power_law_1.2,0.016568319797515867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,4,64,power_law_1.2,0.016567039489746093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,4,64,power_law_1.2,0.016599040031433105
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,4,64,power_law_1.2,0.016811519861221313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,64,power_law_1.2,0.33738113403320313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,4,64,power_law_1.2,0.016563199758529663
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,4,64,power_law_1.2,0.01664639949798584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,4,64,power_law_1.2,0.01684607982635498
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,4,64,power_law_1.2,0.01697535991668701
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,4,64,power_law_1.2,0.017360639572143555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,4,64,power_law_1.2,0.01700096011161804
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,4,64,power_law_1.2,0.016989439725875854
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,4,64,power_law_1.2,0.017006080150604248
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,4,64,power_law_1.2,0.01820799946784973
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,64,power_law_1.2,0.45770622253417964
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,64,power_law_1.2,1.1236236572265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,64,power_law_1.2,0.8351846313476562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,64,power_law_1.2,0.6588966369628906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,64,power_law_1.2,1.3569229125976563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,4,64,power_law_1.2,0.0182259202003479
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,4,64,power_law_1.2,0.020304639339447022
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,4,64,power_law_1.2,0.019014400243759156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,4,64,power_law_1.2,0.02150144100189209
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,4,64,power_law_1.2,0.023522560596466062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,4,64,power_law_1.2,0.032231678962707525
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,8,32,balanced,0.026003201007843018
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,8,32,balanced,0.022653439044952393
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,8,32,balanced,0.02640511989593506
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,8,32,balanced,0.023150079250335694
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,8,32,balanced,0.02523776054382324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,8,32,balanced,0.043215360641479496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,8,32,balanced,0.04547327995300293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,8,32,balanced,0.04601727962493897
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,8,32,balanced,0.04728831768035889
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,8,32,balanced,0.04803455829620361
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,8,32,balanced,0.04968319892883301
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,4,64,power_law_1.2,0.035983359813690184
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,8,32,balanced,0.05066112041473388
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,8,32,balanced,0.05305344104766846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,8,32,balanced,0.05573247909545899
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,8,32,balanced,0.0575270414352417
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,8,32,balanced,0.05625343799591065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,8,32,balanced,0.06124415874481202
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,8,32,balanced,0.06487040042877197
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,8,32,balanced,0.07348480224609374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,8,32,balanced,0.09265664100646973
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,8,32,balanced,0.12036095619201662
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,8,32,balanced,0.14009856224060058
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,8,32,balanced,0.17881343841552735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,8,32,balanced,0.21725568771362305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,8,32,balanced,0.2930444717407227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,8,32,balanced,0.36973697662353516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,8,32,balanced,0.026759679317474368
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,8,32,balanced,0.026725120544433593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,8,32,balanced,0.026823680400848388
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,8,32,balanced,0.049329919815063475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,8,32,balanced,0.027987198829650877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,8,32,balanced,0.03460223913192749
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,8,32,balanced,0.03427072048187256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,8,32,balanced,0.0354150390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,8,32,balanced,0.03533056020736695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,8,32,balanced,0.026757121086120605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,8,32,balanced,0.03598975896835327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,8,32,balanced,0.03682559967041016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,8,32,balanced,0.036248319149017334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,8,32,balanced,0.037036800384521486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,8,32,balanced,0.03831295967102051
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,8,32,balanced,0.04040448188781738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,8,32,balanced,0.04295167922973633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,8,32,balanced,0.039540479183197016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,8,32,balanced,0.04279168128967285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,8,32,balanced,0.04647295951843262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,8,32,balanced,0.04853759765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,8,32,balanced,0.05336703777313232
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,8,32,balanced,0.06905983924865723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,8,32,balanced,0.08098688125610351
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,8,32,balanced,0.10896896362304687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,8,32,balanced,0.13591039657592774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,8,32,balanced,0.1866636848449707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,8,32,balanced,0.23906944274902342
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,4,64,power_law_1.2,0.0555072021484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,4,64,power_law_1.2,0.09838720321655273
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,4,64,power_law_1.2,0.07318784236907959
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,4,64,power_law_1.2,0.16267520904541016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,8,32,balanced,0.016917760372161868
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,8,32,balanced,0.01734272003173828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,8,32,balanced,0.01692415952682495
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,8,32,balanced,0.017831679582595825
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,8,32,balanced,0.01747455954551697
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,8,32,balanced,0.017786879539489747
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,8,32,balanced,0.017791999578475954
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,8,32,balanced,0.01776128053665161
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,8,32,balanced,0.017713919878005982
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,8,32,balanced,0.01788159966468811
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,8,32,balanced,0.01778815984725952
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,8,32,balanced,0.01817855954170227
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,8,32,balanced,0.017040640115737915
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,8,32,balanced,0.017822719812393188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,8,32,balanced,0.018648320436477663
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,8,32,balanced,0.018612480163574217
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,8,32,balanced,0.018965760469436647
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,8,32,balanced,0.01900287985801697
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,4,64,power_law_1.2,0.12592384338378906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,8,32,balanced,0.017422080039978027
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,8,32,balanced,0.017815040349960325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,8,32,power_law_1.01,0.023215360641479492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,8,32,power_law_1.01,0.0247654390335083
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,8,32,power_law_1.01,0.025489919185638428
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,8,32,power_law_1.01,0.042155518531799316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,8,32,power_law_1.01,0.04525568008422852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,8,32,power_law_1.01,0.04810751914978027
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,8,32,power_law_1.01,0.050506238937377934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,8,32,power_law_1.01,0.050544638633728024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,8,32,power_law_1.01,0.05058944225311279
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,8,32,power_law_1.01,0.05222655773162842
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,8,32,power_law_1.01,0.05581439971923828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,8,32,power_law_1.01,0.05744128227233887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,8,32,power_law_1.01,0.06120063781738281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,8,32,power_law_1.01,0.058872318267822264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,8,32,power_law_1.01,0.061199359893798835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,8,32,power_law_1.01,0.06684800148010253
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,8,32,power_law_1.01,0.08731904029846191
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,8,32,power_law_1.01,0.09325183868408203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,8,32,power_law_1.01,0.10723072052001954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,8,32,power_law_1.01,0.13691776275634765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,8,32,power_law_1.01,0.1570355224609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,8,32,power_law_1.01,0.21549568176269532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,8,32,power_law_1.01,0.2615321540832519
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,8,32,power_law_1.01,0.37490177154541016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,8,32,power_law_1.01,0.470384635925293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,8,32,power_law_1.01,0.7221785736083984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,8,32,power_law_1.01,1.0327193450927734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,8,32,power_law_1.01,0.02798847913742065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,8,32,power_law_1.01,0.028440320491790773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,8,32,power_law_1.01,0.028803839683532718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,8,32,power_law_1.01,0.03450367927551269
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,8,32,power_law_1.01,0.035829761028289796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,8,32,power_law_1.01,0.03705856084823608
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,8,32,power_law_1.01,0.039061760902404784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,8,32,power_law_1.01,0.03904128074645996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,8,32,power_law_1.01,0.036221439838409426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,8,32,power_law_1.01,0.03641983985900879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,8,32,power_law_1.01,0.03829119920730591
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,8,32,balanced,0.017749760150909424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,8,32,power_law_1.01,0.03984384059906006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,8,32,power_law_1.01,0.04151167869567871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,8,32,power_law_1.01,0.04313471794128418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,8,32,power_law_1.01,0.04346496105194092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,8,32,power_law_1.01,0.04671232223510742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,8,32,power_law_1.01,0.0480511999130249
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,8,32,power_law_1.01,0.054585599899291994
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,8,32,power_law_1.01,0.059575037956237795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,8,32,power_law_1.01,0.07628928184509277
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,8,32,power_law_1.01,0.0982694435119629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,8,32,power_law_1.01,0.12569727897644042
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,8,32,power_law_1.01,0.1558835220336914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,8,32,power_law_1.01,0.2369548797607422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,8,32,power_law_1.01,0.29615743637084957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,8,32,power_law_1.01,0.43247360229492193
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,8,32,power_law_1.01,0.01773568034172058
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,8,32,power_law_1.01,0.5944921493530273
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,8,32,power_law_1.01,0.017369600534439086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,8,32,power_law_1.01,0.017419519424438475
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,8,32,power_law_1.01,0.017571840286254883
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,8,32,power_law_1.01,0.017745920419692994
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,8,32,power_law_1.01,0.01738495945930481
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,8,32,power_law_1.01,0.017791999578475954
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,8,32,power_law_1.01,0.01773568034172058
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,8,32,power_law_1.01,0.017760000228881835
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,8,32,power_law_1.01,0.017704960107803345
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,8,32,power_law_1.01,0.01792896032333374
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,8,32,balanced,0.02152575969696045
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,8,32,balanced,0.021821439266204834
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,8,32,power_law_1.01,0.017804800271987914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,8,32,power_law_1.01,0.018228479623794556
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,8,32,power_law_1.01,0.01823359966278076
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,8,32,power_law_1.01,0.01989248037338257
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,8,32,balanced,0.034387199878692626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,8,32,power_law_1.01,0.018184959888458252
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,8,32,balanced,0.02477695941925049
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,8,32,balanced,0.027870719432830808
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,8,32,balanced,0.04017792224884033
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,8,32,power_law_1.01,0.01945855975151062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,8,32,power_law_1.01,0.02502784013748169
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,8,32,power_law_1.01,0.020853760242462157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,8,32,power_law_1.01,0.021456639766693115
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,8,32,power_law_1.01,0.0273907208442688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,8,32,power_law_1.2,0.02348927974700928
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,8,32,power_law_1.2,0.024788479804992675
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,8,32,power_law_1.01,0.03606143951416015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,8,32,power_law_1.2,0.02529536008834839
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,8,32,power_law_1.2,0.042813439369201664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,8,32,power_law_1.2,0.045644798278808595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,8,32,power_law_1.2,0.04686079978942871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,8,32,power_law_1.2,0.05155072212219238
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,8,32,power_law_1.2,0.04973951816558838
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,8,32,power_law_1.01,0.04340735912322998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,8,32,power_law_1.2,0.05098624229431152
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,8,32,power_law_1.2,0.052643837928771975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,8,32,power_law_1.2,0.056860160827636716
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,8,32,power_law_1.2,0.06000768184661866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,8,32,power_law_1.2,0.0603878402709961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,8,32,power_law_1.2,0.06328320026397705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,8,32,power_law_1.2,0.07337471961975098
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,8,32,power_law_1.2,0.09101951599121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,8,32,power_law_1.2,0.09663616180419922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,8,32,power_law_1.2,0.10409728050231934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,8,32,power_law_1.2,0.16022655487060547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,8,32,power_law_1.2,0.20654207229614258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,8,32,power_law_1.2,0.2527231979370117
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,8,32,power_law_1.01,0.062154240608215325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,8,32,power_law_1.2,0.35115776062011717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,8,32,power_law_1.2,0.061495041847229
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,8,32,power_law_1.2,0.027645440101623536
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,8,32,power_law_1.2,0.02799232006072998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,8,32,power_law_1.2,0.028382720947265627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,8,32,power_law_1.2,0.03412735939025879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,8,32,power_law_1.2,0.034959359169006346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,8,32,power_law_1.2,0.5203007888793946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,8,32,power_law_1.2,0.03584000110626221
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,8,32,power_law_1.2,0.036240639686584475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,8,32,power_law_1.2,0.03699455976486206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,8,32,power_law_1.2,0.03688064098358154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,8,32,power_law_1.2,0.03676032066345215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,8,32,power_law_1.2,0.040227842330932614
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,8,32,power_law_1.2,0.04329343795776367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,8,32,power_law_1.2,0.04323200225830078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,8,32,power_law_1.2,0.04254591941833496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,8,32,power_law_1.2,0.04483967781066894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,8,32,power_law_1.2,0.04682496070861816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,8,32,power_law_1.2,0.05137792110443116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,8,32,power_law_1.2,0.05862271785736084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,8,32,power_law_1.2,0.06352384090423584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,8,32,power_law_1.2,0.0927462387084961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,8,32,power_law_1.2,0.11810943603515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,8,32,power_law_1.2,0.15056896209716797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,8,32,power_law_1.2,0.6869593811035156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,8,32,power_law_1.2,0.2121369552612305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,8,32,power_law_1.2,0.2980569648742676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,8,32,power_law_1.2,0.44870529174804685
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,8,32,power_law_1.2,0.01740288019180298
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,8,32,power_law_1.2,0.01774719953536987
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,8,32,power_law_1.2,0.017329920530319214
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,8,32,power_law_1.2,0.01755519986152649
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,8,32,power_law_1.2,0.017392640113830564
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,8,32,power_law_1.2,0.017781759500503543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,8,32,power_law_1.2,0.01773695945739746
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,8,32,power_law_1.2,0.017601280212402343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,8,32,power_law_1.2,0.656622085571289
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,8,32,power_law_1.2,0.017824000120162962
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,8,32,power_law_1.2,0.017751040458679198
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,8,32,power_law_1.2,0.017766400575637817
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,8,32,power_law_1.2,0.017786879539489747
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,8,32,power_law_1.01,0.06493824005126954
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,8,32,power_law_1.2,0.01978368043899536
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,8,32,power_law_1.2,0.8028377532958985
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,8,32,power_law_1.2,0.019790079593658447
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,8,32,power_law_1.2,0.0205785608291626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,8,32,power_law_1.01,0.08605952262878418
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,8,32,power_law_1.01,0.14034815788269045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,8,32,power_law_1.2,1.0223104095458984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,8,32,power_law_1.2,1.2966490173339844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,8,32,power_law_1.2,0.01901952028274536
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,8,32,power_law_1.2,0.017800960540771484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,8,32,power_law_1.2,0.021509120464324953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,8,32,power_law_1.2,0.030104320049285888
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,8,32,power_law_1.2,0.024052479267120362
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,16,16,balanced,0.022113280296325682
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,16,16,balanced,0.02232192039489746
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,8,32,power_law_1.2,0.034639360904693606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,16,16,balanced,0.02392064094543457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,16,16,balanced,0.02470911979675293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,16,16,balanced,0.027971839904785155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,16,16,balanced,0.04644480228424072
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,16,16,balanced,0.04888576030731201
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,16,16,balanced,0.05227263927459717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,16,16,balanced,0.05062784194946289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,16,16,balanced,0.05174911975860595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,16,16,balanced,0.05337984085083007
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,16,16,balanced,0.05583360195159912
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,16,16,balanced,0.05835904121398926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,16,16,balanced,0.05486080169677734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,16,16,balanced,0.05834623813629151
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,16,16,balanced,0.060751361846923826
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,16,16,balanced,0.06642303943634034
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,16,16,balanced,0.07288576126098632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,16,16,balanced,0.09529472351074218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,16,16,balanced,0.1200268840789795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,16,16,balanced,0.1377241611480713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,16,16,balanced,0.1760038375854492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,16,16,balanced,0.2132953643798828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,8,32,power_law_1.2,0.04653696060180664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,16,16,balanced,0.2856012725830078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,16,16,balanced,0.35997825622558594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,16,16,balanced,0.5032588958740234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,16,16,balanced,0.026361598968505862
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,16,16,balanced,0.026051840782165527
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,16,16,balanced,0.026403839588165286
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,16,16,balanced,0.027328000068664548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,16,16,balanced,0.6489215850830078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,16,16,balanced,0.029278719425201417
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,16,16,balanced,0.03667072057723999
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,16,16,balanced,0.04074751853942871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,16,16,balanced,0.038731520175933835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,16,16,balanced,0.04155776023864746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,16,16,balanced,0.03955712080001831
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,16,16,balanced,0.04029695987701416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,16,16,balanced,0.04289279937744141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,16,16,balanced,0.04527999877929688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,16,16,balanced,0.04118912220001221
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,16,16,balanced,0.04313087940216064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,16,16,balanced,0.04418047904968262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,16,16,balanced,0.047295999526977536
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,16,16,balanced,0.04855167865753174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,16,16,balanced,0.0529369592666626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,16,16,balanced,0.06810368061065673
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,16,16,balanced,0.07944447994232177
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,16,16,balanced,0.10823936462402343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,16,16,balanced,0.13037952423095703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,16,16,balanced,0.18104448318481445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,16,16,balanced,0.23106304168701172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,16,16,balanced,0.32980865478515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,16,16,balanced,0.4236223983764648
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,8,32,power_law_1.2,0.06033408164978027
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,8,32,power_law_1.2,0.07631103992462158
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,8,32,power_law_1.2,0.10460800170898436
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,16,16,balanced,0.01781888008117676
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,16,16,balanced,0.02076672077178955
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,16,16,balanced,0.017748479843139646
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,16,16,balanced,0.020627200603485107
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,16,16,balanced,0.018184959888458252
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,16,16,balanced,0.021020159721374512
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,16,16,balanced,0.021125121116638185
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,16,16,balanced,0.021086719036102295
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,16,16,balanced,0.021469440460205078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,16,16,balanced,0.021493759155273438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,16,16,balanced,0.02132352113723755
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,16,16,balanced,0.021087999343872073
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,16,16,balanced,0.02144511938095093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,16,16,balanced,0.021510400772094727
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,16,16,balanced,0.021594879627227784
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,16,16,balanced,0.019841279983520508
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,8,32,power_law_1.2,0.15220352172851564
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,16,16,balanced,0.02266752004623413
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,8,32,power_law_1.2,0.1578764820098877
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,16,16,balanced,0.021247999668121336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,16,16,balanced,0.02102400064468384
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,16,16,balanced,0.027307519912719725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,16,16,power_law_1.01,0.027641599178314206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,16,16,power_law_1.01,0.025647358894348143
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,16,16,balanced,0.0296396803855896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,16,16,power_law_1.01,0.02597120046615601
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,16,16,power_law_1.01,0.04539775848388672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,16,16,balanced,0.037484800815582274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,16,16,power_law_1.01,0.04768383979797363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,16,16,power_law_1.01,0.05134592056274414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,16,16,power_law_1.01,0.0530457592010498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,16,16,power_law_1.01,0.054202880859375004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,16,16,power_law_1.01,0.05828864097595214
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,16,16,power_law_1.01,0.06158720016479492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,16,16,power_law_1.01,0.05985536098480224
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,16,16,power_law_1.01,0.02678272008895874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,16,16,power_law_1.01,0.051528959274292
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,16,16,power_law_1.01,0.06098048210144043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,16,16,power_law_1.01,0.06360960006713867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,16,16,power_law_1.01,0.06618112087249756
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,16,16,power_law_1.01,0.08790911674499512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,16,16,power_law_1.01,0.10626943588256836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,16,16,power_law_1.01,0.11047295570373535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,16,16,power_law_1.01,0.14821632385253908
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,16,16,power_law_1.01,0.18159744262695313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,16,16,power_law_1.01,0.2098124885559082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,16,16,power_law_1.01,0.27345535278320315
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,16,16,balanced,0.042906880378723145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,16,16,power_law_1.01,0.41964286804199225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,16,16,power_law_1.01,0.5212403106689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,16,16,power_law_1.01,0.027583999633789064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,16,16,power_law_1.01,0.028476159572601318
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,16,16,power_law_1.01,0.02799232006072998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,16,16,power_law_1.01,0.7524787139892578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,16,16,power_law_1.01,0.028849918842315674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,16,16,power_law_1.01,0.03544447898864746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,16,16,power_law_1.01,0.9176870727539063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,16,16,power_law_1.01,0.03737728118896484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,16,16,power_law_1.01,0.038264319896698
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,16,16,power_law_1.01,0.03863168001174927
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,16,16,power_law_1.01,0.03994112014770508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,16,16,power_law_1.01,0.040344319343566894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,16,16,power_law_1.01,0.04272384166717529
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,16,16,power_law_1.01,0.04601984024047852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,16,16,power_law_1.01,0.04605311870574951
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,16,16,power_law_1.01,0.045781760215759276
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,16,16,power_law_1.01,0.04649472236633301
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,16,16,balanced,0.06091519832611084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,16,16,power_law_1.01,0.04978816032409668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,16,16,power_law_1.01,0.06167808055877686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,16,16,power_law_1.01,0.07099391937255858
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,16,16,power_law_1.01,0.088089599609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,16,16,power_law_1.01,0.10211071968078614
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,16,16,power_law_1.01,0.14689151763916014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,16,16,power_law_1.01,0.1856038475036621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,16,16,power_law_1.01,0.24811008453369138
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,16,16,power_law_1.01,0.04361728191375732
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,16,16,power_law_1.01,0.36329601287841795
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,16,16,power_law_1.01,0.021022720336914064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,16,16,power_law_1.01,0.5216345596313476
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,16,16,power_law_1.01,0.020830719470977782
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,16,16,power_law_1.01,0.020700159072875975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,16,16,power_law_1.01,0.657940444946289
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,16,16,power_law_1.01,0.0210150408744812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,16,16,power_law_1.01,0.02103424072265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,16,16,power_law_1.01,0.021054720878601073
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,16,16,power_law_1.01,0.021080319881439206
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,16,16,power_law_1.01,0.021497600078582764
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,16,16,power_law_1.01,0.021064960956573488
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,16,16,power_law_1.01,0.021419520378112795
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,16,16,power_law_1.01,0.021461760997772215
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,16,16,power_law_1.01,0.02147200107574463
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,16,16,power_law_1.01,0.02243583917617798
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,16,16,balanced,0.07089407920837403
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,16,16,power_law_1.01,0.02069119930267334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,16,16,power_law_1.01,0.020453119277954103
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,16,16,power_law_1.01,0.02349440097808838
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,16,16,power_law_1.01,0.025150721073150632
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,16,16,power_law_1.01,0.028444159030914306
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,16,16,power_law_1.01,0.03521536111831665
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,16,16,power_law_1.01,0.04076543807983398
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,16,16,power_law_1.01,0.055092477798461915
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,16,16,power_law_1.01,0.06811264038085937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,16,16,power_law_1.2,0.028038399219512938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,16,16,power_law_1.2,0.025356800556182862
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,16,16,power_law_1.2,0.025556480884552
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,16,16,power_law_1.2,0.02761856079101562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,16,16,power_law_1.2,0.04488575935363769
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,16,16,power_law_1.2,0.04901887893676758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,16,16,power_law_1.2,0.05063936233520507
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,16,16,power_law_1.2,0.05143807888031006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,16,16,power_law_1.2,0.05265151977539062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,16,16,power_law_1.2,0.05432064056396484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,16,16,power_law_1.2,0.058652157783508296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,16,16,power_law_1.2,0.06121984004974366
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,16,16,power_law_1.2,0.06120831966400146
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,16,16,power_law_1.01,0.08615424156188964
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,16,16,power_law_1.2,0.06480768203735351
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,16,16,power_law_1.2,0.06578303813934326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,16,16,power_law_1.2,0.07678080081939698
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,16,16,power_law_1.2,0.08609279632568359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,16,16,power_law_1.2,0.1086079978942871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,16,16,power_law_1.2,0.12879103660583496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,16,16,power_law_1.2,0.16309759140014649
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,16,16,power_law_1.2,0.2041574478149414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,16,16,power_law_1.2,0.24757759094238282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,16,16,power_law_1.2,0.3394303894042969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,16,16,power_law_1.2,0.5974310302734376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,16,16,power_law_1.2,0.48453121185302733
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,16,16,power_law_1.01,0.11308159828186035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,16,16,power_law_1.2,0.027662079334259033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,16,16,power_law_1.2,0.028423678874969483
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,16,16,power_law_1.2,0.02847872018814087
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,16,16,power_law_1.2,0.028555519580841064
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,16,16,power_law_1.01,0.1836595153808594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,16,16,power_law_1.2,0.03537024021148681
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,16,16,power_law_1.2,0.03953919887542724
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,16,16,power_law_1.2,0.037880320549011234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,16,16,power_law_1.2,0.03960576057434082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,16,16,power_law_1.01,0.15103103637695312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,16,16,power_law_1.2,0.03993983983993531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,16,16,power_law_1.2,0.04124671936035156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,16,16,power_law_1.2,0.04396927833557129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,16,16,power_law_1.2,0.04483071804046631
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,16,16,power_law_1.2,0.043175678253173824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,16,16,power_law_1.2,0.04485631942749023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,16,16,power_law_1.2,0.04690944194793702
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,16,16,power_law_1.2,0.04899839878082275
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,16,16,power_law_1.2,0.05416319847106934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,16,16,power_law_1.2,0.06354688167572022
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,16,16,power_law_1.2,0.07373568058013916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,16,16,power_law_1.2,0.09512319564819335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,16,16,power_law_1.2,0.11406335830688477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,16,16,power_law_1.2,0.17723775863647462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,16,16,power_law_1.2,0.85299072265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,16,16,power_law_1.2,1.2602790069580077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,16,16,power_law_1.2,0.22764928817749022
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,16,16,power_law_1.2,0.020316159725189208
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,16,16,power_law_1.2,0.020339200496673586
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,16,16,power_law_1.2,0.020570878982543943
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,16,16,power_law_1.2,0.3378867340087891
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,16,16,power_law_1.2,0.020261120796203614
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,16,16,power_law_1.2,0.020280320644378663
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,16,16,power_law_1.2,0.02068351984024048
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,16,16,power_law_1.2,0.021063680648803713
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,16,16,power_law_1.2,0.020657920837402345
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,16,16,power_law_1.2,0.021004800796508787
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,16,16,power_law_1.2,0.02084991931915283
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,16,16,power_law_1.2,0.021442561149597167
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,16,16,power_law_1.2,0.02106879949569702
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,16,16,power_law_1.2,0.02108544111251831
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,16,16,power_law_1.2,0.0215116810798645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,16,16,power_law_1.2,0.4136908721923828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,16,16,power_law_1.2,0.023092479705810548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,16,16,power_law_1.2,0.7869516754150391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,16,16,power_law_1.2,0.5726924896240234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,16,16,power_law_1.2,0.02225791931152344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,16,16,power_law_1.2,0.023543040752410888
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,16,16,power_law_1.2,0.026851840019226074
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,16,16,power_law_1.2,0.030791680812835693
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,16,16,power_law_1.2,0.03540735960006714
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,16,16,power_law_1.2,0.16773759841918945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,32,8,balanced,0.023129599094390867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,16,16,power_law_1.2,0.11950207710266114
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,32,8,balanced,0.025208320617675785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,32,8,balanced,0.028405759334564205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,32,8,balanced,0.030896639823913573
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,32,8,balanced,0.05713024139404297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,32,8,balanced,0.055045118331909185
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,32,8,balanced,0.056417279243469245
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,32,8,balanced,0.059169278144836426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,32,8,balanced,0.06116864204406738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,32,8,balanced,0.05794047832489013
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,32,8,balanced,0.06121471881866455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,32,8,balanced,0.06370175838470458
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,32,8,balanced,0.06759679794311524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,32,8,balanced,0.06979199886322021
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,16,16,power_law_1.2,0.04730239868164062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,16,16,power_law_1.2,0.06380032062530518
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,32,8,balanced,0.07360000133514404
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,32,8,balanced,0.09313152313232423
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,32,8,balanced,0.17682943344116211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,32,8,balanced,0.21538047790527343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,32,8,balanced,0.28708480834960937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,32,8,balanced,0.35912960052490234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,32,8,balanced,0.5006771087646484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,32,8,balanced,0.6433318328857422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,32,8,balanced,0.9226534271240234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,32,8,balanced,0.027229440212249757
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,32,8,balanced,0.027644159793853762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,32,8,balanced,1.2048818969726562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,32,8,balanced,0.029230079650878905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,32,8,balanced,0.03064703941345215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,32,8,balanced,0.03704576015472412
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,32,8,balanced,0.04318975925445557
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,32,8,balanced,0.0443942403793335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,32,8,balanced,0.04605696201324463
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,32,8,balanced,0.04746623992919922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,32,8,balanced,0.048935680389404296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,32,8,balanced,0.0456063985824585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,32,8,balanced,0.04839807987213135
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,32,8,balanced,0.04935808181762695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,32,8,balanced,0.050731520652771
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,32,8,balanced,0.05173888206481934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,32,8,balanced,0.053009920120239265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,32,8,balanced,0.11961600303649902
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,32,8,balanced,0.13847552299499513
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,32,8,balanced,0.05792768001556396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,32,8,balanced,0.07102975845336915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,32,8,balanced,0.12949888229370116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,32,8,balanced,0.08060544013977052
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,32,8,balanced,0.18643840789794922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,32,8,balanced,0.2284774398803711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,32,8,balanced,0.3278668975830078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,32,8,balanced,0.4231270217895508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,32,8,power_law_1.01,0.027624959945678713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,32,8,balanced,0.616239356994629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,32,8,power_law_1.01,0.026004478931427
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,32,8,power_law_1.01,0.02716032028198242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,32,8,balanced,0.10569984436035156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,32,8,balanced,0.8031846618652343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,32,8,power_law_1.01,0.029119999408721925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,32,8,power_law_1.01,0.0329638409614563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,32,8,power_law_1.01,0.05419136047363281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,32,8,power_law_1.01,0.05548927783966064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,32,8,power_law_1.01,0.056628479957580566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,32,8,balanced,0.0219980788230896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,32,8,power_law_1.01,0.0595366382598877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,32,8,power_law_1.01,0.06324480056762696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,32,8,power_law_1.01,0.0630515193939209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,32,8,power_law_1.01,0.0755673599243164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,32,8,power_law_1.01,0.0675865602493286
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,32,8,power_law_1.01,0.07490687847137452
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,32,8,power_law_1.01,0.0883238410949707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,32,8,power_law_1.01,0.11336704254150391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,32,8,power_law_1.01,0.12959232330322265
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,16,16,power_law_1.2,0.21673215866088866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,32,8,power_law_1.01,0.16443008422851563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,32,8,power_law_1.01,0.19824127197265623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,32,8,power_law_1.01,0.2648896026611328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,32,8,power_law_1.01,0.31163135528564456
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,32,8,power_law_1.01,0.4627878570556641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,32,8,power_law_1.01,0.06287615776062012
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,32,8,power_law_1.01,0.5891097640991211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,32,8,power_law_1.01,0.02970495939254761
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,32,8,power_law_1.01,0.8897062683105469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,32,8,power_law_1.01,0.02730623960494995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,32,8,power_law_1.01,0.03083904027938843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,32,8,power_law_1.01,1.0724889373779296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,32,8,power_law_1.01,0.04165503978729248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,32,8,power_law_1.01,0.04440192222595215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,32,8,power_law_1.01,0.04600192070007324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,32,8,power_law_1.01,0.04773248195648193
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,32,8,power_law_1.01,0.04848000049591065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,32,8,power_law_1.01,0.048115200996398926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,32,8,power_law_1.01,0.04733695983886719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,32,8,power_law_1.01,0.050079998970031736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,32,8,power_law_1.01,0.037395200729370116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,32,8,power_law_1.01,0.050995202064514154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,32,8,power_law_1.01,0.05139584064483642
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,32,8,power_law_1.01,0.0448422384262085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,32,8,power_law_1.01,0.06996863842010498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,32,8,power_law_1.01,0.055828480720520025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,32,8,power_law_1.01,0.08103551864624023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,32,8,power_law_1.01,0.1273151969909668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,32,8,power_law_1.01,0.10880895614624024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,32,8,power_law_1.01,0.18970624923706053
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,32,8,power_law_1.01,0.21461376190185547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,32,8,power_law_1.01,0.2989632034301758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,32,8,power_law_1.01,0.4121331024169922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,32,8,power_law_1.2,0.023919360637664793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,32,8,power_law_1.2,0.025950720310211183
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,32,8,power_law_1.01,0.6487334442138672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,32,8,power_law_1.2,0.02721280097961426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,32,8,power_law_1.2,0.02929663896560669
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,32,8,power_law_1.01,0.7995123291015626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,32,8,power_law_1.2,0.052620801925659176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,32,8,power_law_1.2,0.03503488063812256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,32,8,power_law_1.2,0.05424384117126465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,32,8,power_law_1.2,0.05722623825073242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,32,8,power_law_1.2,0.05894527912139893
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,32,8,power_law_1.2,0.06467199802398682
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,32,8,power_law_1.2,0.062452478408813475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,32,8,power_law_1.2,0.0624729585647583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,32,8,power_law_1.2,0.06776703834533691
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,32,8,power_law_1.2,0.08840191841125489
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,32,8,power_law_1.2,0.09885055541992187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,32,8,power_law_1.2,0.11472127914428712
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,16,16,power_law_1.2,0.3196326446533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,32,8,power_law_1.2,0.1336780834197998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,32,8,power_law_1.2,0.056707839965820316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,32,8,power_law_1.2,0.21539327621459964
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,32,8,power_law_1.2,0.2753625679016113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,32,8,power_law_1.2,0.07796991825103759
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,32,8,power_law_1.2,0.3436825561523437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,32,8,power_law_1.2,0.1686579132080078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,32,8,power_law_1.2,0.02889983892440796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,32,8,power_law_1.2,0.031418879032135014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,32,8,power_law_1.2,0.032573440074920655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,32,8,power_law_1.2,0.03411200046539307
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,32,8,power_law_1.2,0.03705728054046631
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,32,8,power_law_1.2,0.04766079902648926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,32,8,power_law_1.2,0.04354688167572022
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,32,8,power_law_1.2,0.044791040420532224
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,32,8,power_law_1.2,0.49429374694824213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,32,8,power_law_1.2,0.046243839263916016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,32,8,power_law_1.2,0.046452479362487795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,32,8,power_law_1.2,0.0501580810546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,32,8,power_law_1.2,0.9041715240478515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,32,8,power_law_1.2,0.04809728145599365
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,32,8,power_law_1.2,0.04976640224456787
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,32,8,power_law_1.2,0.05129216194152832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,32,8,power_law_1.2,0.053880319595336915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,32,8,power_law_1.2,0.05997056007385254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,32,8,power_law_1.2,0.6373247909545898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,32,8,power_law_1.2,0.07143040180206299
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,32,8,power_law_1.2,0.08417535781860351
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,32,8,power_law_1.2,0.10752127647399903
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,32,8,power_law_1.2,0.14015999794006348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,32,8,power_law_1.2,0.1929292869567871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,32,8,power_law_1.2,0.04846464157104492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,32,8,power_law_1.2,0.443243522644043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,32,8,power_law_1.2,0.3441689682006836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,32,8,power_law_1.2,0.22811519622802737
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,8,32,balanced,0.02876672029495239
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,8,32,balanced,0.028471040725708007
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,8,32,balanced,0.028513278961181644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,8,32,balanced,0.03092736005783081
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,8,32,balanced,0.03254400014877319
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,8,32,balanced,0.06760447978973388
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,8,32,balanced,0.06816512107849121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,8,32,balanced,0.07008512020111084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,8,32,balanced,0.06588287830352783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,8,32,balanced,0.07446784019470215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,8,32,balanced,0.06800384044647216
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,8,32,balanced,0.06818048000335694
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,8,32,balanced,0.07074304103851319
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,8,32,balanced,0.08308992385864258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,8,32,balanced,0.08635007858276367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,8,32,balanced,0.09193984031677246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,8,32,balanced,0.10658816337585449
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,8,32,balanced,0.11747584342956544
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,32,8,power_law_1.01,0.053427200317382816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,8,32,balanced,0.12490880012512207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,8,32,balanced,0.1408345603942871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,8,32,balanced,0.15432831764221192
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,8,32,balanced,0.1821696090698242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,32,8,power_law_1.01,0.032409598827362066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,8,32,balanced,0.20997119903564454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,8,32,balanced,0.2642278480529785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,8,32,balanced,0.319866886138916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,32,8,power_law_1.2,0.6619251251220704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,8,32,balanced,0.4266611099243164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,8,32,balanced,0.0316761589050293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,8,32,balanced,0.5386304092407227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,8,32,balanced,0.03054975986480713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,8,32,balanced,0.033704960346221925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,8,32,balanced,0.03166591882705688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,8,32,balanced,0.03215872049331665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,8,32,balanced,0.04893824100494385
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,8,32,balanced,0.04318975925445557
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,8,32,balanced,0.04564864158630371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,8,32,balanced,0.0453004789352417
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,8,32,balanced,0.04770304203033447
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,8,32,balanced,0.046178560256958004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,8,32,balanced,0.05137919902801513
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,8,32,balanced,0.05012864112854004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,8,32,balanced,0.05444479942321777
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,8,32,balanced,0.05120639801025391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,8,32,balanced,0.055175681114196774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,8,32,balanced,0.04240511894226075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,8,32,balanced,0.06652544021606446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,8,32,balanced,0.057553920745849606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,8,32,balanced,0.07621632099151612
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,8,32,balanced,0.09594623565673828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,8,32,balanced,0.1146611213684082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,8,32,balanced,0.1506547164916992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,8,32,balanced,0.05077631950378418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,8,32,balanced,0.1880396842956543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,8,32,balanced,0.258917121887207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,8,32,balanced,0.3311296081542969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,32,8,power_law_1.2,0.859842529296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,8,32,balanced,0.0186572802066803
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,8,32,balanced,0.02067199945449829
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,8,32,balanced,0.02073983907699585
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,8,32,balanced,0.020710399150848387
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,8,32,balanced,0.020610558986663818
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,8,32,balanced,0.021025280952453613
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,8,32,balanced,0.021112320423126222
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,8,32,balanced,0.02149631977081299
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,8,32,balanced,0.021427199840545655
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,8,32,balanced,0.021406719684600832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,8,32,balanced,0.021425919532775877
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,8,32,balanced,0.021416959762573243
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,8,32,balanced,0.021468160152435304
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,8,32,balanced,0.019011839628219604
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,8,32,balanced,0.02190592050552368
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,8,32,balanced,0.02189055919647217
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,8,32,balanced,0.023136000633239746
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,8,32,balanced,0.019074560403823854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,32,8,power_law_1.2,1.186270751953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,8,32,balanced,0.02050303936004639
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,8,32,balanced,0.018602240085601806
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,8,32,balanced,0.020678400993347168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,8,32,power_law_1.01,0.03456768035888672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,8,32,power_law_1.01,0.033290240764617916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,8,32,power_law_1.01,0.032983040809631346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,8,32,power_law_1.01,0.03381119966506958
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,8,32,power_law_1.01,0.06071424007415771
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,8,32,power_law_1.01,0.06283904075622558
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,8,32,power_law_1.01,0.06380928039550782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,8,32,power_law_1.01,0.06984576225280761
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,8,32,power_law_1.01,0.07364480018615722
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,8,32,power_law_1.01,0.08315648078918457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,8,32,power_law_1.01,0.09773951530456543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,8,32,power_law_1.01,0.09816320419311524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,8,32,power_law_1.01,0.09547904014587402
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,8,32,power_law_1.01,0.11294719696044922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,8,32,power_law_1.01,0.11919232368469239
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,8,32,power_law_1.01,0.1266329574584961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,8,32,power_law_1.01,0.13544447898864748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,8,32,power_law_1.01,0.14356096267700197
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,8,32,power_law_1.01,0.1601433563232422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,8,32,power_law_1.01,0.20131711959838866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,8,32,power_law_1.01,0.24991487503051757
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,8,32,balanced,0.026460158824920654
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,8,32,balanced,0.028115200996398925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,8,32,power_law_1.01,0.3095603179931641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,8,32,power_law_1.01,0.408776969909668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,8,32,power_law_1.01,0.5596915054321289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,8,32,power_law_1.01,0.7133824157714843
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,8,32,balanced,0.03321856021881103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,8,32,power_law_1.01,0.03343744039535522
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,8,32,power_law_1.01,0.03417855978012085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,8,32,power_law_1.01,0.03346303939819336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,8,32,power_law_1.01,0.03493376016616821
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,8,32,power_law_1.01,1.0690943908691406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,8,32,power_law_1.01,0.043956480026245116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,8,32,power_law_1.01,0.042639360427856446
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,8,32,balanced,0.0684441614151001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,8,32,power_law_1.01,0.04977791786193848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,8,32,power_law_1.01,1.4437350463867187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,8,32,power_law_1.01,0.04699135780334472
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,8,32,power_law_1.01,0.04978559970855713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,8,32,power_law_1.01,0.04730368137359619
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,8,32,balanced,0.052460799217224116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,8,32,power_law_1.01,0.0522380781173706
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,8,32,power_law_1.01,0.05543424129486084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,8,32,power_law_1.01,0.054577918052673334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,8,32,balanced,0.042466559410095216
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,8,32,power_law_1.01,0.054915838241577156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,8,32,power_law_1.01,0.059494400024414064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,8,32,power_law_1.01,0.0647488021850586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,8,32,power_law_1.01,0.07369472026824951
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,8,32,power_law_1.01,0.08333567619323731
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,8,32,power_law_1.01,0.1078604793548584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,8,32,power_law_1.01,0.12705408096313478
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,8,32,power_law_1.01,0.17525375366210938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,8,32,power_law_1.01,0.2559244728088379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,8,32,power_law_1.01,0.04605440139770508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,8,32,power_law_1.01,0.3249280166625977
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,8,32,power_law_1.01,0.02038912057876587
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,8,32,power_law_1.01,0.02026623964309692
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,8,32,power_law_1.01,0.4637350463867188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,8,32,power_law_1.01,0.020625920295715333
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,8,32,power_law_1.01,0.02025599956512451
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,8,32,power_law_1.01,0.02109055995941162
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,8,32,power_law_1.01,0.7045657348632812
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,8,32,power_law_1.01,0.0210099196434021
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,8,32,power_law_1.01,0.020760319232940673
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,8,32,power_law_1.01,0.021072640419006347
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,8,32,power_law_1.01,0.021054720878601073
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,8,32,power_law_1.01,0.8761984252929688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,8,32,power_law_1.01,0.02150144100189209
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,8,32,power_law_1.01,0.02143104076385498
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,8,32,power_law_1.01,0.021457920074462893
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,8,32,power_law_1.01,0.021877760887146
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,8,32,power_law_1.01,0.02217600107192993
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,8,32,power_law_1.01,0.02598400115966797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,8,32,power_law_1.01,0.03381759881973266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,8,32,power_law_1.01,0.0214847993850708
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,8,32,power_law_1.01,0.0202892804145813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,8,32,power_law_1.01,0.028037118911743163
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,8,32,power_law_1.01,0.023522560596466062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,8,32,power_law_1.01,0.06308608055114746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,8,32,power_law_1.2,0.034719998836517336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,8,32,power_law_1.01,0.03719295978546143
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,8,32,power_law_1.01,0.04976640224456787
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,8,32,power_law_1.2,0.03293567895889282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,8,32,power_law_1.2,0.03576575994491577
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,8,32,power_law_1.2,0.05750783920288086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,8,32,power_law_1.2,0.06304128170013427
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,8,32,power_law_1.2,0.07090303897857667
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,8,32,power_law_1.2,0.07990911960601807
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,8,32,power_law_1.2,0.07385983943939209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,8,32,power_law_1.2,0.07502463817596436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,8,32,power_law_1.2,0.09046655654907226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,8,32,power_law_1.2,0.09986559867858887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,8,32,power_law_1.2,0.09666943550109863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,8,32,power_law_1.2,0.10347647666931152
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,8,32,power_law_1.2,0.12060031890869141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,8,32,power_law_1.2,0.12076416015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,8,32,power_law_1.2,0.1298419189453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,8,32,power_law_1.2,0.14330623626708985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,8,32,power_law_1.2,0.15400447845458984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,8,32,power_law_1.2,0.16996992111206055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,8,32,power_law_1.2,0.24285568237304686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,8,32,power_law_1.2,0.2614899253845215
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,8,32,power_law_1.01,0.08528639793395995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,8,32,power_law_1.2,0.4193459320068359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,8,32,power_law_1.2,0.39440383911132815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,8,32,power_law_1.2,0.6220467376708985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,8,32,power_law_1.2,0.033783040046691894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,8,32,power_law_1.2,0.0333132791519165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,8,32,power_law_1.2,0.03464319944381714
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,8,32,power_law_1.2,0.9627686309814454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,8,32,power_law_1.2,0.041519360542297365
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,8,32,power_law_1.2,0.044016637802124024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,8,32,power_law_1.2,0.04318848133087158
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,8,32,power_law_1.2,0.04594048023223877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,8,32,power_law_1.2,0.04838912010192871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,8,32,power_law_1.2,0.04662015914916992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,8,32,power_law_1.2,0.04863359928131104
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,8,32,power_law_1.2,0.052459521293640135
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,8,32,power_law_1.2,1.409812469482422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,8,32,power_law_1.2,0.05311744213104248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,8,32,power_law_1.2,0.05343232154846191
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,8,32,power_law_1.2,0.05448832035064697
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,8,32,power_law_1.2,0.06022655963897705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,8,32,power_law_1.2,0.0623795223236084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,8,32,power_law_1.2,0.07005951881408692
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,8,32,power_law_1.2,0.07915520191192627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,8,32,power_law_1.2,0.09469056129455566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,8,32,power_law_1.2,0.12932095527648926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,8,32,power_law_1.2,0.16194944381713866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,8,32,power_law_1.2,0.22309503555297852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,8,32,power_law_1.2,0.3247040176391601
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,8,32,power_law_1.2,0.43540607452392577
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,8,32,power_law_1.2,0.6082880020141601
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,8,32,power_law_1.2,0.020250880718231203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,8,32,power_law_1.2,0.020817921161651612
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,8,32,power_law_1.2,0.8151103973388671
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,8,32,power_law_1.2,0.02023808002471924
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,8,32,power_law_1.2,0.02025343894958496
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,8,32,power_law_1.2,0.020563199520111083
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,8,32,power_law_1.2,0.020536320209503172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,8,32,power_law_1.2,2.07140869140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,8,32,power_law_1.2,0.020885760784149172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,8,32,power_law_1.2,0.021056001186370847
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,8,32,power_law_1.2,0.021087999343872073
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,8,32,power_law_1.2,0.021087999343872073
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,8,32,power_law_1.2,0.021118719577789307
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,8,32,power_law_1.2,0.02135551929473877
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,8,32,power_law_1.2,0.02133888006210327
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,8,32,power_law_1.2,0.021943039894104004
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,8,32,power_law_1.2,0.022277119159698485
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,8,32,power_law_1.2,0.02282367944717407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,8,32,power_law_1.2,0.024747519493103026
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,8,32,power_law_1.2,1.196874237060547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,8,32,power_law_1.01,0.10666367530822754
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,8,32,power_law_1.2,0.08254847526550294
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,8,32,power_law_1.01,0.18846208572387696
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,8,32,power_law_1.01,0.18466175079345704
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,8,32,power_law_1.2,0.03069567918777466
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,8,32,power_law_1.2,0.0386406397819519
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,16,16,balanced,0.028424959182739257
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,8,32,power_law_1.2,0.04531455993652343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,16,16,balanced,0.028023040294647215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,16,16,balanced,0.030862081050872802
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,16,16,balanced,0.03132159948348999
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,16,16,balanced,0.0366374397277832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,16,16,balanced,0.0627839994430542
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,16,16,balanced,0.07385856151580811
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,16,16,balanced,0.06855296134948731
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,16,16,balanced,0.07017600059509277
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,16,16,balanced,0.0719155216217041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,16,16,balanced,0.0785753583908081
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,16,16,balanced,0.08417792320251465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,16,16,balanced,0.10128640174865722
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,16,16,balanced,0.09852928161621094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,16,16,balanced,0.11195903778076173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,16,16,balanced,0.11119487762451172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,16,16,balanced,0.11883904457092284
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,8,32,power_law_1.2,0.02720256090164185
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,16,16,balanced,0.13793279647827147
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,16,16,balanced,0.15089152336120606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,16,16,balanced,0.17916160583496094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,16,16,balanced,0.20360832214355468
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,16,16,balanced,0.255600643157959
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,8,32,power_law_1.2,0.06528255939483643
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,16,16,balanced,0.31062271118164064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,16,16,balanced,0.41222270965576174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,16,16,balanced,0.03118720054626465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,16,16,balanced,0.5127731323242187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,16,16,balanced,0.03129343986511231
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,16,16,balanced,0.03293823957443237
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,16,16,balanced,0.7180134582519531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,16,16,balanced,0.03339263916015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,16,16,balanced,0.039106559753417966
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,16,16,balanced,0.9214041900634765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,16,16,balanced,0.04892159938812256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,16,16,balanced,0.05058559894561767
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,16,16,balanced,0.04974463939666748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,16,16,balanced,0.05028992176055909
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,16,16,balanced,0.05188352108001709
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,16,16,balanced,0.053838081359863274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,16,16,balanced,0.054722561836242675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,16,16,balanced,0.05776895999908447
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,16,16,balanced,0.05594367980957031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,16,16,balanced,0.057244157791137694
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,16,16,balanced,0.05945216178894043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,16,16,balanced,0.06023168087005615
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,16,16,balanced,0.0696230411529541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,16,16,balanced,0.07877120018005371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,16,16,balanced,0.09499775886535644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,16,16,balanced,0.11117568016052246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,16,16,balanced,0.1515379238128662
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,16,16,balanced,0.18454143524169922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,16,16,balanced,0.25443071365356446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,16,16,balanced,0.32441726684570316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,16,16,balanced,0.4549593734741211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,16,16,balanced,0.595676155090332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,8,32,power_law_1.2,0.12597248077392578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,8,32,power_law_1.2,0.24499711990356446
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,8,32,power_law_1.2,0.15780863761901856
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,8,32,power_law_1.2,0.3398348617553711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,16,16,balanced,0.021500160694122313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,16,16,balanced,0.024364800453186036
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,16,16,balanced,0.019595520496368406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,16,16,balanced,0.024354560375213625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,16,16,balanced,0.02461695909500122
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,16,16,balanced,0.01943295955657959
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,16,16,balanced,0.024840960502624514
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,16,16,balanced,0.024291839599609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,16,16,balanced,0.024761600494384767
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,16,16,balanced,0.02551039934158325
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,16,16,balanced,0.01991935968399048
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,16,16,balanced,0.025163519382476806
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,16,16,balanced,0.025968639850616454
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,16,16,balanced,0.02687488079071045
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,16,16,balanced,0.027708160877227783
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,16,16,balanced,0.029323520660400393
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,16,16,balanced,0.023146240711212157
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,16,16,balanced,0.02396928071975708
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,16,16,balanced,0.024221439361572266
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,16,16,balanced,0.05609471797943115
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,16,16,power_law_1.01,0.03305088043212891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,16,16,power_law_1.01,0.03461632013320923
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,16,16,power_law_1.01,0.038257920742034913
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,16,16,balanced,0.043193597793579105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,16,16,power_law_1.01,0.03579904079437256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,16,16,power_law_1.01,0.06622079849243165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,16,16,power_law_1.01,0.06604032039642334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,16,16,power_law_1.01,0.06819712162017823
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,16,16,power_law_1.01,0.07320576190948487
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,16,16,power_law_1.01,0.07833856105804443
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,16,16,power_law_1.01,0.08702848434448243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,16,16,power_law_1.01,0.09181183815002442
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,16,16,power_law_1.01,0.10629119873046874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,16,16,power_law_1.01,0.11488256454467774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,16,16,power_law_1.01,0.11428352355957032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,16,16,power_law_1.01,0.12570112228393554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,16,16,power_law_1.01,0.12877440452575684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,16,16,power_law_1.01,0.13940223693847656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,16,16,power_law_1.01,0.15658495903015138
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,16,16,power_law_1.01,0.17624319076538086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,16,16,power_law_1.01,0.2081203269958496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,16,16,power_law_1.01,0.2511052894592285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,16,16,power_law_1.01,0.3313190460205078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,16,16,balanced,0.06799232006072999
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,16,16,power_law_1.01,0.39044097900390623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,16,16,power_law_1.01,0.5846105575561523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,16,16,power_law_1.01,1.1462413024902343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,16,16,power_law_1.01,0.03663615942001343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,16,16,power_law_1.01,0.03786240100860595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,16,16,power_law_1.01,1.4434803771972657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,16,16,power_law_1.01,0.036871678829193115
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,16,16,power_law_1.01,0.03831039905548096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,16,16,power_law_1.01,0.045898241996765135
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,16,16,power_law_1.01,0.04891903877258301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,16,16,power_law_1.01,0.05014272212982178
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,16,16,power_law_1.01,0.05046144008636475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,16,16,power_law_1.01,0.05102975845336914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,16,16,power_law_1.01,0.055129599571228025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,16,16,power_law_1.01,0.05405055999755859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,16,16,power_law_1.01,0.05885439872741699
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,16,16,power_law_1.01,0.058362879753112794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,16,16,power_law_1.01,0.060038399696350095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,16,16,power_law_1.01,0.0632204818725586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,16,16,power_law_1.01,0.06537727832794189
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,16,16,power_law_1.01,0.7052851104736328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,16,16,power_law_1.01,0.08354432106018066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,16,16,power_law_1.01,0.07086976051330567
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,16,16,power_law_1.01,0.1274508762359619
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,16,16,power_law_1.01,0.0965452766418457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,16,16,power_law_1.01,0.1530470371246338
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,16,16,power_law_1.01,0.19864704132080077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,16,16,power_law_1.01,0.2619033622741699
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,16,16,power_law_1.01,0.37481342315673827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,16,16,power_law_1.01,0.48992767333984377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,16,16,power_law_1.01,0.804501724243164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,16,16,power_law_1.01,0.023932158946990967
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,16,16,power_law_1.01,1.1174527740478515
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,16,16,power_law_1.01,0.022835199832916257
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,16,16,balanced,0.08743295669555665
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,16,16,power_law_1.01,0.023089919090270996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,16,16,power_law_1.01,0.024286720752716064
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,16,16,power_law_1.01,0.023491840362548828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,16,16,power_law_1.01,0.023942399024963378
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,16,16,power_law_1.01,0.024309759140014646
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,16,16,power_law_1.01,0.024810240268707276
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,16,16,power_law_1.01,0.02443903923034668
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,16,16,power_law_1.01,0.024550399780273437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,16,16,power_law_1.01,0.024241919517517092
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,16,16,power_law_1.01,0.026112000942230224
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,16,16,power_law_1.01,0.025408000946044923
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,16,16,power_law_1.01,0.02644479990005493
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,16,16,power_law_1.01,0.03089920043945312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,16,16,power_law_1.01,0.028129279613494873
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,16,16,power_law_1.01,0.036444160938262936
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,16,16,power_law_1.01,0.044408321380615234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,16,16,power_law_1.01,0.056721920967102046
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,16,16,power_law_1.01,0.07047039985656738
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,16,16,power_law_1.01,0.1351116752624512
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,16,16,power_law_1.01,0.08801792144775392
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,16,16,power_law_1.2,0.03293951988220215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,16,16,power_law_1.2,0.036432640552520754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,16,16,power_law_1.2,0.06819968223571778
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,16,16,power_law_1.2,0.07191167831420898
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,16,16,power_law_1.01,0.1487168025970459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,16,16,power_law_1.2,0.07158912181854248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,16,16,power_law_1.2,0.07797120094299317
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,16,16,power_law_1.2,0.0779417610168457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,16,16,power_law_1.2,0.10045696258544923
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,16,16,power_law_1.2,0.10052607536315919
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,16,16,power_law_1.2,0.11146368026733397
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,16,16,power_law_1.2,0.11799551963806151
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,16,16,power_law_1.2,0.1272332763671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,16,16,power_law_1.2,0.13214464187622071
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,16,16,power_law_1.2,0.14041855812072754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,16,16,power_law_1.2,0.18706560134887695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,16,16,power_law_1.2,0.2323699188232422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,16,16,power_law_1.2,0.27101184844970705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,16,16,power_law_1.2,0.3545177459716797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,16,16,power_law_1.2,0.44531711578369143
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,16,16,power_law_1.2,0.6528614044189454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,16,16,power_law_1.2,0.9407360076904296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,16,16,power_law_1.2,1.3690573120117189
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,16,16,power_law_1.2,0.03342592000961304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,16,16,power_law_1.2,0.16299520492553712
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,16,16,power_law_1.2,0.035037438869476314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,16,16,power_law_1.2,0.035825920104980466
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,16,16,power_law_1.2,0.03800575971603394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,16,16,power_law_1.2,0.07307136058807373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,16,16,power_law_1.2,0.04593664169311523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,16,16,power_law_1.2,0.0468390417098999
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,16,16,power_law_1.2,0.04975103855133057
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,16,16,power_law_1.2,0.04975359916687012
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,16,16,power_law_1.2,0.051662077903747564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,16,16,power_law_1.2,0.053903360366821286
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,16,16,power_law_1.2,0.056268801689147954
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,16,16,power_law_1.01,0.26767103195190434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,16,16,power_law_1.2,0.04091392040252685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,16,16,power_law_1.2,0.05798399925231934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,16,16,power_law_1.2,0.0567526388168335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,16,16,power_law_1.2,0.03559936046600341
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,16,16,power_law_1.2,0.06271872043609619
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,16,16,power_law_1.2,0.07291647911071777
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,16,16,power_law_1.2,0.06704768180847168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,16,16,power_law_1.2,0.08109312057495117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,16,16,power_law_1.2,0.10113792419433594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,16,16,power_law_1.2,0.13862272262573244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,16,16,power_law_1.2,0.06086527824401855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,16,16,power_law_1.2,0.17380352020263673
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,16,16,power_law_1.2,0.2264089584350586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,16,16,power_law_1.2,0.30113792419433594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,16,16,power_law_1.2,1.9032754516601564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,16,16,power_law_1.2,0.4567769622802734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,16,16,power_law_1.2,0.023125760555267334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,16,16,power_law_1.2,0.023941121101379394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,16,16,power_law_1.2,0.8798873901367188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,16,16,power_law_1.2,0.6219865417480468
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,16,16,power_law_1.2,0.022812159061431886
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,16,16,power_law_1.2,1.215326690673828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,16,16,power_law_1.2,0.023978240489959717
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,16,16,power_law_1.2,0.024350719451904295
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,16,16,power_law_1.2,0.024773120880126953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,16,16,power_law_1.2,0.024714241027832033
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,16,16,power_law_1.2,0.025207040309906004
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,16,16,power_law_1.2,0.025363199710845948
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,16,16,power_law_1.2,0.026430718898773194
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,16,16,power_law_1.2,0.026853120326995848
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,16,16,power_law_1.2,0.030566399097442624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,16,16,power_law_1.2,0.036788480281829836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,16,16,power_law_1.2,0.024019200801849366
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,16,16,power_law_1.2,0.023907840251922607
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,16,16,power_law_1.2,0.02404736042022705
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,16,16,power_law_1.2,0.04676095962524414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,16,16,power_law_1.2,0.06973567962646485
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,16,16,power_law_1.2,0.10483967781066894
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,16,16,power_law_1.2,0.04873216152191162
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,16,16,power_law_1.2,0.16222591400146485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,32,8,balanced,0.028843519687652586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,32,8,balanced,0.03049855947494507
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,32,8,balanced,0.032143359184265134
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,32,8,balanced,0.03469568014144898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,32,8,balanced,0.04030335903167724
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,32,8,balanced,0.07206272125244141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,32,8,balanced,0.07669248104095458
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,32,8,balanced,0.08070272445678711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,32,8,balanced,0.10546560287475586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,32,8,balanced,0.11437952041625976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,32,8,balanced,0.1098303985595703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,32,8,balanced,0.11166975975036622
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,32,8,balanced,0.11650431632995606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,32,8,balanced,0.12771583557128907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,32,8,balanced,0.13204352378845213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,32,8,balanced,0.1402457618713379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,32,8,balanced,0.15407999992370605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,32,8,balanced,0.17937536239624025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,32,8,balanced,0.205664005279541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,32,8,balanced,0.2574822425842285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,32,8,balanced,0.3081945610046387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,32,8,balanced,0.411541748046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,32,8,balanced,0.5137267303466797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,32,8,balanced,0.7108684539794922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,32,8,balanced,0.9103411102294923
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,32,8,balanced,1.3122329711914062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,32,8,balanced,0.03126912117004395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,32,8,balanced,0.033314559459686276
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,16,16,power_law_1.2,0.21120512008666995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,32,8,balanced,0.03674367904663086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,32,8,balanced,0.04360832214355469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,32,8,balanced,0.0596889591217041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,32,8,balanced,0.05916543960571289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,32,8,balanced,0.0611955213546753
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,32,8,balanced,0.03871104001998901
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,32,8,balanced,0.06242432117462158
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,32,8,balanced,0.06454144001007081
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,32,8,balanced,0.0608460807800293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,32,8,balanced,0.0634496021270752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,32,8,balanced,0.06532991886138916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,32,8,balanced,0.07015168190002441
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,32,8,balanced,0.07098368167877198
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,32,8,balanced,0.07375103950500488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,32,8,balanced,0.0838105583190918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,32,8,balanced,0.09905407905578614
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,32,8,balanced,0.11694463729858398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,32,8,balanced,0.14878463745117188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,32,8,balanced,0.18035711288452147
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,32,8,balanced,0.25542783737182617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,32,8,balanced,0.3168691253662109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,32,8,balanced,0.44875137329101566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,32,8,balanced,0.5839244842529296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,32,8,balanced,0.8361753845214844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,32,8,balanced,1.0998028564453126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,32,8,balanced,1.6984396362304686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,32,8,power_law_1.01,0.035022079944610596
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,32,8,power_law_1.01,0.03647615909576416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,32,8,power_law_1.01,0.04443647861480713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,32,8,power_law_1.01,0.07457536220550537
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,32,8,power_law_1.01,0.07718783855438233
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,32,8,power_law_1.01,0.0362611198425293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,32,8,power_law_1.01,0.08416255950927734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,32,8,power_law_1.01,0.08066047668457031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,32,8,power_law_1.01,0.09854847908020019
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,32,8,power_law_1.01,0.11209983825683593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,32,8,power_law_1.01,0.11036031723022462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,32,8,power_law_1.01,0.11677824020385744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,32,8,power_law_1.01,0.1251097583770752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,32,8,power_law_1.01,0.13011712074279785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,32,8,power_law_1.01,0.134901762008667
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,32,8,power_law_1.01,0.14203392028808595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,32,8,power_law_1.01,0.16781951904296874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,32,8,power_law_1.01,0.19068031311035155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,32,8,power_law_1.01,0.2351513671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,32,8,power_law_1.01,0.2820582389831543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,32,8,power_law_1.01,0.35730049133300784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,32,8,power_law_1.01,0.4853363037109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,32,8,power_law_1.01,0.6521459197998046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,32,8,power_law_1.01,0.7977804565429688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,32,8,power_law_1.01,1.2570047760009766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,32,8,power_law_1.01,0.036015360355377196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,32,8,power_law_1.01,0.041094398498535155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,32,8,power_law_1.01,1.6281701660156251
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,32,8,power_law_1.01,0.03909120082855225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,32,8,power_law_1.01,0.04205440044403076
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,32,8,power_law_1.01,0.04397439956665039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,32,8,power_law_1.01,0.057921280860900884
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,32,8,power_law_1.01,0.05724671840667724
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,32,8,power_law_1.01,0.058686718940734864
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,32,8,power_law_1.01,0.060695037841796876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,32,8,power_law_1.01,0.03342080116271973
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,32,8,power_law_1.01,0.0626688003540039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,32,8,power_law_1.01,0.06536320209503174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,32,8,power_law_1.01,0.06369279861450196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,32,8,power_law_1.01,0.06858751773834229
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,32,8,power_law_1.01,0.062463998794555664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,32,8,power_law_1.01,0.07049215793609619
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,32,8,power_law_1.01,0.07252096176147461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,32,8,power_law_1.01,0.07856512069702148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,32,8,power_law_1.01,0.09382911682128905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,32,8,power_law_1.01,0.11505663871765137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,32,8,power_law_1.01,0.14162431716918947
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,32,8,power_law_1.01,0.17039487838745118
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,32,8,power_law_1.01,0.24024063110351562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,32,8,power_law_1.01,0.294069766998291
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,32,8,power_law_1.01,0.3975385665893555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,32,8,power_law_1.01,0.5066662216186524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,32,8,power_law_1.2,0.034946560859680176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,32,8,power_law_1.01,0.7945996856689452
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,32,8,power_law_1.2,0.03299455881118775
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,32,8,power_law_1.2,0.040316162109375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,32,8,power_law_1.01,1.0152435302734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,32,8,power_law_1.2,0.035000319480896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,32,8,power_law_1.2,0.047791361808776855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,32,8,power_law_1.2,0.07312255859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,32,8,power_law_1.2,0.08472831726074219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,32,8,power_law_1.2,0.10650624275207518
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,32,8,power_law_1.2,0.09040255546569824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,32,8,power_law_1.2,0.08339839935302734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,32,8,power_law_1.2,0.10798848152160645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,32,8,power_law_1.2,0.11400832176208495
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,32,8,power_law_1.2,0.12548735618591308
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,32,8,power_law_1.2,0.11193984031677247
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,32,8,power_law_1.2,0.13301888465881348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,32,8,power_law_1.2,0.13942784309387207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,32,8,power_law_1.2,0.14874752044677736
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,32,8,power_law_1.2,0.17693824768066407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,32,8,power_law_1.2,0.20385919570922853
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,32,8,power_law_1.2,0.24039680480957032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,32,8,power_law_1.2,0.29755647659301754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,32,8,power_law_1.2,0.3928729629516602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,32,8,power_law_1.2,0.520398063659668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,32,8,power_law_1.2,0.8996685028076172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,32,8,power_law_1.2,0.03632256031036377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,32,8,power_law_1.2,0.03750911951065063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,32,8,power_law_1.2,0.03965183973312378
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,32,8,power_law_1.2,0.04209536075592041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,32,8,power_law_1.2,1.366746826171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,32,8,power_law_1.2,0.04361472129821777
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,32,8,power_law_1.2,0.05672832012176514
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,32,8,power_law_1.2,0.05583487987518311
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,32,8,power_law_1.2,0.058731517791748046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,32,8,power_law_1.2,0.06247424125671387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,32,8,power_law_1.2,0.0650816011428833
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,32,8,power_law_1.2,0.06249855995178223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,32,8,power_law_1.2,0.06288256168365478
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,32,8,power_law_1.2,0.06992127895355224
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,32,8,power_law_1.2,0.060096001625061034
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,32,8,power_law_1.2,0.07516032218933105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,32,8,power_law_1.2,0.0828825569152832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,32,8,power_law_1.2,0.09454848289489745
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,32,8,power_law_1.2,0.695367660522461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,32,8,power_law_1.2,0.1132761573791504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,32,8,power_law_1.2,1.8385267639160154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,32,8,power_law_1.2,0.14241536140441896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,32,8,power_law_1.2,0.07065855979919433
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,32,8,power_law_1.2,0.25297664642333983
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,32,8,power_law_1.2,0.3306022262573242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,32,8,power_law_1.2,0.46695808410644535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,32,8,power_law_1.2,0.18280319213867186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,32,8,power_law_1.2,0.6240716934204101
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,32,8,power_law_1.2,1.220016632080078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,128,balanced,0.03359744071960449
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,128,balanced,0.033890559673309326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,128,balanced,0.03412735939025879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,128,balanced,0.035016961097717285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,128,balanced,0.036981759071350095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,128,balanced,0.06266751766204834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,128,balanced,0.1242905616760254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,128,balanced,0.12430080413818359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,128,balanced,0.1264409637451172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,128,balanced,0.1267097568511963
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,128,balanced,0.12695679664611817
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,128,balanced,0.12623104095458984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,128,balanced,0.13009407997131348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,128,balanced,0.13736063957214356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,128,balanced,0.13050880432128906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,128,balanced,0.1323148822784424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,128,balanced,0.13680255889892579
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,128,balanced,0.14187007904052734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,128,balanced,0.14973055839538574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,128,balanced,0.14755968093872068
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,128,balanced,0.15488384246826173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,128,balanced,0.16564735412597656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,128,balanced,0.17899391174316406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,128,balanced,0.2046988868713379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,128,balanced,0.24198911666870115
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,128,balanced,0.28434431076049804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,128,balanced,0.03504640102386475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,128,balanced,0.03499775886535644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,128,balanced,0.03535871982574463
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,128,balanced,0.03500927925109863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,128,balanced,0.03500927925109863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,128,balanced,0.03752831935882568
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,128,balanced,0.037557759284973145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,128,balanced,0.04765439987182617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,128,balanced,0.04600063800811768
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,128,balanced,0.04768511772155762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,128,balanced,0.04642560005187989
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,128,balanced,0.06851200103759766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,128,balanced,0.05051904201507569
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,128,balanced,0.04671999931335449
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,128,balanced,0.048089599609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,128,balanced,0.05669760227203369
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,128,balanced,0.050091519355773925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,128,balanced,0.049848318099975586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,128,balanced,0.05217919826507569
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,128,balanced,0.054656000137329096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,128,balanced,0.06405888080596925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,128,balanced,0.05948287963867187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,128,balanced,0.06731776237487794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,128,balanced,0.07109504222869874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,128,balanced,0.0821951961517334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,128,balanced,0.09542655944824219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,128,balanced,0.11686143875122071
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,128,balanced,0.14320639610290525
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,32,8,power_law_1.2,0.8932249450683594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,2,128,balanced,0.01979519963264465
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,2,128,balanced,0.021030399799346924
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,2,128,balanced,0.02153599977493286
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,2,128,balanced,0.021149439811706545
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,2,128,balanced,0.019514880180358886
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,2,128,balanced,0.021384959220886228
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,2,128,balanced,0.02117759943008423
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,2,128,balanced,0.021418240070343018
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,2,128,balanced,0.02108288049697876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,2,128,balanced,0.021424639225006103
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,2,128,balanced,0.01994495987892151
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,2,128,balanced,0.02115839958190918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,2,128,balanced,0.021207039356231687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,2,128,balanced,0.021507840156555175
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,2,128,balanced,0.021466879844665526
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,2,128,balanced,0.021505279541015627
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,2,128,balanced,0.021858561038970947
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,2,128,balanced,0.021822719573974608
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,2,128,balanced,0.022572801113128663
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,2,128,balanced,0.022720000743865966
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,2,128,balanced,0.02346112012863159
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,2,128,balanced,0.018759679794311524
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,2,128,balanced,0.019854079484939575
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,128,power_law_1.01,0.0696127986907959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,128,power_law_1.01,0.06673664093017578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,128,power_law_1.01,0.06963327884674073
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,128,power_law_1.01,0.12310015678405761
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,128,power_law_1.01,0.12356224060058593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,128,power_law_1.01,0.1266470432281494
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,128,power_law_1.01,0.06955008029937744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,128,power_law_1.01,0.12770432472229004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,128,power_law_1.01,0.13199872016906739
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,128,power_law_1.01,0.12840448379516603
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,128,power_law_1.01,0.13916031837463377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,128,power_law_1.01,0.1371891212463379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,128,power_law_1.01,0.13363583564758302
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,128,power_law_1.01,0.1354905605316162
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,128,power_law_1.01,0.1412236785888672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,128,power_law_1.01,0.14417792320251466
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,128,power_law_1.01,0.14973952293395995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,128,power_law_1.01,0.15205632209777833
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,128,power_law_1.01,0.16305791854858398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,128,power_law_1.01,0.17786880493164064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,128,power_law_1.01,0.2115519905090332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,128,power_law_1.01,0.24815231323242187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,128,power_law_1.01,0.27853439331054686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,128,power_law_1.01,0.4499622344970703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,128,power_law_1.01,0.5267212677001953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,128,power_law_1.01,0.8523980712890624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,128,power_law_1.01,0.03699455976486206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,128,power_law_1.01,1.1548070526123047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,128,power_law_1.01,0.037922561168670654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,128,power_law_1.01,0.03788928031921387
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,128,power_law_1.01,0.04564352035522461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,128,power_law_1.01,0.03791359901428223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,128,power_law_1.01,0.046407680511474605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,128,power_law_1.01,0.04981247901916504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,128,power_law_1.01,0.04756864070892334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,128,power_law_1.01,0.052645120620727535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,128,power_law_1.01,0.04888448238372803
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,128,power_law_1.01,0.05756288051605225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,128,power_law_1.01,0.05231488227844239
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,128,power_law_1.01,0.05225471973419189
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,128,power_law_1.01,0.0526310396194458
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,128,power_law_1.01,0.055429120063781735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,128,power_law_1.01,0.059261441230773926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,128,power_law_1.01,0.06081791877746582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,128,power_law_1.01,0.06819200038909912
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,128,power_law_1.01,0.07287168025970459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,128,power_law_1.01,0.0832806396484375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,128,power_law_1.01,0.09480575561523438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,128,power_law_1.01,0.13422335624694823
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,128,power_law_1.01,0.15319935798645018
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,128,power_law_1.01,0.2058572769165039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,128,power_law_1.01,0.2755955123901367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,128,power_law_1.01,0.4656358337402343
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,2,128,power_law_1.01,0.021096959114074706
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,128,power_law_1.01,0.5772249603271484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,2,128,power_law_1.01,0.021038079261779787
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,2,128,power_law_1.01,0.021340160369873046
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,2,128,power_law_1.01,0.021340160369873046
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,2,128,power_law_1.01,0.020819199085235596
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,2,128,power_law_1.01,0.020625920295715333
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,2,128,power_law_1.01,0.02104448080062866
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,2,128,power_law_1.01,0.021437439918518066
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,2,128,power_law_1.01,0.021020159721374512
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,2,128,balanced,0.029889280796051028
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,2,128,power_law_1.01,0.02096895933151245
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,2,128,power_law_1.01,0.021381120681762695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,2,128,power_law_1.01,0.021483519077301026
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,2,128,power_law_1.01,0.021182720661163328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,2,128,power_law_1.01,0.021016321182250976
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,2,128,power_law_1.01,0.02149631977081299
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,2,128,power_law_1.01,0.02184319972991943
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,2,128,power_law_1.01,0.022164480686187742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,2,128,power_law_1.01,0.022163200378417968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,2,128,power_law_1.01,0.024739840030670167
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,2,128,balanced,0.026818559169769284
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,2,128,balanced,0.035938560962677
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,2,128,balanced,0.019481600522994996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,2,128,power_law_1.01,0.03915519952774048
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,2,128,power_law_1.01,0.02780416011810303
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,2,128,power_law_1.01,0.023343360424041747
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,128,power_law_1.2,0.06861055850982666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,128,power_law_1.2,0.06886911869049073
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,128,power_law_1.2,0.07196800231933594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,128,power_law_1.2,0.06917632102966309
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,128,power_law_1.2,0.12453760147094725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,128,power_law_1.2,0.12268032073974608
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,2,128,power_law_1.01,0.04627200126647949
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,128,power_law_1.2,0.1292569637298584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,128,power_law_1.2,0.12623488426208496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,128,power_law_1.2,0.1339750385284424
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,2,128,power_law_1.01,0.06376959800720214
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,128,power_law_1.2,0.13513216018676757
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,128,power_law_1.2,0.13260800361633301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,128,power_law_1.2,0.13613696098327638
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,128,power_law_1.2,0.13753087997436525
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,128,power_law_1.2,0.1435878372192383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,128,power_law_1.2,0.13775872230529784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,128,power_law_1.2,0.14581631660461425
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,128,power_law_1.2,0.15363327980041502
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,128,power_law_1.2,0.1614668846130371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,128,power_law_1.2,0.17414783477783202
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,128,power_law_1.2,0.20717824935913084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,128,power_law_1.2,0.23844863891601564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,128,power_law_1.2,0.3368588638305664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,128,power_law_1.2,0.3791475296020508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,128,power_law_1.2,0.588535041809082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,128,power_law_1.2,0.03780479907989502
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,128,power_law_1.2,0.7967833709716797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,128,power_law_1.2,0.03787519931793213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,128,power_law_1.2,0.038242559432983395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,128,power_law_1.2,0.0378329610824585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,128,power_law_1.2,0.046964478492736814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,128,power_law_1.2,0.04895616054534912
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,128,power_law_1.2,1.166118392944336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,128,power_law_1.2,0.04688384056091309
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,128,power_law_1.2,0.05313280105590821
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,128,power_law_1.2,0.050895361900329596
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,128,power_law_1.2,0.05091072082519531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,128,power_law_1.2,0.049839358329772945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,128,power_law_1.2,0.051016960144042965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,128,power_law_1.2,1.68553466796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,2,128,power_law_1.01,0.06586495876312257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,128,power_law_1.2,0.05432831764221191
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,128,power_law_1.2,0.05979904174804688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,128,power_law_1.2,0.06857728004455567
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,128,power_law_1.2,0.06689663887023925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,128,power_law_1.2,0.07967872142791747
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,128,power_law_1.2,0.10844287872314454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,128,power_law_1.2,0.112542724609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,128,power_law_1.2,0.16107135772705078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,128,power_law_1.2,0.198470401763916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,128,power_law_1.2,0.31415807723999023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,128,power_law_1.2,0.050681600570678706
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,128,power_law_1.2,0.4165184020996094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,2,128,power_law_1.2,0.02135551929473877
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,2,128,power_law_1.2,0.021004800796508787
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,128,power_law_1.2,0.6775820922851563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,2,128,power_law_1.2,0.02120448112487793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,2,128,power_law_1.2,0.021064960956573488
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,2,128,power_law_1.2,0.021020159721374512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,128,power_law_1.2,0.8394560241699219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,2,128,power_law_1.2,0.020687360763549805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,2,128,power_law_1.2,0.020684800148010253
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,2,128,power_law_1.2,0.020992000102996827
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,2,128,power_law_1.2,0.021029119491577146
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,2,128,power_law_1.2,0.021086719036102295
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,2,128,power_law_1.2,0.020771839618682862
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,2,128,power_law_1.2,0.0210099196434021
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,2,128,power_law_1.2,0.021135358810424804
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,2,128,power_law_1.2,0.021477119922637938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,2,128,power_law_1.2,0.021483519077301026
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,2,128,power_law_1.2,0.02191231966018677
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,2,128,power_law_1.2,0.021867520809173584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,2,128,power_law_1.2,0.023956480026245116
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,2,128,power_law_1.2,0.03749248027801514
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,128,power_law_1.2,0.07169151782989502
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,2,128,power_law_1.01,0.09064703941345215
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,2,128,power_law_1.01,0.11957247734069824
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,2,128,power_law_1.2,0.023294720649719238
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,2,128,power_law_1.2,0.030499839782714845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,64,balanced,0.031752960681915285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,2,128,power_law_1.2,0.058823680877685545
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,2,128,power_law_1.2,0.08888319969177247
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,64,balanced,0.03175551891326904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,64,balanced,0.03536384105682373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,64,balanced,0.03580159902572631
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,64,balanced,0.0602623987197876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,64,balanced,0.07187967777252198
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,64,balanced,0.12120320320129394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,64,balanced,0.12190591812133789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,4,64,balanced,0.0331328010559082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,64,balanced,0.12552831649780274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,64,balanced,0.12302080154418946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,64,balanced,0.12616703987121583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,64,balanced,0.12940287590026855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,64,balanced,0.1306726360321045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,64,balanced,0.13328895568847657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,64,balanced,0.1376371192932129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,64,balanced,0.14483839988708497
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,64,balanced,0.14125823974609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,64,balanced,0.14916735649108886
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,64,balanced,0.15443455696105957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,64,balanced,0.16702335357666015
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,2,128,power_law_1.2,0.10873087882995605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,64,balanced,0.1885491180419922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,64,balanced,0.21840255737304687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,64,balanced,0.2571724891662598
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,64,balanced,0.3167500877380371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,64,balanced,0.38862846374511717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,64,balanced,0.03380095958709717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,64,balanced,0.03419264078140259
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,64,balanced,0.03375744104385376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,64,balanced,0.03572096109390259
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,64,balanced,0.0365068793296814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,64,balanced,0.040712962150573725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,64,balanced,0.05027200222015381
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,64,balanced,0.050623998641967774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,64,balanced,0.05067520141601563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,64,balanced,0.05226751804351807
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,64,balanced,0.12803071975708008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,64,balanced,0.05201663970947266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,64,balanced,0.0528000020980835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,64,balanced,0.052988162040710454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,64,balanced,0.05467904090881348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,64,balanced,0.05715712070465088
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,64,balanced,0.06023039817810059
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,2,128,power_law_1.2,0.050763521194458004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,64,balanced,0.061064958572387695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,64,balanced,0.06736256122589111
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,64,balanced,0.06927103996276855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,64,balanced,0.08000255584716796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,64,balanced,0.03340928077697754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,64,balanced,0.08869888305664062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,64,balanced,0.10980735778808595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,64,balanced,0.13445119857788085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,64,balanced,0.17120000839233399
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,64,balanced,0.2185907173156738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,64,balanced,0.05306367874145508
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,2,128,power_law_1.2,0.13782912254333496
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,2,128,power_law_1.2,0.1909734344482422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,4,64,balanced,0.01949952006340027
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,4,64,balanced,0.022709760665893555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,4,64,balanced,0.019560960531234742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,4,64,balanced,0.02235136032104492
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,4,64,balanced,0.02266752004623413
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,4,64,balanced,0.022708480358123777
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,4,64,balanced,0.02268160104751587
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,4,64,balanced,0.02261888027191162
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,4,64,balanced,0.022741758823394777
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,4,64,balanced,0.022726399898529055
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,4,64,balanced,0.022593278884887696
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,4,64,balanced,0.02267647981643677
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,4,64,balanced,0.019600640535354614
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,4,64,balanced,0.022833919525146483
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,4,64,balanced,0.023948800563812257
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,4,64,balanced,0.019403519630432128
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,4,64,balanced,0.023582720756530763
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,4,64,balanced,0.02114432096481323
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,4,64,balanced,0.02479487895965576
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,4,64,balanced,0.025144319534301757
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,4,64,balanced,0.027200000286102293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,4,64,balanced,0.021518080234527587
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,4,64,power_law_1.01,0.06856832027435303
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,64,power_law_1.01,0.07094528198242187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,64,power_law_1.01,0.12105855941772461
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,4,64,balanced,0.02000511884689331
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,64,power_law_1.01,0.11986559867858886
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,64,power_law_1.01,0.1252787208557129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,64,power_law_1.01,0.12134528160095215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,64,power_law_1.01,0.1293452835083008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,64,power_law_1.01,0.13285887718200684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,64,power_law_1.01,0.12520575523376465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,64,power_law_1.01,0.129039363861084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,64,power_law_1.01,0.13088255882263183
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,64,power_law_1.01,0.135928316116333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,64,power_law_1.01,0.14252927780151367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,64,power_law_1.01,0.1476159954071045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,64,power_law_1.01,0.14537343978881836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,64,power_law_1.01,0.15320704460144044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,64,power_law_1.01,0.15925888061523436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,64,power_law_1.01,0.18393856048583984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,64,power_law_1.01,0.19287679672241212
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,64,power_law_1.01,0.067740159034729
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,64,power_law_1.01,0.276810245513916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,64,power_law_1.01,0.30448768615722654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,64,power_law_1.01,0.44603134155273433
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,64,power_law_1.01,0.5399257659912109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,64,power_law_1.01,0.03960959911346436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,64,power_law_1.01,0.7880204772949219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,64,power_law_1.01,0.04033792018890381
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,64,power_law_1.01,0.03990528106689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,64,power_law_1.01,1.0965670776367187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,64,power_law_1.01,0.04279551982879638
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,64,power_law_1.01,0.049329919815063475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,64,power_law_1.01,0.04854144096374512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,64,power_law_1.01,0.05466879844665527
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,64,power_law_1.01,0.05220096111297608
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,64,power_law_1.01,0.052185602188110344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,64,power_law_1.01,0.06079616069793701
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,64,power_law_1.01,0.05505407810211181
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,64,power_law_1.01,0.05515903949737548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,64,power_law_1.01,0.05876224040985108
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,64,power_law_1.01,0.06700032234191895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,64,power_law_1.01,0.06632063865661622
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,64,power_law_1.01,0.052225279808044436
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,4,64,balanced,0.021875200271606447
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,64,power_law_1.01,0.0709555196762085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,64,power_law_1.01,0.08717823982238769
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,64,power_law_1.01,0.1058841609954834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,64,power_law_1.01,0.064236798286438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,64,power_law_1.01,0.13090047836303711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,64,power_law_1.01,0.1734899139404297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,64,power_law_1.01,0.2395801544189453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,64,power_law_1.01,0.30161535263061523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,64,power_law_1.01,0.43338623046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,64,power_law_1.01,0.5645478439331055
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,4,64,power_law_1.01,0.022371840476989747
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,4,64,power_law_1.01,0.02238464117050171
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,4,64,power_law_1.01,0.022342400550842283
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,4,64,power_law_1.01,0.022663679122924805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,4,64,power_law_1.01,0.021857280731201172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,4,64,power_law_1.01,0.022307839393615723
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,4,64,power_law_1.01,0.022728960514068603
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,4,64,power_law_1.01,0.02232192039489746
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,4,64,power_law_1.01,0.022254080772399903
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,4,64,power_law_1.01,0.021908481121063233
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,4,64,power_law_1.01,0.02191231966018677
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,4,64,power_law_1.01,0.022428159713745115
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,4,64,power_law_1.01,0.021996800899505613
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,4,64,power_law_1.01,0.022142720222473145
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,4,64,power_law_1.01,0.02230015993118286
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,64,power_law_1.01,0.07478271961212159
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,4,64,power_law_1.01,0.022717440128326417
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,4,64,power_law_1.01,0.023335680961608887
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,4,64,power_law_1.01,0.02393984079360962
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,64,power_law_1.01,0.06940288066864013
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,4,64,balanced,0.03487744092941284
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,4,64,balanced,0.03788160085678101
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,4,64,balanced,0.049181442260742184
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,4,64,power_law_1.01,0.039631359577178955
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,4,64,power_law_1.01,0.025605120658874513
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,4,64,power_law_1.01,0.030001919269561767
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,4,64,power_law_1.01,0.033070080280303955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,64,power_law_1.2,0.0681766414642334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,64,power_law_1.2,0.06903168201446533
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,64,power_law_1.2,0.07175936222076415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,64,power_law_1.2,0.12087936401367187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,64,power_law_1.2,0.12368512153625488
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,4,64,power_law_1.01,0.04574463844299316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,64,power_law_1.2,0.12120703697204589
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,64,power_law_1.2,0.1266521644592285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,64,power_law_1.2,0.12757887840270996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,64,power_law_1.2,0.1300211238861084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,4,64,power_law_1.2,0.06727807998657227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,64,power_law_1.2,0.13066752433776857
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,64,power_law_1.2,0.1334348773956299
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,64,power_law_1.2,0.14576383590698244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,64,power_law_1.2,0.14006272315979004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,64,power_law_1.2,0.1409881591796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,64,power_law_1.2,0.15577983856201172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,64,power_law_1.2,0.14799872398376465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,64,power_law_1.2,0.1274841594696045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,64,power_law_1.2,0.17378816604614258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,64,power_law_1.2,0.19309568405151367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,64,power_law_1.2,0.22587392807006834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,64,power_law_1.2,0.30460159301757816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,64,power_law_1.2,0.36083072662353516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,64,power_law_1.2,0.5247513580322265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,64,power_law_1.2,0.7687628936767579
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,64,power_law_1.2,0.039504640102386475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,64,power_law_1.2,1.0126092529296875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,64,power_law_1.2,0.03913856029510498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,64,power_law_1.2,0.05052671909332276
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,64,power_law_1.2,0.04108799934387207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,64,power_law_1.2,1.4741069030761718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,64,power_law_1.2,0.04963456153869629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,64,power_law_1.2,0.050584321022033696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,64,power_law_1.2,0.05586048126220703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,64,power_law_1.2,0.052291841506958
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,64,power_law_1.2,0.05286655902862549
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,64,power_law_1.2,0.05801727771759033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,64,power_law_1.2,0.03991168022155762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,64,power_law_1.2,0.05469056129455566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,64,power_law_1.2,0.05501823902130128
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,64,power_law_1.2,0.060497918128967286
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,64,power_law_1.2,0.06706943988800049
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,64,power_law_1.2,0.06745471954345703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,64,power_law_1.2,0.06048255920410156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,64,power_law_1.2,0.07285120010375976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,64,power_law_1.2,0.08227840423583985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,64,power_law_1.2,0.09892736434936523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,64,power_law_1.2,0.16795135498046876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,64,power_law_1.2,0.10623616218566893
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,64,power_law_1.2,0.20620927810668946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,64,power_law_1.2,0.3235353469848633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,64,power_law_1.2,0.356890869140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,4,64,power_law_1.2,0.0223091197013855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,64,power_law_1.2,0.5494579315185547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,4,64,power_law_1.2,0.022305281162261964
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,4,64,power_law_1.2,0.022339839935302735
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,4,64,power_law_1.2,0.022348799705505372
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,4,64,power_law_1.2,0.022476799488067627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,64,power_law_1.2,0.7780134582519531
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,4,64,power_law_1.2,0.0215065598487854
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,4,64,power_law_1.2,0.02181504011154175
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,4,64,power_law_1.2,0.02267008066177368
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,4,64,power_law_1.2,0.022247679233551025
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,4,64,power_law_1.2,0.02194175958633423
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,4,64,power_law_1.2,0.0219865608215332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,4,64,power_law_1.2,0.02219520092010498
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,4,64,power_law_1.2,0.022242560386657714
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,4,64,power_law_1.2,0.02235136032104492
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,4,64,power_law_1.2,0.022650880813598635
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,4,64,power_law_1.2,0.023902719020843503
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,4,64,power_law_1.2,0.027159039974212644
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,4,64,power_law_1.2,0.021990399360656738
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,4,64,power_law_1.01,0.07706495761871338
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,4,64,power_law_1.01,0.06341504096984864
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,4,64,power_law_1.01,0.10446335792541503
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,4,64,power_law_1.01,0.13333120346069335
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,4,64,power_law_1.2,0.03916800022125244
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,4,64,power_law_1.2,0.034035201072692874
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,4,64,power_law_1.2,0.02801919937133789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,8,32,balanced,0.034202880859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,8,32,balanced,0.03372159957885742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,8,32,balanced,0.03340800046920776
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,8,32,balanced,0.032983040809631346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,8,32,balanced,0.03951744079589844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,8,32,balanced,0.06214655876159668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,8,32,balanced,0.07596928119659424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,8,32,balanced,0.12840831756591797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,8,32,balanced,0.13616767883300782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,8,32,balanced,0.12804863929748536
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,8,32,balanced,0.12821887969970702
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,8,32,balanced,0.13316351890563966
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,8,32,balanced,0.13263360023498535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,8,32,balanced,0.1377305603027344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,8,32,balanced,0.14270848274230957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,8,32,balanced,0.14403072357177735
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,4,64,power_law_1.2,0.0502348804473877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,8,32,balanced,0.14015744209289552
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,8,32,balanced,0.14851712226867675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,8,32,balanced,0.1653811264038086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,8,32,balanced,0.15411840438842775
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,8,32,balanced,0.18041215896606447
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,8,32,balanced,0.2099238395690918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,8,32,balanced,0.24587263107299803
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,8,32,balanced,0.3007321548461914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,8,32,balanced,0.3634342575073242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,8,32,balanced,0.48317951202392584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,8,32,balanced,0.03370879888534546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,8,32,balanced,0.033784320354461675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,8,32,balanced,0.6073164749145508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,8,32,balanced,0.03543551921844482
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,8,32,balanced,0.03746815919876099
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,8,32,balanced,0.04313983917236328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,8,32,balanced,0.04524928092956543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,8,32,balanced,0.06574079990386963
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,8,32,balanced,0.0583897590637207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,8,32,balanced,0.06614528179168702
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,8,32,balanced,0.0577561616897583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,8,32,balanced,0.05903488159179687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,8,32,balanced,0.05805439949035644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,8,32,balanced,0.062027521133422844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,8,32,balanced,0.033826560974121095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,8,32,balanced,0.06313727855682373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,8,32,balanced,0.06698239803314208
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,8,32,balanced,0.06518784046173096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,8,32,balanced,0.07233407974243164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,4,64,power_law_1.2,0.08204928398132325
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,4,64,power_law_1.2,0.06227968215942383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,8,32,balanced,0.07448575973510743
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,8,32,balanced,0.08492032051086426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,8,32,balanced,0.09219712257385254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,8,32,balanced,0.11094528198242189
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,8,32,balanced,0.12871808052062989
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,8,32,balanced,0.16394367218017578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,8,32,balanced,0.21320064544677733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,8,32,balanced,0.28248832702636717
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,4,64,power_law_1.2,0.10338687896728516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,8,32,balanced,0.3686502456665039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,4,64,power_law_1.2,0.14049920082092285
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,4,64,power_law_1.2,0.19539072036743163
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,8,32,balanced,0.020281600952148437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,8,32,balanced,0.021574399471282958
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,8,32,balanced,0.019939839839935303
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,8,32,balanced,0.025443840026855468
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,8,32,balanced,0.02598144054412842
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,8,32,balanced,0.026385920047760008
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,8,32,balanced,0.026746881008148194
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,8,32,balanced,0.02392319917678833
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,8,32,balanced,0.020071680545806887
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,8,32,balanced,0.02033535957336426
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,8,32,balanced,0.0275161600112915
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,8,32,balanced,0.027297279834747314
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,8,32,balanced,0.027294719219207765
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,8,32,balanced,0.02792831897735596
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,8,32,balanced,0.03033344030380249
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,8,32,balanced,0.0297216010093689
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,8,32,balanced,0.03172096014022827
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,8,32,balanced,0.03373823881149292
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,8,32,balanced,0.037406721115112306
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,8,32,balanced,0.026027519702911377
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,8,32,balanced,0.026371839046478274
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,8,32,balanced,0.02683903932571411
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,8,32,power_law_1.01,0.05916543960571289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,8,32,power_law_1.01,0.06956416130065918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,8,32,power_law_1.01,0.0702566385269165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,8,32,power_law_1.01,0.07300608158111573
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,8,32,power_law_1.01,0.07235583782196045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,8,32,power_law_1.01,0.1294924831390381
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,8,32,power_law_1.01,0.1253209590911865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,8,32,power_law_1.01,0.13353471755981444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,8,32,power_law_1.01,0.12779135704040528
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,8,32,power_law_1.01,0.12845567703247068
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,8,32,power_law_1.01,0.12852607727050783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,8,32,power_law_1.01,0.13478783607482908
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,8,32,power_law_1.01,0.13813119888305664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,8,32,power_law_1.01,0.1425766372680664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,8,32,power_law_1.01,0.1426150417327881
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,8,32,power_law_1.01,0.1439129638671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,8,32,power_law_1.01,0.1474944019317627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,8,32,power_law_1.01,0.15539711952209473
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,8,32,power_law_1.01,0.1741811180114746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,8,32,power_law_1.01,0.20248960494995116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,8,32,power_law_1.01,0.22206335067749022
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,8,32,power_law_1.01,0.2960576057434082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,8,32,power_law_1.01,0.31687551498413086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,8,32,power_law_1.01,0.47202304840087894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,8,32,power_law_1.01,0.5805184173583984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,8,32,power_law_1.01,0.730218276977539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,8,32,power_law_1.01,0.039548161029815676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,8,32,power_law_1.01,0.04333439826965332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,8,32,power_law_1.01,0.04365439891815186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,8,32,power_law_1.01,1.0811007690429686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,8,32,power_law_1.01,0.04477312088012696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,8,32,power_law_1.01,0.045701122283935545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,8,32,power_law_1.01,0.060037121772766114
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,8,32,power_law_1.01,0.05648767948150635
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,8,32,power_law_1.01,0.05713024139404297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,8,32,power_law_1.01,0.06827904224395752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,8,32,power_law_1.01,0.06079360008239746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,8,32,power_law_1.01,0.060803837776184075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,8,32,power_law_1.01,0.06337024211883545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,8,32,power_law_1.01,0.0708902406692505
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,8,32,power_law_1.01,0.06990464210510254
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,8,32,power_law_1.01,0.06816512107849121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,8,32,power_law_1.01,0.07291903972625732
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,8,32,power_law_1.01,0.05809408187866211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,8,32,power_law_1.01,0.0781990385055542
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,8,32,power_law_1.01,0.08929280281066895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,8,32,power_law_1.01,0.10257408142089844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,8,32,power_law_1.01,0.119550724029541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,8,32,power_law_1.01,0.15515135765075683
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,8,32,power_law_1.01,0.17905920028686523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,8,32,power_law_1.01,0.24626176834106445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,8,32,power_law_1.01,0.3351923370361328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,8,32,power_law_1.01,0.023889920711517333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,8,32,power_law_1.01,0.5034342575073242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,8,32,power_law_1.01,0.025488638877868654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,8,32,power_law_1.01,0.7007347106933594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,8,32,power_law_1.01,0.025125119686126705
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,8,32,power_law_1.01,0.025740799903869627
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,8,32,power_law_1.01,0.026295039653778073
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,8,32,power_law_1.01,0.02509952068328857
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,8,32,power_law_1.01,0.02600575923919678
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,8,32,power_law_1.01,0.027723519802093505
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,8,32,power_law_1.01,0.025996799468994143
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,8,32,power_law_1.01,0.026565120220184323
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,8,32,power_law_1.01,0.027962880134582518
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,8,32,power_law_1.01,0.026818559169769284
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,8,32,power_law_1.01,0.027951359748840332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,8,32,power_law_1.01,0.02682751893997192
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,8,32,power_law_1.01,0.028679680824279786
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,8,32,power_law_1.01,0.028428800106048584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,8,32,power_law_1.01,0.030565121173858643
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,8,32,power_law_1.01,0.02626944065093994
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,8,32,balanced,0.025199360847473144
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,8,32,balanced,0.026388480663299563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,8,32,balanced,0.05278336048126221
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,8,32,balanced,0.05890816211700439
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,8,32,balanced,0.06981247901916504
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,8,32,power_law_1.01,0.033831679821014406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,8,32,power_law_1.01,0.03615488052368164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,8,32,power_law_1.2,0.06065663814544677
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,8,32,power_law_1.01,0.04319744110107422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,8,32,power_law_1.2,0.06990975856781005
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,8,32,power_law_1.2,0.07020927906036377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,8,32,power_law_1.2,0.07334911823272705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,8,32,power_law_1.2,0.07197055816650391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,8,32,power_law_1.2,0.1274611186981201
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,8,32,power_law_1.2,0.12464256286621093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,8,32,power_law_1.2,0.13411456108093261
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,8,32,power_law_1.2,0.12758272171020507
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,8,32,power_law_1.01,0.05645567893981933
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,8,32,power_law_1.2,0.12944255828857423
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,8,32,power_law_1.2,0.12928895950317382
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,8,32,power_law_1.2,0.13328000068664553
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,8,32,power_law_1.2,0.1372159957885742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,8,32,power_law_1.2,0.14566911697387694
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,8,32,power_law_1.2,0.14210944175720214
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,8,32,power_law_1.2,0.1432652759552002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,8,32,power_law_1.2,0.14835328102111817
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,8,32,power_law_1.2,0.15753472328186036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,8,32,power_law_1.2,0.18051071166992189
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,8,32,power_law_1.2,0.21579776763916017
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,8,32,power_law_1.2,0.2621363258361816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,8,32,power_law_1.2,0.3248473739624024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,8,32,power_law_1.2,0.3757235336303711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,8,32,power_law_1.2,0.5291072082519531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,8,32,power_law_1.2,0.694250259399414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,8,32,power_law_1.2,1.1178931427001952
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,8,32,power_law_1.2,0.039468801021575926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,8,32,power_law_1.2,0.04378880023956298
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,8,32,power_law_1.2,0.04501503944396973
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,8,32,power_law_1.2,1.3873881530761718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,8,32,power_law_1.2,0.04464767932891846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,8,32,power_law_1.2,0.05917695999145508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,8,32,power_law_1.2,0.05553535938262939
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,8,32,power_law_1.2,0.05587327957153321
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,8,32,power_law_1.2,0.06653312206268311
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,8,32,power_law_1.2,0.058878722190856936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,8,32,power_law_1.2,0.058184962272644046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,8,32,power_law_1.2,0.06113024234771729
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,8,32,power_law_1.2,0.06219776153564453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,8,32,power_law_1.2,0.06509056091308593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,8,32,power_law_1.2,0.07218688011169434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,8,32,power_law_1.2,0.0715558385848999
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,8,32,power_law_1.2,0.07201791763305664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,8,32,power_law_1.2,0.08030719757080078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,8,32,power_law_1.2,0.07364992141723634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,8,32,power_law_1.2,0.09788543701171874
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,8,32,power_law_1.01,0.06660863876342774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,8,32,power_law_1.2,0.13645055770874023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,8,32,power_law_1.2,0.19123712539672852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,8,32,power_law_1.2,0.22838783264160156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,8,32,power_law_1.2,0.33845886230468747
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,8,32,power_law_1.2,0.44367488861083987
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,8,32,power_law_1.2,0.12172672271728516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,8,32,power_law_1.2,0.6351475143432618
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,8,32,power_law_1.2,0.025649919509887698
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,8,32,power_law_1.2,0.025614080429077146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,8,32,power_law_1.2,0.8612723541259765
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,8,32,power_law_1.2,0.025588479042053226
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,8,32,power_law_1.2,0.025656321048736573
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,8,32,power_law_1.2,0.026394879817962645
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,8,32,power_law_1.2,0.02542464017868042
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,8,32,power_law_1.2,0.02614527940750122
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,8,32,power_law_1.2,0.027613439559936524
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,8,32,power_law_1.2,0.026263039112091064
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,8,32,power_law_1.2,0.02359424114227295
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,8,32,power_law_1.2,0.026986238956451414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,8,32,power_law_1.2,0.026566400527954104
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,8,32,power_law_1.2,0.02721407890319824
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,8,32,power_law_1.2,0.02721407890319824
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,8,32,power_law_1.2,0.0281548810005188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,8,32,power_law_1.2,0.02759167909622192
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,8,32,power_law_1.2,0.02936320066452026
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,8,32,power_law_1.01,0.0880025577545166
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,8,32,power_law_1.01,0.11315967559814452
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,8,32,power_law_1.01,0.12577280044555664
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,8,32,power_law_1.01,0.18772863388061523
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,8,32,power_law_1.2,0.03108223915100098
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,8,32,power_law_1.2,0.045228800773620605
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,8,32,power_law_1.2,0.03601408004760742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,8,32,power_law_1.2,0.05556096076965332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,16,16,balanced,0.03172991991043091
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,16,16,balanced,0.03372544050216675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,16,16,balanced,0.03649408102035522
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,16,16,balanced,0.03894144058227539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,16,16,balanced,0.03529088020324707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,16,16,balanced,0.06754687786102295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,16,16,balanced,0.07956607818603516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,16,16,balanced,0.13808768272399902
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,16,16,balanced,0.13931520462036134
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,16,16,balanced,0.13936256408691405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,16,16,balanced,0.14586879730224608
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,16,16,balanced,0.15000191688537598
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,16,16,balanced,0.15300607681274414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,16,16,balanced,0.14563967704772948
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,16,16,balanced,0.15178751945495605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,16,16,balanced,0.1530406379699707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,16,16,balanced,0.15791359901428223
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,8,32,power_law_1.2,0.06884607791900635
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,16,16,balanced,0.17291263580322264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,16,16,balanced,0.18733312606811522
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,16,16,balanced,0.21389440536499021
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,16,16,balanced,0.24458879470825196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,16,16,balanced,0.3014707183837891
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,8,32,power_law_1.2,0.0770521593093872
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,8,32,power_law_1.2,0.11193344116210938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,16,16,balanced,0.36281471252441405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,16,16,balanced,0.4777344131469727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,16,16,balanced,0.03379199981689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,16,16,balanced,0.034114561080932616
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,16,16,balanced,0.589752311706543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,16,16,balanced,0.03741055965423584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,16,16,balanced,0.041579518318176266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,16,16,balanced,0.04443520069122314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,16,16,balanced,0.8180979156494141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,16,16,balanced,0.05645952224731445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,16,16,balanced,0.05716224193572998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,16,16,balanced,0.06779136180877685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,16,16,balanced,0.06869631767272949
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,16,16,balanced,0.06969088077545166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,16,16,balanced,0.07215231895446778
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,16,16,balanced,0.0762931203842163
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,16,16,balanced,1.0445977783203124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,16,16,balanced,0.07893375873565674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,16,16,balanced,0.07483391761779785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,16,16,balanced,0.07811071872711181
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,16,16,balanced,0.08132096290588378
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,16,16,balanced,0.08350336074829101
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,16,16,balanced,0.09175168037414551
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,16,16,balanced,0.10227456092834472
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,16,16,balanced,0.12063103675842286
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,16,16,balanced,0.1377228832244873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,16,16,balanced,0.1723302459716797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,16,16,balanced,0.20670848846435547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,16,16,balanced,0.27800703048706055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,16,16,balanced,0.366558723449707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,16,16,balanced,0.5002841567993164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,16,16,balanced,0.6557810974121094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,8,32,power_law_1.2,0.13967103958129884
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,8,32,power_law_1.2,0.26131071090698244
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,8,32,power_law_1.2,0.2891865539550781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,16,16,balanced,0.022480640411376953
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,16,16,balanced,0.022259199619293214
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,16,16,balanced,0.03617536067962647
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,16,16,balanced,0.03545344114303589
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,16,16,balanced,0.022648320198059083
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,16,16,balanced,0.037253119945526124
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,16,16,balanced,0.03765503883361816
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,16,16,balanced,0.03704063892364502
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,16,16,balanced,0.03741568088531494
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,16,16,balanced,0.03747328042984009
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,16,16,balanced,0.036856319904327396
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,16,16,balanced,0.038056960105896
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,16,16,balanced,0.039512319564819334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,16,16,balanced,0.03968127965927124
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,16,16,balanced,0.04249599933624267
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,16,16,balanced,0.04460288047790527
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,16,16,balanced,0.04817024230957031
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,16,16,balanced,0.05263872146606445
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,16,16,balanced,0.059182081222534176
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,16,16,balanced,0.02513279914855957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,16,16,power_law_1.01,0.035352320671081544
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,16,16,balanced,0.023568639755249025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,16,16,power_law_1.01,0.06100863933563232
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,16,16,power_law_1.01,0.07568384170532226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,16,16,power_law_1.01,0.07215231895446778
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,16,16,power_law_1.01,0.08189311981201172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,16,16,power_law_1.01,0.13538304328918457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,16,16,power_law_1.01,0.14812288284301758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,16,16,power_law_1.01,0.13393407821655273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,16,16,power_law_1.01,0.13295488357543944
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,16,16,power_law_1.01,0.13573887825012207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,16,16,power_law_1.01,0.14141056060791016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,16,16,power_law_1.01,0.14507007598876953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,16,16,power_law_1.01,0.14983296394348145
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,16,16,power_law_1.01,0.14454400062561035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,16,16,power_law_1.01,0.15159680366516112
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,16,16,power_law_1.01,0.15463552474975586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,16,16,power_law_1.01,0.16024703979492189
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,16,16,power_law_1.01,0.17931903839111327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,16,16,power_law_1.01,0.19758848190307615
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,16,16,power_law_1.01,0.23754112243652345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,16,16,power_law_1.01,0.2643788719177246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,16,16,power_law_1.01,0.3258393478393555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,16,16,power_law_1.01,0.405847053527832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,16,16,power_law_1.01,0.5359270477294922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,16,16,power_law_1.01,0.661185302734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,16,16,power_law_1.01,0.9960921478271484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,16,16,power_law_1.01,0.03908224105834961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,16,16,power_law_1.01,0.04609151840209961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,16,16,power_law_1.01,1.3365542602539062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,16,16,balanced,0.029671680927276612
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,16,16,power_law_1.01,0.05217152118682862
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,16,16,power_law_1.01,0.055828480720520025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,16,16,power_law_1.01,0.05505919933319092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,16,16,power_law_1.01,0.07427584171295167
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,16,16,power_law_1.01,0.06729599952697754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,16,16,power_law_1.01,0.0664409589767456
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,16,16,power_law_1.01,0.06860032081604003
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,16,16,power_law_1.01,0.06830719947814942
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,16,16,power_law_1.01,0.0758847999572754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,16,16,power_law_1.01,0.07901696205139161
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,16,16,power_law_1.01,0.08407296180725098
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,16,16,power_law_1.01,0.07722623825073242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,16,16,power_law_1.01,0.07894400119781494
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,16,16,power_law_1.01,0.08270336151123046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,16,16,power_law_1.01,0.08438912391662598
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,16,16,power_law_1.01,0.0950220775604248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,16,16,power_law_1.01,0.105862398147583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,16,16,power_law_1.01,0.1261964797973633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,16,16,power_law_1.01,0.15524224281311033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,16,16,power_law_1.01,0.19322240829467774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,16,16,power_law_1.01,0.25443967819213864
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,16,16,power_law_1.01,0.3390041732788086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,16,16,power_law_1.01,0.44960384368896483
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,16,16,power_law_1.01,0.02514559984207153
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,16,16,power_law_1.01,0.6031731033325196
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,16,16,power_law_1.01,0.02936959981918335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,16,16,power_law_1.01,0.7849625396728516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,16,16,power_law_1.01,0.03598464012145996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,16,16,power_law_1.01,0.03554687976837158
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,16,16,power_law_1.01,0.03556735992431641
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,16,16,power_law_1.01,0.03572864055633545
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,16,16,power_law_1.01,0.034766080379486086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,16,16,power_law_1.01,0.034895360469818115
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,16,16,power_law_1.01,0.03580672025680542
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,16,16,power_law_1.01,0.03462912082672119
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,16,16,power_law_1.01,0.03419519901275635
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,16,16,power_law_1.01,0.03458048105239868
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,16,16,power_law_1.01,0.03623807907104492
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,16,16,power_law_1.01,0.03631743907928467
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,16,16,power_law_1.01,0.03635328054428101
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,16,16,power_law_1.01,0.03699584007263183
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,16,16,power_law_1.01,0.03777663946151734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,16,16,power_law_1.01,0.04230912208557129
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,16,16,balanced,0.03566720008850098
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,16,16,balanced,0.036062719821929934
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,16,16,balanced,0.08802816390991211
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,16,16,balanced,0.07676415920257569
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,16,16,balanced,0.11326335906982421
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,16,16,power_law_1.01,0.05638144016265869
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,16,16,power_law_1.01,0.048787198066711425
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,16,16,power_law_1.01,0.06711040019989015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,16,16,power_law_1.2,0.03790719985961914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,16,16,power_law_1.2,0.06077184200286865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,16,16,power_law_1.2,0.0717299222946167
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,16,16,power_law_1.2,0.07588096141815186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,16,16,power_law_1.2,0.0764569616317749
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,16,16,power_law_1.2,0.14412927627563477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,16,16,power_law_1.2,0.13527039527893064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,16,16,power_law_1.2,0.13225472450256348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,16,16,power_law_1.2,0.13387519836425782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,16,16,power_law_1.2,0.13730048179626464
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,16,16,power_law_1.2,0.14257535934448243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,16,16,power_law_1.2,0.1463283157348633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,16,16,power_law_1.2,0.15263487815856933
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,16,16,power_law_1.2,0.14608384132385255
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,16,16,power_law_1.2,0.1513587188720703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,16,16,power_law_1.2,0.15629695892333983
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,16,16,power_law_1.2,0.15910783767700196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,16,16,power_law_1.2,0.18112127304077147
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,16,16,power_law_1.2,0.20565631866455075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,16,16,power_law_1.2,0.23533695220947265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,16,16,power_law_1.2,0.29373439788818356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,16,16,power_law_1.2,0.3741145706176758
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,16,16,power_law_1.01,0.08517375946044922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,16,16,power_law_1.2,0.40497665405273436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,16,16,power_law_1.2,0.598790397644043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,16,16,power_law_1.2,0.834588165283203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,16,16,power_law_1.01,0.10141568183898926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,16,16,power_law_1.2,0.039921920299530025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,16,16,power_law_1.2,1.237127685546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,16,16,power_law_1.2,0.046416640281677246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,16,16,power_law_1.2,0.05161215782165527
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,16,16,power_law_1.2,0.052056322097778315
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,16,16,power_law_1.2,0.06123007774353027
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,16,16,power_law_1.2,1.531988525390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,16,16,power_law_1.2,0.06794112205505372
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,16,16,power_law_1.2,0.06764287948608398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,16,16,power_law_1.2,0.06779136180877685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,16,16,power_law_1.2,0.07048831939697266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,16,16,power_law_1.2,0.07624703884124756
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,16,16,power_law_1.2,0.08354047775268555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,16,16,power_law_1.2,0.08188287734985351
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,16,16,power_law_1.2,0.07779712200164794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,16,16,power_law_1.2,0.08054911613464355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,16,16,power_law_1.2,0.08364800453186036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,16,16,power_law_1.2,0.08450176239013671
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,16,16,power_law_1.2,0.10948863983154297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,16,16,power_law_1.2,0.1348851203918457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,16,16,power_law_1.2,0.06574719905853271
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,16,16,power_law_1.2,0.17467136383056642
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,16,16,power_law_1.2,0.23899648666381834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,16,16,power_law_1.2,0.2959859275817871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,16,16,power_law_1.2,0.3842854309082031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,16,16,power_law_1.2,0.5448678588867187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,16,16,power_law_1.2,0.025542399883270263
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,16,16,power_law_1.2,0.8210393524169921
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,16,16,power_law_1.2,0.9364083099365235
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,16,16,power_law_1.2,0.034603519439697264
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,16,16,power_law_1.2,0.03571327924728394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,16,16,power_law_1.2,0.09755647659301758
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,16,16,power_law_1.2,0.03533951997756958
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,16,16,power_law_1.2,0.03638272047042847
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,16,16,power_law_1.2,0.0340070390701294
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,16,16,power_law_1.2,0.035608320236206054
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,16,16,power_law_1.2,0.028995840549468993
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,16,16,power_law_1.2,0.0337612795829773
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,16,16,power_law_1.2,0.0346777606010437
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,16,16,power_law_1.2,0.03512192010879517
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,16,16,power_law_1.2,0.03411200046539307
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,16,16,power_law_1.01,0.13007360458374023
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,16,16,power_law_1.2,0.035100159645080564
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,16,16,power_law_1.2,0.037402880191802976
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,16,16,power_law_1.2,0.036974079608917236
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,16,16,power_law_1.2,0.04093183994293213
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,16,16,power_law_1.01,0.16606464385986328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,16,16,power_law_1.01,0.22583168029785156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,16,16,power_law_1.2,0.03592704057693481
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,16,16,power_law_1.01,0.3083302307128906
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,16,16,power_law_1.2,0.047134718894958495
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,16,16,power_law_1.2,0.0531276798248291
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,16,16,power_law_1.2,0.09367551803588867
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,16,16,power_law_1.2,0.10267264366149902
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,16,16,power_law_1.2,0.06967936038970948
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,32,8,balanced,0.03662847995758057
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,32,8,balanced,0.03504256010055542
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,32,8,balanced,0.03741055965423584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,32,8,balanced,0.04240128040313721
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,32,8,balanced,0.04686079978942871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,32,8,balanced,0.07729663848876953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,32,8,balanced,0.09148799896240234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,32,8,balanced,0.15912704467773436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,32,8,balanced,0.16236543655395508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,32,8,balanced,0.1652454376220703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,32,8,balanced,0.16032512664794923
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,16,16,power_law_1.2,0.1483187198638916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,32,8,balanced,0.1657318305969238
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,32,8,balanced,0.1694553565979004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,32,8,balanced,0.17704959869384768
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,32,8,balanced,0.1799488067626953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,32,8,balanced,0.18686464309692383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,32,8,balanced,0.20025983810424805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,32,8,balanced,0.2286783981323242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,32,8,balanced,0.2577497673034668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,32,8,balanced,0.314019832611084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,32,8,balanced,0.3704076766967773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,32,8,balanced,0.4859545516967773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,32,8,balanced,0.5969561767578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,32,8,balanced,0.8190054321289063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,32,8,balanced,0.03506688117980957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,32,8,balanced,1.0402662658691406
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,16,16,power_law_1.2,0.1930905532836914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,32,8,balanced,0.03823744058609009
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,32,8,balanced,0.044906239509582516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,32,8,balanced,0.05266304016113281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,32,8,balanced,0.05787775993347168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,32,8,balanced,0.06912511825561524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,32,8,balanced,1.4908901977539064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,32,8,balanced,0.09408767700195313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,32,8,balanced,0.09536383628845216
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,32,8,balanced,0.09832575798034668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,32,8,balanced,0.09993215560913085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,32,8,balanced,0.10408831596374513
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,32,8,balanced,0.1096063995361328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,32,8,balanced,0.1097920036315918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,32,8,balanced,1.946197814941406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,32,8,balanced,0.1153651237487793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,32,8,balanced,0.12352512359619142
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,32,8,balanced,0.14180224418640136
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,32,8,balanced,0.15926783561706542
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,32,8,balanced,0.19527551651000977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,32,8,balanced,0.07749760150909424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,32,8,balanced,0.2278041648864746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,32,8,balanced,0.29793792724609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,32,8,balanced,0.09440896034240723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,32,8,balanced,0.36487167358398437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,32,8,balanced,0.5013068771362305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,32,8,balanced,0.6699314880371093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,32,8,balanced,0.9376089477539062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,32,8,balanced,1.2386124420166016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,16,16,power_law_1.2,0.24955135345458984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,16,16,power_law_1.2,0.3894233703613281
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,16,16,power_law_1.2,0.5227148818969727
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,32,8,balanced,0.026762239933013916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,32,8,balanced,0.02759552001953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,32,8,balanced,0.053455362319946284
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,32,8,balanced,0.0543232011795044
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,32,8,balanced,0.05425663948059082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,32,8,balanced,0.05422719955444336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,32,8,balanced,0.05478784084320069
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,32,8,balanced,0.028409600257873535
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,32,8,balanced,0.05511295795440674
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,32,8,balanced,0.05500544071197509
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,32,8,balanced,0.05700352191925049
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,32,8,balanced,0.030379519462585446
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,32,8,balanced,0.057055997848510745
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,32,8,balanced,0.059357438087463375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,32,8,balanced,0.06128128051757813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,32,8,balanced,0.06408576011657716
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,32,8,balanced,0.06680319786071778
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,32,8,balanced,0.07384448051452637
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,32,8,balanced,0.07857791900634765
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,32,8,balanced,0.09005696296691894
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,32,8,balanced,0.033898239135742185
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,32,8,power_law_1.01,0.03903743982315063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,32,8,power_law_1.01,0.03987200021743774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,32,8,power_law_1.01,0.06457087993621827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,32,8,power_law_1.01,0.07005951881408692
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,32,8,power_law_1.01,0.07861120223999023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,32,8,power_law_1.01,0.08588671684265137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,32,8,power_law_1.01,0.08667903900146484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,32,8,power_law_1.01,0.15446271896362304
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,32,8,power_law_1.01,0.15531904220581055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,32,8,power_law_1.01,0.1570406436920166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,32,8,power_law_1.01,0.16106880187988282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,32,8,power_law_1.01,0.16293632507324218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,32,8,power_law_1.01,0.16373504638671874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,32,8,power_law_1.01,0.1658892822265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,32,8,power_law_1.01,0.1712179183959961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,32,8,power_law_1.01,0.17517824172973634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,32,8,power_law_1.01,0.18606847763061524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,32,8,power_law_1.01,0.20755456924438476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,32,8,power_law_1.01,0.23555200576782226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,32,8,power_law_1.01,0.27745920181274414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,32,8,power_law_1.01,0.314259204864502
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,32,8,power_law_1.01,0.42729854583740234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,32,8,power_law_1.01,0.5002572631835938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,32,8,power_law_1.01,0.692959976196289
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,32,8,balanced,0.043755521774292
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,32,8,power_law_1.01,0.8685491180419922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,32,8,power_law_1.01,0.04108287811279297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,32,8,power_law_1.01,0.04799488067626953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,32,8,power_law_1.01,0.06011392116546631
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,32,8,power_law_1.01,0.0628172779083252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,32,8,power_law_1.01,1.5931149291992188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,32,8,power_law_1.01,0.06665215969085694
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,32,8,power_law_1.01,0.07304704189300537
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,32,8,power_law_1.01,0.07653247833251953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,32,8,power_law_1.01,1.26901123046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,32,8,power_law_1.01,0.09106176376342774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,32,8,power_law_1.01,0.0939072036743164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,32,8,power_law_1.01,0.09862272262573242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,32,8,power_law_1.01,0.09810943603515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,32,8,power_law_1.01,0.09934080123901368
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,32,8,power_law_1.01,0.0982425594329834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,32,8,power_law_1.01,0.10033151626586914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,32,8,power_law_1.01,0.10068223953247071
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,32,8,power_law_1.01,0.09206656455993653
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,32,8,power_law_1.01,0.11125887870788573
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,32,8,power_law_1.01,0.12390527725219727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,32,8,power_law_1.01,0.14138751983642578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,32,8,power_law_1.01,0.1737049674987793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,32,8,power_law_1.01,0.2104140853881836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,32,8,power_law_1.01,0.2707276725769043
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,32,8,power_law_1.01,0.33052032470703124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,32,8,power_law_1.01,0.45283329010009765
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,32,8,power_law_1.01,0.5579468917846679
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,32,8,power_law_1.01,0.030179839134216308
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,32,8,power_law_1.01,0.8345983886718751
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,32,8,power_law_1.01,0.03299328088760376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,32,8,power_law_1.01,0.04468095779418945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,32,8,power_law_1.01,1.097539825439453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,32,8,power_law_1.01,0.04652800083160401
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,32,8,power_law_1.01,0.05067903995513916
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,32,8,power_law_1.01,0.050068478584289554
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,32,8,power_law_1.01,0.053091840744018556
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,32,8,power_law_1.01,0.053509120941162114
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,32,8,power_law_1.01,0.04240128040313721
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,32,8,power_law_1.01,0.050887680053710936
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,32,8,power_law_1.01,0.05136256217956543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,32,8,power_law_1.01,0.052967681884765624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,32,8,power_law_1.01,0.05311103820800781
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,32,8,power_law_1.01,0.05491199970245362
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,32,8,power_law_1.01,0.056556801795959476
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,32,8,power_law_1.01,0.058617601394653315
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,32,8,power_law_1.01,0.06138112068176269
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,32,8,power_law_1.01,0.06650879859924316
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,32,8,balanced,0.05440767765045166
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,32,8,balanced,0.053858561515808104
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,32,8,balanced,0.12629504203796388
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,32,8,balanced,0.15272319793701172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,32,8,balanced,0.2028351974487305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,32,8,power_law_1.01,0.08899456024169922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,32,8,power_law_1.01,0.07977344036102295
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,32,8,power_law_1.01,0.10560768127441407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,32,8,power_law_1.2,0.038717439174652105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,32,8,power_law_1.2,0.03988480091094971
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,32,8,power_law_1.2,0.061903362274169926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,32,8,power_law_1.2,0.06949376106262208
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,32,8,power_law_1.2,0.0785420799255371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,32,8,power_law_1.2,0.08509311676025391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,32,8,power_law_1.2,0.08830975532531739
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,32,8,power_law_1.2,0.15544192314147948
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,32,8,power_law_1.2,0.15521535873413086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,32,8,power_law_1.2,0.15703935623168946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,32,8,power_law_1.2,0.16116479873657225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,32,8,power_law_1.2,0.16459264755249023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,32,8,power_law_1.2,0.15929344177246094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,32,8,power_law_1.2,0.16622976303100584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,32,8,power_law_1.2,0.17256832122802734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,32,8,power_law_1.2,0.17513088226318357
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,32,8,power_law_1.01,0.14424703598022462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,32,8,power_law_1.2,0.18472831726074218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,32,8,power_law_1.2,0.2129062461853027
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,32,8,power_law_1.2,0.23869184494018553
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,32,8,power_law_1.2,0.2854963111877441
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,32,8,power_law_1.2,0.3361663818359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,32,8,power_law_1.2,0.4527872085571289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,32,8,power_law_1.2,0.5148620986938477
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,32,8,power_law_1.01,0.18004735946655273
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,32,8,power_law_1.2,0.7836019134521484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,32,8,power_law_1.2,0.040838398933410645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,32,8,power_law_1.2,0.910645751953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,32,8,power_law_1.2,0.048083200454711914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,32,8,power_law_1.2,0.05425280094146728
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,32,8,power_law_1.2,0.06263679981231689
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,32,8,power_law_1.2,1.4147290039062501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,32,8,power_law_1.2,0.06886784076690675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,32,8,power_law_1.2,0.07593215942382812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,32,8,power_law_1.2,0.07478015899658204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,32,8,power_law_1.2,1.8478399658203124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,32,8,power_law_1.2,0.0926144027709961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,32,8,power_law_1.2,0.09117312431335449
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,32,8,power_law_1.2,0.09595264434814453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,32,8,power_law_1.2,0.09071359634399415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,32,8,power_law_1.2,0.0985536003112793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,32,8,power_law_1.2,0.09466239929199219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,32,8,power_law_1.2,0.09870079994201661
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,32,8,power_law_1.2,0.10208383560180664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,32,8,power_law_1.2,0.10082816123962401
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,32,8,power_law_1.2,0.11005439758300781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,32,8,power_law_1.2,0.12951807975769042
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,32,8,power_law_1.2,0.14185983657836915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,32,8,power_law_1.2,0.17758975982666017
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,32,8,power_law_1.2,0.20689920425415037
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,32,8,power_law_1.2,0.27446144104003906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,32,8,power_law_1.2,0.3505971145629883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,32,8,power_law_1.2,0.4849843215942383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,32,8,power_law_1.2,0.5816857528686523
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,32,8,power_law_1.2,0.030105600357055663
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,32,8,power_law_1.2,0.03337599992752075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,32,8,power_law_1.2,0.9095999908447265
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,32,8,power_law_1.2,0.03940736055374146
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,32,8,power_law_1.2,0.042466559410095216
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,32,8,power_law_1.2,0.04774911880493164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,32,8,power_law_1.2,1.1806169891357423
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,32,8,power_law_1.2,0.050888957977294924
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,32,8,power_law_1.2,0.05288576126098633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,32,8,power_law_1.01,0.2415551948547363
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,32,8,power_law_1.2,0.05287424087524414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,32,8,power_law_1.01,0.28281856536865235
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,32,8,power_law_1.2,0.05140992164611816
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,32,8,power_law_1.2,0.05065343856811524
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,32,8,power_law_1.2,0.05019008159637452
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,32,8,power_law_1.2,0.053112320899963386
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,32,8,power_law_1.2,0.054009599685668944
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,32,8,power_law_1.2,0.06319488048553466
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,32,8,power_law_1.2,0.06035840034484863
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,32,8,power_law_1.2,0.07677951812744141
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,32,8,power_law_1.2,0.08464896202087402
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,32,8,power_law_1.01,0.41687808990478514
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,32,8,power_law_1.2,0.20155519485473633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,32,8,power_law_1.2,0.05307392120361328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,32,8,power_law_1.2,0.0588595199584961
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,32,8,power_law_1.01,0.48290176391601564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,256,balanced,0.02436352014541626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,256,balanced,0.023343360424041747
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,256,balanced,0.02401024103164673
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,256,balanced,0.0237555193901062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,256,balanced,0.02396928071975708
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,256,balanced,0.024023039340972902
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,256,balanced,0.040375041961669925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,256,balanced,0.04138751983642578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,256,balanced,0.042803201675415035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,256,balanced,0.04313087940216064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,256,balanced,0.043600640296936034
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,256,balanced,0.04480512142181396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,256,balanced,0.04545407772064209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,256,balanced,0.04656256198883056
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,256,balanced,0.04721920013427734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,256,balanced,0.04726784229278565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,256,balanced,0.048508157730102544
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,256,balanced,0.04934656143188477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,256,balanced,0.05146495819091797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,256,balanced,0.05286911964416504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,256,balanced,0.05521408081054687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,256,balanced,0.06121600151062012
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,256,balanced,0.05811071872711182
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,256,balanced,0.06613247871398927
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,256,balanced,0.07475456237792968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,256,balanced,0.08672127723693848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,256,balanced,0.10380160331726074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,256,balanced,0.027999999523162844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,256,balanced,0.027674880027771
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,256,balanced,0.02797312021255493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,256,balanced,0.02772608041763306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,256,balanced,0.027783679962158202
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,256,balanced,0.028066558837890627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,256,balanced,0.03387648105621338
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,256,balanced,0.0341593599319458
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,256,balanced,0.03463680028915405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,256,balanced,0.0347763204574585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,256,balanced,0.034584319591522215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,256,balanced,0.035033600330352785
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,256,balanced,0.034595839977264405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,256,balanced,0.03496831893920899
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,256,balanced,0.03539583921432495
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,256,balanced,0.03540607929229737
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,256,balanced,0.03581696033477783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,256,balanced,0.03639935970306397
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,256,balanced,0.03705471992492676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,256,balanced,0.03789184093475342
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,32,8,power_law_1.2,0.11003392219543458
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,256,balanced,0.03996031999588013
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,256,balanced,0.042315521240234376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,256,balanced,0.03932800054550171
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,256,balanced,0.046069760322570805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,32,8,power_law_1.2,0.13463168144226073
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,32,8,power_law_1.2,0.1714521598815918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,256,balanced,0.04811391830444336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,256,balanced,0.054952960014343265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,256,balanced,0.06077695846557617
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,32,8,power_law_1.2,0.35887359619140624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,384,32,8,power_law_1.2,0.45107967376708985
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,384,32,8,power_law_1.2,0.6438130950927734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,384,32,8,power_law_1.2,1.030307846069336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,1,256,balanced,0.015379199981689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,1,256,balanced,0.0157695996761322
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,1,256,balanced,0.015738879442214963
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,1,256,balanced,0.015301120281219483
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,1,256,balanced,0.015568640232086182
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,1,256,balanced,0.015307519435882568
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,1,256,balanced,0.01574399948120117
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,1,256,balanced,0.015363839864730836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,1,256,balanced,0.015367679595947264
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,1,256,balanced,0.015348479747772217
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,1,256,balanced,0.015340800285339356
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,1,256,balanced,0.015705599784851074
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,1,256,balanced,0.015717120170593263
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,1,256,balanced,0.015715839862823485
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,1,256,balanced,0.015705599784851074
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,1,256,balanced,0.015773439407348634
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,1,256,balanced,0.015754239559173582
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,1,256,balanced,0.015776000022888183
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,1,256,balanced,0.015717120170593263
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,1,256,balanced,0.01561087965965271
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,1,256,balanced,0.015700479745864866
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,1,256,balanced,0.015691519975662233
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,1,256,balanced,0.01535104036331177
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,256,power_law_1.01,0.02595455884933472
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,256,power_law_1.01,0.04314367771148682
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,256,power_law_1.01,0.026757121086120605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,256,power_law_1.01,0.04495872020721435
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,256,power_law_1.01,0.04609663963317871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,256,power_law_1.01,0.026024959087371825
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,256,power_law_1.01,0.04726272106170654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,256,power_law_1.01,0.024309759140014646
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,256,power_law_1.01,0.048879361152648924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,256,power_law_1.01,0.051074562072753904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,256,power_law_1.01,0.04606719970703125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,256,power_law_1.01,0.052250881195068356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,256,power_law_1.01,0.05399680137634277
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,256,power_law_1.01,0.057896962165832525
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,256,power_law_1.01,0.06178048133850098
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,256,power_law_1.01,0.06346111774444581
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,256,power_law_1.01,0.050122241973876956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,256,power_law_1.01,0.06942336082458496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,256,power_law_1.01,0.09224960327148438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,256,power_law_1.01,0.10791808128356935
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,256,power_law_1.01,0.14034560203552246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,256,power_law_1.01,0.16657024383544922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,256,power_law_1.01,0.24007808685302737
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,256,power_law_1.01,0.3344780731201172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,256,power_law_1.01,0.047736320495605465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,256,power_law_1.01,0.481657600402832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,256,power_law_1.01,0.02810879945755005
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,256,power_law_1.01,0.02797823905944824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,256,power_law_1.01,0.6633523559570312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,256,power_law_1.01,0.02817280054092407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,256,power_law_1.01,0.03444479942321778
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,256,power_law_1.01,0.034942719936370846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,256,power_law_1.01,0.03495296001434326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,256,power_law_1.01,0.03494784116744995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,256,power_law_1.01,0.03550719976425171
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,256,power_law_1.01,0.03578752040863037
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,256,power_law_1.01,0.029678080081939694
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,256,power_law_1.01,0.03700992107391358
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,256,power_law_1.01,0.03707520008087158
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,256,power_law_1.01,0.038238720893859865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,256,power_law_1.01,0.03870464086532593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,256,power_law_1.01,0.04010496139526367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,256,power_law_1.01,0.042961921691894535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,256,power_law_1.01,0.04238463878631592
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,256,power_law_1.01,0.0476697587966919
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,256,power_law_1.01,0.053267202377319335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,256,power_law_1.01,0.06090367794036865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,256,power_law_1.01,0.07679872035980224
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,256,power_law_1.01,0.09479935646057129
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,256,power_law_1.01,0.14221440315246583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,256,power_law_1.01,0.18300416946411133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,256,power_law_1.01,0.03628416061401367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,256,power_law_1.01,0.2889177513122559
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,1,256,power_law_1.01,0.015125759840011596
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,1,256,power_law_1.01,0.01571455955505371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,256,power_law_1.01,0.3644441604614258
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,1,256,power_law_1.01,0.015731199979782104
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,1,256,power_law_1.01,0.015776000022888183
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,1,256,power_law_1.01,0.015356160402297975
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,1,256,power_law_1.01,0.015665919780731202
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,1,256,power_law_1.01,0.01578495979309082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,1,256,power_law_1.01,0.015500799417495728
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,1,256,power_law_1.01,0.01574399948120117
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,1,256,power_law_1.01,0.01572864055633545
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,1,256,power_law_1.01,0.01573632001876831
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,1,256,power_law_1.01,0.015669759511947632
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,1,256,power_law_1.01,0.016094720363616942
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,1,256,power_law_1.01,0.015768320560455323
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,1,256,power_law_1.01,0.016104960441589357
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,1,256,power_law_1.01,0.015793919563293457
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,1,256,power_law_1.01,0.01529728055000305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,1,256,balanced,0.016596479415893556
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,1,256,balanced,0.016560640335083008
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,1,256,balanced,0.018831360340118408
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,1,256,balanced,0.017420799732208253
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,1,256,power_law_1.01,0.01657855987548828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,1,256,power_law_1.01,0.017749760150909424
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,1,256,power_law_1.01,0.016543999910354615
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,1,256,power_law_1.2,0.024798719882965087
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,1,256,power_law_1.2,0.024751360416412356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,1,256,power_law_1.2,0.02472064018249512
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,1,256,power_law_1.2,0.04267136096954346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,1,256,power_law_1.2,0.04446080207824707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,1,256,power_law_1.2,0.04609920024871826
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,1,256,power_law_1.01,0.019040000438690186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,1,256,power_law_1.2,0.04601600170135498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,1,256,power_law_1.2,0.04807551860809326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,1,256,power_law_1.2,0.048124160766601566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,1,256,power_law_1.2,0.04978687763214111
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,1,256,power_law_1.2,0.05116672039031982
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,1,256,power_law_1.2,0.0515008020401001
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,1,256,power_law_1.2,0.04686592102050781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,1,256,power_law_1.2,0.054527997970581055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,1,256,power_law_1.2,0.060093441009521485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,1,256,power_law_1.2,0.06161151885986328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,1,256,power_law_1.2,0.0625100803375244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,1,256,power_law_1.2,0.06977280139923095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,1,256,power_law_1.2,0.09411711692810058
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,1,256,power_law_1.2,0.12150527954101562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,1,256,power_law_1.2,0.14729599952697753
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,1,256,power_law_1.2,0.20602624893188476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,1,256,power_law_1.2,0.2980991935729981
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,1,256,power_law_1.2,0.4438272094726562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,1,256,power_law_1.2,0.5856972885131836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,1,256,power_law_1.2,0.9948480224609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,1,256,power_law_1.2,0.02765183925628662
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,1,256,power_law_1.2,0.028015360832214353
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,1,256,power_law_1.2,1.3430642700195312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,1,256,power_law_1.2,0.027991039752960207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,1,256,power_law_1.2,0.0349401593208313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,1,256,power_law_1.2,0.03421695947647095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,1,256,power_law_1.2,0.03519871950149536
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,1,256,power_law_1.2,0.034639360904693606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,1,256,power_law_1.2,0.03568511962890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,1,256,power_law_1.2,0.035795199871063235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,1,256,power_law_1.2,0.035837440490722655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,1,256,power_law_1.2,0.03667200088500976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,1,256,power_law_1.2,0.03706624031066895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,1,256,power_law_1.2,0.037840640544891356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,1,256,power_law_1.2,0.03961344003677368
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,1,256,power_law_1.2,0.0410969591140747
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,1,256,power_law_1.2,0.04324863910675049
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,1,256,power_law_1.2,0.04110079765319824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,1,256,power_law_1.2,0.04771455764770508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,1,256,power_law_1.2,0.05363071918487548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,1,256,power_law_1.2,0.06866559982299805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,1,256,power_law_1.2,0.08369919776916504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,1,256,power_law_1.2,0.13205504417419434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,1,256,power_law_1.2,0.15561599731445314
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,1,256,power_law_1.2,0.2745446395874024
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,1,256,power_law_1.01,0.022283520698547363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,1,256,power_law_1.2,0.3676262283325195
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,1,256,power_law_1.01,0.025743360519409182
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,1,256,power_law_1.2,0.01529088020324707
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,1,256,power_law_1.2,0.015315200090408326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,1,256,power_law_1.2,0.5901516723632813
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,1,256,power_law_1.2,0.01513856053352356
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,1,256,power_law_1.01,0.03019392013549805
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,1,256,power_law_1.2,0.01574527978897095
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,1,256,power_law_1.01,0.03886080026626587
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,1,256,power_law_1.2,0.01537279963493347
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,1,256,power_law_1.2,0.015703040361404418
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,1,256,power_law_1.2,0.01570688009262085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,1,256,power_law_1.2,0.729381103515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,1,256,power_law_1.2,0.015360000133514403
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,1,256,power_law_1.2,0.01536128044128418
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,1,256,power_law_1.2,0.0156876802444458
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,1,256,power_law_1.2,0.01569408059120178
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,1,256,power_law_1.2,0.015882240533828734
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,1,256,power_law_1.2,0.01582592010498047
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,1,256,power_law_1.2,0.015918079614639282
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,1,256,power_law_1.2,0.015742720365524293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,1,256,power_law_1.2,0.015705599784851074
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,1,256,power_law_1.01,0.05074048042297363
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,1,256,power_law_1.01,0.0717964792251587
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,1,256,power_law_1.2,0.01651968002319336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,1,256,power_law_1.2,0.017957119941711425
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,1,256,power_law_1.2,0.019452159404754636
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,1,256,power_law_1.2,0.023873279094696047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,2,128,balanced,0.02514048099517822
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,2,128,balanced,0.023884799480438233
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,2,128,balanced,0.02470144033432007
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,2,128,balanced,0.02430847883224487
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,2,128,balanced,0.0248473596572876
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,1,256,power_law_1.2,0.01697535991668701
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,2,128,balanced,0.022714879512786865
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,1,256,power_law_1.2,0.028092160224914554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,2,128,balanced,0.03749504089355469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,2,128,balanced,0.039457280635833744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,2,128,balanced,0.03845247983932495
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,2,128,balanced,0.03987584114074707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,2,128,balanced,0.04070655822753906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,2,128,balanced,0.04111104011535645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,2,128,balanced,0.0376473593711853
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,2,128,balanced,0.04158207893371582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,2,128,balanced,0.04237696170806885
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,2,128,balanced,0.04279679775238037
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,2,128,balanced,0.04495359897613525
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,2,128,balanced,0.04627583980560303
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,2,128,balanced,0.04990208148956299
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,2,128,balanced,0.05321472167968751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,2,128,balanced,0.05017727851867676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,2,128,balanced,0.056418561935424806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,2,128,balanced,0.05974400043487549
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,2,128,balanced,0.06797823905944825
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,2,128,balanced,0.08778495788574218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,2,128,balanced,0.1115839958190918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,2,128,balanced,0.1329088020324707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,2,128,balanced,0.0271616005897522
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,2,128,balanced,0.025960960388183595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,2,128,balanced,0.02699903964996338
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,2,128,balanced,0.02560256004333496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,2,128,balanced,0.02632575988769531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,2,128,balanced,0.0320793604850769
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,2,128,balanced,0.032467200756073
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,2,128,balanced,0.032258560657501226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,2,128,balanced,0.03297791957855224
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,2,128,balanced,0.032505600452423095
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,2,128,balanced,0.033061120510101315
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,2,128,balanced,0.03260288000106811
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,2,128,balanced,0.03372159957885742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,2,128,balanced,0.03373183965682983
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,2,128,balanced,0.03416960000991821
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,2,128,balanced,0.03478784084320068
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,2,128,balanced,0.03543040037155151
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,2,128,balanced,0.0369868803024292
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,2,128,balanced,0.03944704055786133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,2,128,balanced,0.035770881175994876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,2,128,balanced,0.04001920223236084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,2,128,balanced,0.025982720851898195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,2,128,balanced,0.04323328018188476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,2,128,balanced,0.04844543933868408
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,2,128,balanced,0.05105152130126953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,2,128,balanced,0.06535808086395264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,2,128,balanced,0.0789132785797119
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,1,256,power_law_1.2,0.03468031883239746
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,1,256,power_law_1.2,0.0450278377532959
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,1,256,power_law_1.2,0.10246144294738771
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,1,256,power_law_1.2,0.06633855819702147
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,1,256,power_law_1.2,0.1435315227508545
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,2,128,balanced,0.015326720476150513
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,2,128,balanced,0.015818239450454713
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,2,128,balanced,0.015741440057754515
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,2,128,balanced,0.015257600545883179
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,2,128,balanced,0.015349760055541992
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,2,128,balanced,0.01574399948120117
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,2,128,balanced,0.015790079832077024
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,2,128,balanced,0.015756800174713134
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,2,128,balanced,0.01571455955505371
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,2,128,balanced,0.01568127989768982
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,2,128,balanced,0.015797760486602783
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,2,128,balanced,0.016087039709091186
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,2,128,balanced,0.0157913601398468
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,2,128,balanced,0.016180479526519777
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,2,128,balanced,0.016117759943008424
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,2,128,balanced,0.016136959791183472
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,2,128,balanced,0.016140799522399902
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,2,128,balanced,0.015394560098648072
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,2,128,balanced,0.015738879442214963
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,2,128,balanced,0.015292160511016844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,2,128,power_law_1.01,0.024739840030670167
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,2,128,power_law_1.01,0.02398848056793213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,2,128,power_law_1.01,0.023486719131469727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,2,128,power_law_1.01,0.02314239978790283
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,2,128,power_law_1.01,0.0378278398513794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,2,128,power_law_1.01,0.04025599956512451
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,2,128,power_law_1.01,0.04035071849822998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,2,128,power_law_1.01,0.041113600730896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,2,128,power_law_1.01,0.041510400772094724
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,2,128,power_law_1.01,0.043399682044982915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,2,128,power_law_1.01,0.044686079025268555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,2,128,power_law_1.01,0.045126399993896484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,2,128,power_law_1.01,0.047193598747253415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,2,128,power_law_1.01,0.050010881423950194
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,2,128,power_law_1.01,0.05300608158111573
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,2,128,power_law_1.01,0.0541260814666748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,2,128,power_law_1.01,0.055983362197875974
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,2,128,power_law_1.01,0.0634444808959961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,2,128,power_law_1.01,0.07007232189178467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,2,128,power_law_1.01,0.09535231590270996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,2,128,power_law_1.01,0.11084287643432618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,2,128,power_law_1.01,0.1468569564819336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,2,128,power_law_1.01,0.1765939140319824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,2,128,power_law_1.01,0.23181312561035156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,2,128,power_law_1.01,0.3250636672973633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,2,128,power_law_1.01,0.46831104278564456
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,2,128,power_law_1.01,0.6030643081665039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,2,128,balanced,0.015312639474868776
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,2,128,power_law_1.01,0.025985279083251954
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,2,128,balanced,0.015873279571533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,2,128,power_law_1.01,0.026419200897216798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,2,128,power_law_1.01,0.026343679428100585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,2,128,power_law_1.01,0.026431999206542968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,2,128,power_law_1.01,0.03220096111297607
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,2,128,power_law_1.01,0.03291520118713379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,2,128,power_law_1.01,0.03307264089584351
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,2,128,power_law_1.01,0.03352191925048828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,2,128,power_law_1.01,0.03373568058013916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,2,128,power_law_1.01,0.034213120937347415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,2,128,power_law_1.01,0.034961919784545895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,2,128,power_law_1.01,0.034220800399780274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,2,128,power_law_1.01,0.03576191902160645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,2,128,power_law_1.01,0.03659264087677002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,2,128,power_law_1.01,0.03801215887069702
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,2,128,power_law_1.01,0.04068480014801025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,2,128,power_law_1.01,0.03709952116012573
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,2,128,power_law_1.01,0.04117631912231445
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,2,128,balanced,0.0154367995262146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,2,128,power_law_1.01,0.0448960018157959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,2,128,power_law_1.01,0.05632512092590332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,2,128,power_law_1.01,0.06348544120788574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,2,128,power_law_1.01,0.07896192073822021
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,2,128,power_law_1.01,0.0984883213043213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,2,128,power_law_1.01,0.14460543632507322
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,2,128,power_law_1.01,0.1772096061706543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,2,128,power_law_1.01,0.23320959091186522
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,2,128,power_law_1.01,0.015708160400390626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,2,128,power_law_1.01,0.3258585739135742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,2,128,power_law_1.01,0.01539072036743164
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,2,128,power_law_1.01,0.015672320127487184
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,2,128,power_law_1.01,0.015719679594039918
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,2,128,power_law_1.01,0.015764479637145994
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,2,128,power_law_1.01,0.015768320560455323
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,2,128,power_law_1.01,0.01566208004951477
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,2,128,power_law_1.01,0.015756800174713134
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,2,128,power_law_1.01,0.015773439407348634
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,2,128,power_law_1.01,0.015750399827957152
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,2,128,power_law_1.01,0.0157695996761322
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,2,128,power_law_1.01,0.016095999479293823
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,2,128,power_law_1.01,0.016161279678344725
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,2,128,power_law_1.01,0.015864319801330566
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,2,128,power_law_1.01,0.015927040576934816
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,2,128,power_law_1.01,0.016135679483413694
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,2,128,power_law_1.01,0.015815680027008054
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,2,128,balanced,0.017429759502410887
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,2,128,balanced,0.01776128053665161
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,2,128,balanced,0.01947391986846924
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,2,128,balanced,0.021184000968933105
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,2,128,power_law_1.01,0.018227200508117675
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,2,128,power_law_1.01,0.017004799842834473
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,2,128,power_law_1.01,0.01730687975883484
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,2,128,power_law_1.2,0.02316416025161743
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,2,128,power_law_1.2,0.022714879512786865
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,2,128,power_law_1.01,0.019443199634552003
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,2,128,power_law_1.2,0.022704639434814454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,2,128,power_law_1.2,0.040203518867492676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,2,128,power_law_1.2,0.039792640209197996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,2,128,power_law_1.2,0.042650880813598635
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,2,128,power_law_1.2,0.041634559631347656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,2,128,power_law_1.2,0.04436351776123047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,2,128,power_law_1.2,0.04275968074798584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,2,128,power_law_1.2,0.047953920364379884
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,2,128,power_law_1.2,0.045780482292175295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,2,128,power_law_1.2,0.047248640060424806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,2,128,power_law_1.2,0.04969344139099121
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,2,128,power_law_1.01,0.03245568037033081
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,2,128,power_law_1.01,0.02150399923324585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,2,128,power_law_1.2,0.05279232025146484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,2,128,power_law_1.01,0.027006719112396237
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,2,128,power_law_1.2,0.05599487781524658
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,2,128,power_law_1.2,0.054963197708129886
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,2,128,power_law_1.2,0.07278592109680175
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,2,128,power_law_1.2,0.08588159561157227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,2,128,power_law_1.2,0.11051648139953614
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,2,128,power_law_1.01,0.037363200187683104
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,2,128,power_law_1.2,0.12729727745056152
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,2,128,power_law_1.2,0.1664806365966797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,2,128,power_law_1.2,0.23712896347045898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,2,128,power_law_1.2,0.3437094497680664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,2,128,power_law_1.2,0.026184959411621096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,2,128,power_law_1.2,0.5061081695556641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,2,128,power_law_1.2,0.026772480010986328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,2,128,power_law_1.2,0.05486080169677734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,2,128,power_law_1.2,0.026776320934295654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,2,128,power_law_1.2,0.8162175750732421
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,2,128,power_law_1.2,0.03212032079696655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,2,128,power_law_1.2,0.032486400604248046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,2,128,power_law_1.2,0.033136639595031735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,2,128,power_law_1.2,0.03273216009140015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,2,128,power_law_1.2,0.03337215900421143
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,2,128,power_law_1.2,1.0420403289794922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,2,128,power_law_1.2,0.034283521175384524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,2,128,power_law_1.2,0.03500416040420532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,2,128,power_law_1.2,0.03497215986251831
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,2,128,power_law_1.2,0.03624576091766357
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,2,128,power_law_1.2,0.03904383897781372
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,2,128,power_law_1.2,0.03904000043869019
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,2,128,power_law_1.2,0.03910144090652466
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,2,128,power_law_1.2,0.03914367914199829
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,2,128,power_law_1.2,0.04208127975463867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,2,128,power_law_1.2,0.052070398330688474
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,2,128,power_law_1.2,0.0632038402557373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,2,128,power_law_1.2,0.07679743766784668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,2,128,power_law_1.2,0.11264639854431154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,2,128,power_law_1.2,0.14410752296447754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,2,128,power_law_1.2,0.2217215919494629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,2,128,power_law_1.2,0.29440000534057614
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,2,128,power_law_1.2,0.0338048005104065
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,2,128,power_law_1.2,0.015754239559173582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,2,128,power_law_1.2,0.4268211364746094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,2,128,power_law_1.2,0.0153766405582428
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,2,128,power_law_1.2,0.015742720365524293
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,2,128,power_law_1.2,0.015731199979782104
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,2,128,power_law_1.2,0.015768320560455323
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,2,128,power_law_1.2,0.015701760053634644
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,2,128,power_law_1.2,0.015763200521469116
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,2,128,power_law_1.2,0.015750399827957152
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,2,128,power_law_1.2,0.7053183746337891
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,2,128,power_law_1.2,0.015703040361404418
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,2,128,power_law_1.2,0.015686399936676025
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,2,128,power_law_1.2,0.015873279571533203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,2,128,power_law_1.2,0.01609984040260315
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,2,128,power_law_1.2,0.015737600326538086
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,2,128,power_law_1.2,0.016097279787063597
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,2,128,power_law_1.2,0.01617535948753357
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,2,128,power_law_1.2,0.01578879952430725
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,2,128,power_law_1.01,0.05047935962677001
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,2,128,power_law_1.01,0.06419328212738037
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,2,128,power_law_1.2,0.016799999475479128
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,2,128,power_law_1.2,0.017749760150909424
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,2,128,power_law_1.2,0.019133440256118774
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,2,128,power_law_1.2,0.018172160387039185
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,4,64,balanced,0.019809279441833496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,4,64,balanced,0.01941375970840454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,4,64,balanced,0.019809279441833496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,4,64,balanced,0.01948415994644165
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,2,128,power_law_1.2,0.026910719871520994
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,4,64,balanced,0.021868801116943358
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,2,128,power_law_1.2,0.021916160583496096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,4,64,balanced,0.022955520153045656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,4,64,balanced,0.037406721115112306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,4,64,balanced,0.03912575960159302
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,4,64,balanced,0.03910784006118774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,4,64,balanced,0.03930880069732666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,4,64,balanced,0.042746877670288085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,4,64,balanced,0.041141757965087886
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,4,64,balanced,0.041118721961975094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,4,64,balanced,0.047306241989135744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,4,64,balanced,0.04283135890960693
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,4,64,balanced,0.04381311893463134
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,4,64,balanced,0.045432319641113286
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,4,64,balanced,0.049327359199523926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,4,64,balanced,0.047424001693725584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,4,64,balanced,0.05270527839660645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,4,64,balanced,0.05613696098327636
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,4,64,balanced,0.06394495964050292
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,4,64,balanced,0.0758732795715332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,4,64,balanced,0.10096256256103515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,4,64,balanced,0.11890048027038573
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,4,64,balanced,0.15256447792053224
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,4,64,balanced,0.024006400108337402
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,4,64,balanced,0.1874777603149414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,4,64,balanced,0.02478208065032959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,4,64,balanced,0.024389119148254396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,4,64,balanced,0.0247705602645874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,4,64,balanced,0.025585920810699464
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,4,64,balanced,0.02639616012573242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,4,64,balanced,0.0313971209526062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,4,64,balanced,0.03217664003372193
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,4,64,balanced,0.0315827202796936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,4,64,balanced,0.03172096014022827
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,4,64,balanced,0.03429888010025024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,4,64,balanced,0.03216768026351929
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,4,64,balanced,0.032624640464782716
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,4,64,balanced,0.03864448070526123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,4,64,balanced,0.03331072092056274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,4,64,balanced,0.03457279920578003
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,4,64,balanced,0.03575680017471313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,4,64,balanced,0.03784703969955444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,4,64,balanced,0.035000319480896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,4,64,balanced,0.038616321086883544
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,4,64,balanced,0.04021376132965088
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,4,64,balanced,0.04363135814666748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,4,64,balanced,0.04656511783599854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,4,64,balanced,0.06182144165039063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,4,64,balanced,0.07235455989837647
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,4,64,balanced,0.09443967819213868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,4,64,balanced,0.11815423965454101
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,2,128,power_law_1.2,0.03132800102233887
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,2,128,power_law_1.2,0.051824641227722165
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,2,128,power_law_1.2,0.053287677764892585
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,2,128,power_law_1.2,0.08128255844116211
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,2,128,power_law_1.2,0.12238719940185547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,4,64,balanced,0.015912959575653078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,4,64,balanced,0.016591360569000242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,4,64,balanced,0.016098560094833375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,4,64,balanced,0.016413439512252808
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,4,64,balanced,0.015690239667892455
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,4,64,balanced,0.016567039489746093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,4,64,balanced,0.016615680456161498
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,4,64,balanced,0.01657088041305542
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,4,64,balanced,0.016576000452041627
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,4,64,balanced,0.016718080043792723
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,4,64,balanced,0.016590080261230468
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,4,64,balanced,0.01655295968055725
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,4,64,balanced,0.016913919448852538
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,4,64,balanced,0.016981760263442995
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,4,64,balanced,0.017337599992752077
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,4,64,balanced,0.017034239768981933
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,4,64,balanced,0.01734655976295471
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,4,64,balanced,0.017798399925231932
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,4,64,balanced,0.016136959791183472
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,4,64,balanced,0.01610751986503601
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,4,64,power_law_1.01,0.023892478942871092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,4,64,power_law_1.01,0.02269695997238159
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,4,64,power_law_1.01,0.02317568063735962
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,4,64,power_law_1.01,0.021862399578094483
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,4,64,power_law_1.01,0.03576575994491577
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,4,64,power_law_1.01,0.03839231967926025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,4,64,power_law_1.01,0.03928191900253296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,4,64,power_law_1.01,0.040468478202819826
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,4,64,power_law_1.01,0.04141952037811279
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,4,64,power_law_1.01,0.042315521240234376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,4,64,power_law_1.01,0.04371071815490722
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,4,64,power_law_1.01,0.04519680023193359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,4,64,power_law_1.01,0.046449918746948246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,4,64,power_law_1.01,0.04971903800964356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,4,64,power_law_1.01,0.05180799961090088
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,4,64,power_law_1.01,0.05181695938110351
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,4,64,power_law_1.01,0.058296318054199214
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,4,64,power_law_1.01,0.07010047912597657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,4,64,power_law_1.01,0.07449600219726563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,4,64,power_law_1.01,0.09603839874267578
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,4,64,balanced,0.016573439836502075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,4,64,power_law_1.01,0.10612223625183106
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,4,64,power_law_1.01,0.13758591651916502
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,4,64,power_law_1.01,0.1672127914428711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,4,64,power_law_1.01,0.22607744216918943
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,4,64,power_law_1.01,0.28585472106933596
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,4,64,balanced,0.016669440269470214
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,4,64,balanced,0.016510720252990725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,4,64,power_law_1.01,0.026433279514312746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,4,64,power_law_1.01,0.4920345687866211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,4,64,power_law_1.01,0.0255948805809021
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,4,64,power_law_1.01,0.0257587194442749
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,4,64,power_law_1.01,0.7494630432128906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,4,64,power_law_1.01,0.02597759962081909
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,4,64,power_law_1.01,0.031343359947204594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,4,64,power_law_1.01,0.031746559143066407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,4,64,power_law_1.01,0.032119040489196775
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,4,64,power_law_1.01,0.03311487913131714
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,4,64,power_law_1.01,0.03330048084259033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,4,64,power_law_1.01,0.03347840070724487
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,4,64,power_law_1.01,0.0345472002029419
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,4,64,power_law_1.01,0.03527551889419556
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,4,64,power_law_1.01,0.037423360347747806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,4,64,power_law_1.01,0.04068607807159424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,4,64,power_law_1.01,0.0374399995803833
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,4,64,power_law_1.01,0.03864703893661499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,4,64,power_law_1.01,0.04155007839202881
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,4,64,power_law_1.01,0.045633277893066405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,4,64,power_law_1.01,0.05626239776611328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,4,64,power_law_1.01,0.068437762260437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,4,64,power_law_1.01,0.08674943923950196
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,4,64,power_law_1.01,0.03291520118713379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,4,64,power_law_1.01,0.11183615684509278
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,4,64,power_law_1.01,0.14166015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,4,64,power_law_1.01,0.19582080841064453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,4,64,power_law_1.01,0.016559360027313234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,4,64,power_law_1.01,0.30436351776123044
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,4,64,power_law_1.01,0.01657855987548828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,4,64,power_law_1.01,0.40509311676025395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,4,64,power_law_1.01,0.01650048017501831
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,4,64,power_law_1.01,0.016524800062179566
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,4,64,power_law_1.01,0.016584960222244264
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,4,64,power_law_1.01,0.0165721595287323
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,4,64,power_law_1.01,0.016844799518585203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,4,64,power_law_1.01,0.01655679941177368
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,4,64,power_law_1.01,0.01650431990623474
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,4,64,power_law_1.01,0.016522239446640014
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,4,64,power_law_1.01,0.016921600103378294
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,4,64,power_law_1.01,0.017224960327148438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,4,64,power_law_1.01,0.017344000339508055
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,4,64,power_law_1.01,0.016962560415267943
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,4,64,power_law_1.01,0.017089279890060424
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,4,64,power_law_1.01,0.016524800062179566
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,4,64,balanced,0.02028287887573242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,4,64,balanced,0.019953919649124147
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,4,64,power_law_1.01,0.01692800045013428
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,4,64,balanced,0.022254080772399903
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,4,64,balanced,0.0246451210975647
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,4,64,power_law_1.01,0.018163199424743655
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,4,64,power_law_1.01,0.01863168001174927
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,4,64,power_law_1.01,0.019878400564193727
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,4,64,power_law_1.01,0.021077759265899658
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,4,64,power_law_1.2,0.02358016014099121
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,4,64,power_law_1.01,0.028622078895568847
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,4,64,power_law_1.2,0.021575679779052732
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,4,64,power_law_1.2,0.023065600395202637
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,4,64,power_law_1.2,0.03543423891067505
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,4,64,power_law_1.2,0.037052159309387205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,4,64,power_law_1.2,0.040249600410461425
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,4,64,power_law_1.2,0.04032639980316162
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,4,64,power_law_1.2,0.04119296073913574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,4,64,power_law_1.2,0.0417190408706665
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,4,64,power_law_1.01,0.02512383937835693
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,4,64,power_law_1.2,0.04269184112548828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,4,64,power_law_1.2,0.04491263866424561
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,4,64,power_law_1.2,0.04507904052734375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,4,64,power_law_1.01,0.03684223890304565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,4,64,power_law_1.2,0.052645120620727535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,4,64,power_law_1.2,0.051361279487609865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,4,64,power_law_1.2,0.05176959991455078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,4,64,power_law_1.2,0.06661632061004638
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,4,64,power_law_1.2,0.07912191867828369
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,4,64,power_law_1.2,0.08982272148132324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,4,64,power_law_1.2,0.10917887687683106
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,4,64,power_law_1.2,0.1156287956237793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,4,64,power_law_1.2,0.1656447982788086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,4,64,power_law_1.2,0.20947967529296876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,4,64,power_law_1.2,0.3340198516845703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,4,64,power_law_1.2,0.04852608203887939
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,4,64,power_law_1.2,0.41838718414306636
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,4,64,power_law_1.2,0.02523391962051392
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,4,64,power_law_1.2,0.026003201007843018
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,4,64,power_law_1.2,0.025986559391021728
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,4,64,power_law_1.2,0.6250624084472657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,4,64,power_law_1.2,0.02595455884933472
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,4,64,power_law_1.2,0.7462924957275391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,4,64,power_law_1.2,0.031307520866394045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,4,64,power_law_1.2,0.03252480030059814
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,4,64,power_law_1.2,0.03216511964797973
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,4,64,power_law_1.2,0.033085439205169674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,4,64,power_law_1.2,0.033786880970001223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,4,64,power_law_1.2,0.03453952074050903
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,4,64,power_law_1.2,0.03430783987045288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,4,64,power_law_1.2,0.03618688106536865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,4,64,power_law_1.2,0.03911936044692993
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,4,64,power_law_1.2,0.03802623987197876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,4,64,power_law_1.2,0.03783807992935181
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,4,64,power_law_1.2,0.03989759922027588
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,4,64,power_law_1.2,0.04818048000335694
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,4,64,power_law_1.2,0.05417471885681152
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,4,64,power_law_1.2,0.06811903953552247
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,4,64,power_law_1.2,0.07740032196044921
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,4,64,power_law_1.2,0.1097599983215332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,4,64,power_law_1.2,0.14147583961486815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,4,64,power_law_1.2,0.2122060775756836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,4,64,power_law_1.2,0.31343616485595704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,4,64,power_law_1.2,0.41887359619140624
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,4,64,power_law_1.2,0.016625920534133913
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,4,64,power_law_1.2,0.5400870513916016
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,4,64,power_law_1.2,0.016579840183258057
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,4,64,power_law_1.2,0.016520960330963137
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,4,64,power_law_1.2,0.016162559986114503
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,4,64,power_law_1.2,0.016346880197525025
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,4,64,power_law_1.2,0.016520960330963137
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,4,64,power_law_1.2,0.016596479415893556
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,4,64,power_law_1.2,0.033255679607391356
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,4,64,power_law_1.2,0.016579840183258057
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,4,64,power_law_1.01,0.04457471847534179
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,4,64,power_law_1.2,0.01658112049102783
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,4,64,power_law_1.2,0.016582399606704712
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,4,64,power_law_1.2,0.016574720144271853
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,4,64,power_law_1.2,0.01701632022857666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,4,64,power_law_1.2,0.01651455998420715
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,4,64,power_law_1.2,0.016743680238723753
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,4,64,power_law_1.2,0.017224960327148438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,4,64,power_law_1.2,0.018585599660873413
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,4,64,power_law_1.2,0.01693439960479736
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,4,64,power_law_1.01,0.0658188819885254
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,4,64,power_law_1.2,0.05297023773193359
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,4,64,power_law_1.01,0.08228351593017577
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,4,64,power_law_1.2,0.02190079927444458
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,4,64,power_law_1.2,0.023911681175231934
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,4,64,power_law_1.2,0.03128576040267944
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,8,32,balanced,0.02021503925323486
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,4,64,power_law_1.2,0.019481600522994996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,8,32,balanced,0.020670719146728515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,8,32,balanced,0.019854079484939575
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,8,32,balanced,0.02105855941772461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,8,32,balanced,0.02189440011978149
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,8,32,balanced,0.023957760334014894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,8,32,balanced,0.03962752103805542
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,8,32,balanced,0.03953536033630371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,8,32,balanced,0.04100607872009278
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,8,32,balanced,0.04113152027130127
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,8,32,balanced,0.04275199890136719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,8,32,balanced,0.042675199508666994
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,8,32,balanced,0.04362880229949952
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,8,32,balanced,0.04523647785186767
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,8,32,balanced,0.04780928134918213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,8,32,balanced,0.04968063831329346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,8,32,balanced,0.0464844799041748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,8,32,balanced,0.05050111770629882
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,8,32,balanced,0.05443583965301514
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,4,64,power_law_1.2,0.03896064043045044
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,4,64,power_law_1.2,0.01783552050590515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,8,32,balanced,0.06162176132202148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,8,32,balanced,0.06852992057800293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,8,32,balanced,0.09757056236267089
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,8,32,balanced,0.1131545639038086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,8,32,balanced,0.14315008163452148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,8,32,balanced,0.1753049659729004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,8,32,balanced,0.02473344087600708
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,8,32,balanced,0.23673599243164062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,8,32,balanced,0.02477695941925049
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,8,32,balanced,0.2996505546569824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,8,32,balanced,0.02481663942337036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,8,32,balanced,0.02556159973144531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,8,32,balanced,0.024783360958099365
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,8,32,balanced,0.02806528091430664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,8,32,balanced,0.03253632068634033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,8,32,balanced,0.033323519229888916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,8,32,balanced,0.03668479919433594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,8,32,balanced,0.034149119853973384
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,8,32,balanced,0.034639360904693606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,8,32,balanced,0.034556159973144526
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,8,32,balanced,0.036601600646972654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,8,32,balanced,0.0388428807258606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,8,32,balanced,0.04075263977050782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,8,32,balanced,0.036550400257110593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,8,32,balanced,0.03957887887954712
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,8,32,balanced,0.04239359855651855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,8,32,balanced,0.04364927768707275
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,8,32,balanced,0.04781824111938476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,8,32,balanced,0.058684158325195315
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,8,32,balanced,0.06885119915008545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,8,32,balanced,0.09305983543395996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,8,32,balanced,0.11372032165527343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,8,32,balanced,0.15790719985961915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,8,32,balanced,0.032698879241943354
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,8,32,balanced,0.20105344772338868
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,4,64,power_law_1.2,0.0650816011428833
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,4,64,power_law_1.2,0.08570752143859864
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,4,64,power_law_1.2,0.1164787197113037
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,8,32,balanced,0.016559360027313234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,8,32,balanced,0.016783360242843628
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,8,32,balanced,0.018214399814605715
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,8,32,balanced,0.018213119506835938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,8,32,balanced,0.018152960538864137
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,8,32,balanced,0.017341439723968503
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,8,32,balanced,0.018305280208587647
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,8,32,balanced,0.01823232054710388
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,8,32,balanced,0.01814784049987793
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,8,32,balanced,0.018612480163574217
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,8,32,balanced,0.018607360124588013
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,8,32,balanced,0.018647040128707888
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,8,32,balanced,0.01896448016166687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,8,32,balanced,0.018972159624099733
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,8,32,balanced,0.016947200298309324
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,8,32,balanced,0.016510720252990725
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,8,32,balanced,0.02024832010269165
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,8,32,balanced,0.020385279655456542
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,8,32,balanced,0.018170880079269407
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,8,32,balanced,0.019461120367050173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,8,32,power_law_1.01,0.022753279209136962
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,8,32,power_law_1.01,0.023605759143829348
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,8,32,power_law_1.01,0.023827199935913083
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,8,32,power_law_1.01,0.02395263910293579
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,8,32,power_law_1.01,0.037465600967407225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,8,32,power_law_1.01,0.03952896118164063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,8,32,power_law_1.01,0.04074111938476562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,8,32,power_law_1.01,0.040940799713134766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,8,32,power_law_1.01,0.04273791790008545
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,8,32,power_law_1.01,0.04295680046081543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,8,32,power_law_1.01,0.044441599845886234
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,8,32,power_law_1.01,0.04559743881225586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,8,32,power_law_1.01,0.04844672203063965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,8,32,power_law_1.01,0.051333122253417976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,8,32,power_law_1.01,0.04876543998718262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,8,32,power_law_1.01,0.05115647792816162
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,8,32,power_law_1.01,0.054918398857116704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,8,32,power_law_1.01,0.0735091209411621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,8,32,power_law_1.01,0.08594047546386718
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,8,32,balanced,0.018173439502716066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,8,32,power_law_1.01,0.09833984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,8,32,power_law_1.01,0.11446271896362306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,8,32,power_law_1.01,0.16015615463256835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,8,32,power_law_1.01,0.19576192855834962
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,8,32,power_law_1.01,0.2565107154846191
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,8,32,balanced,0.01815935969352722
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,8,32,power_law_1.01,0.316309757232666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,8,32,power_law_1.01,0.02678656101226807
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,8,32,power_law_1.01,0.4746521759033203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,8,32,power_law_1.01,0.026776320934295654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,8,32,power_law_1.01,0.026785280704498295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,8,32,power_law_1.01,0.6720896148681641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,8,32,power_law_1.01,0.027321600914001466
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,8,32,power_law_1.01,0.032147200107574464
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,8,32,power_law_1.01,0.03256704092025757
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,8,32,power_law_1.01,0.032924160957336426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,8,32,power_law_1.01,0.03379199981689453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,8,32,power_law_1.01,0.03411328077316284
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,8,32,power_law_1.01,0.034603519439697264
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,8,32,power_law_1.01,0.03625855922698974
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,8,32,power_law_1.01,0.038279678821563724
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,8,32,power_law_1.01,0.039481599330902103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,8,32,power_law_1.01,0.039504640102386475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,8,32,power_law_1.01,0.03971199989318848
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,8,32,power_law_1.01,0.03992448091506958
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,8,32,power_law_1.01,0.04195072174072266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,8,32,power_law_1.01,0.04402815818786621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,8,32,power_law_1.01,0.05345664024353027
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,8,32,power_law_1.01,0.06041600227355957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,8,32,power_law_1.01,0.07569536209106445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,8,32,power_law_1.01,0.11451007843017577
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,8,32,power_law_1.01,0.1254271984100342
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,8,32,power_law_1.01,0.16615167617797852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,8,32,power_law_1.01,0.20970495223999022
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,8,32,power_law_1.01,0.3059071922302246
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,8,32,power_law_1.01,0.01822975993156433
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,8,32,power_law_1.01,0.4483673477172852
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,8,32,power_law_1.01,0.018202879428863526
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,8,32,power_law_1.01,0.01816447973251343
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,8,32,power_law_1.01,0.018201600313186645
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,8,32,power_law_1.01,0.018141440153121947
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,8,32,power_law_1.01,0.017830400466918944
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,8,32,power_law_1.01,0.018156800270080566
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,8,32,power_law_1.01,0.018228479623794556
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,8,32,power_law_1.01,0.018216960430145264
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,8,32,power_law_1.01,0.018188799619674685
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,8,32,power_law_1.01,0.01819391965866089
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,8,32,power_law_1.01,0.01862912058830261
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,8,32,power_law_1.01,0.01864959955215454
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,8,32,power_law_1.01,0.018685439825057985
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,8,32,power_law_1.01,0.01894271969795227
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,8,32,power_law_1.01,0.01904512047767639
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,8,32,balanced,0.018263039588928224
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,8,32,power_law_1.01,0.018539520502090453
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,8,32,balanced,0.030086400508880617
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,8,32,balanced,0.024378879070281984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,8,32,balanced,0.03481215953826904
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,8,32,balanced,0.02371583938598633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,8,32,power_law_1.01,0.02061311960220337
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,8,32,power_law_1.01,0.01987712025642395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,8,32,power_law_1.01,0.022714879512786865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,8,32,power_law_1.2,0.022353920936584473
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,8,32,power_law_1.01,0.023938560485839845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,8,32,power_law_1.2,0.02315648078918457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,8,32,power_law_1.2,0.024303359985351564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,8,32,power_law_1.2,0.023560960292816163
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,8,32,power_law_1.2,0.03860863924026489
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,8,32,power_law_1.2,0.03953023910522461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,8,32,power_law_1.2,0.04076288223266601
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,8,32,power_law_1.2,0.042383360862731936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,8,32,power_law_1.2,0.04312575817108154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,8,32,power_law_1.2,0.04356863975524902
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,8,32,power_law_1.2,0.04596479892730713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,8,32,power_law_1.2,0.05066368103027343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,8,32,power_law_1.2,0.04811391830444336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,8,32,power_law_1.2,0.050937600135803215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,8,32,power_law_1.2,0.052692480087280266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,8,32,power_law_1.2,0.057473278045654295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,8,32,power_law_1.2,0.08192768096923828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,8,32,power_law_1.2,0.047720961570739746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,8,32,power_law_1.2,0.09340415954589844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,8,32,power_law_1.2,0.11179007530212401
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,8,32,power_law_1.2,0.13460991859436036
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,8,32,power_law_1.2,0.18403583526611328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,8,32,power_law_1.2,0.20656896591186524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,8,32,power_law_1.2,0.3288076782226562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,8,32,power_law_1.2,0.44173950195312506
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,8,32,power_law_1.01,0.02861696004867554
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,8,32,power_law_1.01,0.03593600034713745
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,8,32,power_law_1.2,0.6417318725585938
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,8,32,power_law_1.01,0.04838784217834473
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,8,32,power_law_1.2,0.02678656101226807
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,8,32,power_law_1.01,0.056732158660888675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,8,32,power_law_1.2,0.02645247936248779
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,8,32,power_law_1.2,0.026859519481658933
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,8,32,power_law_1.2,0.02758271932601929
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,8,32,power_law_1.2,0.03213696002960205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,8,32,power_law_1.2,0.8711206054687499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,8,32,power_law_1.2,0.03289472103118897
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,8,32,power_law_1.2,0.033780479431152345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,8,32,power_law_1.2,0.034115839004516604
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,8,32,power_law_1.2,0.036830720901489256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,8,32,power_law_1.2,0.037840640544891356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,8,32,power_law_1.2,0.039874560832977295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,8,32,power_law_1.2,0.04114304065704345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,8,32,power_law_1.2,0.03828736066818238
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,8,32,power_law_1.2,0.03992063999176025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,8,32,power_law_1.2,0.04403327941894532
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,8,32,power_law_1.2,0.05627903938293457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,8,32,power_law_1.2,0.0680448007583618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,8,32,power_law_1.2,0.08670080184936524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,8,32,power_law_1.2,0.11642368316650391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,8,32,power_law_1.2,0.15822976112365722
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,8,32,power_law_1.2,0.22757375717163084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,8,32,power_law_1.2,0.033783040046691894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,8,32,power_law_1.2,0.3268185424804687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,8,32,power_law_1.2,0.018145279884338377
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,8,32,power_law_1.2,0.018297599554061888
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,8,32,power_law_1.2,0.01823359966278076
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,8,32,power_law_1.2,0.018201600313186645
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,8,32,power_law_1.2,0.628223991394043
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,8,32,power_law_1.2,0.017840640544891356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,8,32,power_law_1.2,0.03480959892272949
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,8,32,power_law_1.2,0.01822975993156433
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,8,32,power_law_1.2,0.018216960430145264
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,8,32,power_law_1.2,0.018154239654541014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,8,32,power_law_1.2,0.48573055267333987
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,8,32,power_law_1.2,0.018187520503997804
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,8,32,power_law_1.2,0.018572800159454346
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,8,32,power_law_1.2,0.018581759929656983
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,8,32,power_law_1.2,0.017382400035858152
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,8,32,power_law_1.2,0.018602240085601806
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,8,32,power_law_1.2,0.018999040126800537
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,8,32,power_law_1.2,0.018999040126800537
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,8,32,power_law_1.2,0.019111679792404173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,8,32,power_law_1.2,0.05138944149017334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,8,32,power_law_1.01,0.07084415912628174
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,8,32,power_law_1.01,0.08491135597229003
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,8,32,power_law_1.2,0.019587839841842654
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,8,32,power_law_1.2,0.025599999427795412
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,8,32,power_law_1.2,0.02030080080032349
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,8,32,power_law_1.2,0.021794559955596923
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,16,16,balanced,0.01982079982757568
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,16,16,balanced,0.019544320106506346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,16,16,balanced,0.02145024061203003
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,16,16,balanced,0.021880319118499754
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,8,32,power_law_1.2,0.02927103996276855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,16,16,balanced,0.023962879180908205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,16,16,balanced,0.030513920783996583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,16,16,balanced,0.044372482299804686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,16,16,balanced,0.04978432178497315
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,16,16,balanced,0.04568960189819336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,16,16,balanced,0.04566527843475342
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,16,16,balanced,0.04771584033966064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,16,16,balanced,0.04970240116119385
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,8,32,power_law_1.2,0.03879168033599854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,16,16,balanced,0.050993919372558594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,16,16,balanced,0.04772992134094238
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,16,16,balanced,0.0500761604309082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,16,16,balanced,0.05201407909393311
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,16,16,balanced,0.05515903949737548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,16,16,balanced,0.06387072086334229
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,16,16,balanced,0.07221248149871826
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,16,16,balanced,0.09458944320678711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,16,16,balanced,0.11207807540893555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,16,16,balanced,0.14099072456359862
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,16,16,balanced,0.17175424575805665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,16,16,balanced,0.23032703399658203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,16,16,balanced,0.29139455795288083
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,16,16,balanced,0.024399359226226807
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,16,16,balanced,0.40754943847656244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,16,16,balanced,0.024309759140014646
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,16,16,balanced,0.025194239616394044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,16,16,balanced,0.5239513778686524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,16,16,balanced,0.025523200035095218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,16,16,balanced,0.02801408052444458
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,16,16,balanced,0.03134848117828369
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,16,16,balanced,0.038329598903656004
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,16,16,balanced,0.043612160682678223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,16,16,balanced,0.03912064075469971
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,16,16,balanced,0.03905280113220215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,16,16,balanced,0.040352001190185546
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,16,16,balanced,0.041991682052612306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,16,16,balanced,0.04279168128967285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,16,16,balanced,0.03853823900222778
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,16,16,balanced,0.04068480014801025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,16,16,balanced,0.04274303913116455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,16,16,balanced,0.043956480026245116
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,16,16,balanced,0.045980157852172854
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,16,16,balanced,0.049308161735534664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,16,16,balanced,0.0591641616821289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,16,16,balanced,0.070731520652771
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,16,16,balanced,0.09055744171142578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,16,16,balanced,0.10863871574401855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,16,16,balanced,0.1566335964202881
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,16,16,balanced,0.19201919555664063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,16,16,balanced,0.27587968826293946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,16,16,balanced,0.35833984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,8,32,power_law_1.2,0.044951038360595705
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,8,32,power_law_1.2,0.06149759769439698
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,8,32,power_law_1.2,0.07753856182098388
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,8,32,power_law_1.2,0.10810367584228515
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,8,32,power_law_1.2,0.16219776153564452
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,16,16,balanced,0.018100479841232302
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,16,16,balanced,0.021466879844665526
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,16,16,balanced,0.018559999465942383
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,16,16,balanced,0.017849600315093993
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,16,16,balanced,0.021583359241485595
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,16,16,balanced,0.02148736000061035
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,16,16,balanced,0.021848320960998535
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,16,16,balanced,0.02167936086654663
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,16,16,balanced,0.022164480686187742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,16,16,balanced,0.02231679916381836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,16,16,balanced,0.022704639434814454
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,16,16,balanced,0.022993919849395753
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,16,16,balanced,0.021982719898223875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,16,16,balanced,0.02389631986618042
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,16,16,balanced,0.02430464029312134
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,16,16,balanced,0.02554879903793335
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,16,16,balanced,0.02665343999862671
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,16,16,balanced,0.018644479513168336
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,16,16,balanced,0.019791359901428225
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,16,16,balanced,0.021470720767974856
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,16,16,power_law_1.01,0.02436352014541626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,16,16,power_law_1.01,0.024305920600891113
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,16,16,balanced,0.021760001182556152
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,16,16,power_law_1.01,0.02476032018661499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,16,16,power_law_1.01,0.025946879386901857
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,16,16,power_law_1.01,0.02632575988769531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,16,16,power_law_1.01,0.04076672077178955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,16,16,power_law_1.01,0.04383232116699219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,16,16,power_law_1.01,0.04502528190612793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,16,16,power_law_1.01,0.0460595178604126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,16,16,power_law_1.01,0.04673535823822021
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,16,16,power_law_1.01,0.04974463939666748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,16,16,power_law_1.01,0.051589121818542485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,16,16,power_law_1.01,0.051360001564025884
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,16,16,power_law_1.01,0.05050879955291748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,16,16,power_law_1.01,0.053471999168396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,16,16,power_law_1.01,0.055253758430480956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,16,16,power_law_1.01,0.06735360145568847
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,16,16,power_law_1.01,0.0838374423980713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,16,16,power_law_1.01,0.09762432098388671
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,16,16,power_law_1.01,0.11715840339660644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,16,16,power_law_1.01,0.13154047966003418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,16,16,power_law_1.01,0.19159807205200194
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,16,16,power_law_1.01,0.22307071685791016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,16,16,power_law_1.01,0.29157888412475585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,16,16,power_law_1.01,0.39382911682128907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,16,16,power_law_1.01,0.5981875228881836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,16,16,power_law_1.01,0.026208000183105467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,16,16,power_law_1.01,0.7801407623291016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,16,16,power_law_1.01,0.02973184108734131
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,16,16,power_law_1.01,0.02968832015991211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,16,16,power_law_1.01,0.030159358978271482
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,16,16,power_law_1.01,0.03127295970916748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,16,16,power_law_1.01,0.03720448017120361
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,16,16,power_law_1.01,0.038242559432983395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,16,16,power_law_1.01,0.03877248048782349
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,16,16,power_law_1.01,0.03914495944976807
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,16,16,power_law_1.01,0.04073855876922607
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,16,16,power_law_1.01,0.04151552200317383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,16,16,power_law_1.01,0.04350080013275147
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,16,16,power_law_1.01,0.04081024169921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,16,16,power_law_1.01,0.04114560127258301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,16,16,power_law_1.01,0.04281983852386474
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,16,16,power_law_1.01,0.04361472129821777
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,16,16,power_law_1.01,0.04853759765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,16,16,power_law_1.01,0.05539455890655518
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,16,16,balanced,0.03543168067932129
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,16,16,balanced,0.037911040782928465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,16,16,power_law_1.01,0.06205952167510986
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,16,16,balanced,0.0619814395904541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,16,16,power_law_1.01,0.07248511791229248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,16,16,power_law_1.01,0.12675456047058103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,16,16,power_law_1.01,0.08732159614562988
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,16,16,power_law_1.01,0.14286335945129394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,16,16,power_law_1.01,0.205743350982666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,16,16,power_law_1.01,0.2606591987609863
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,16,16,power_law_1.01,0.01989248037338257
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,16,16,power_law_1.01,0.021446399688720703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,16,16,power_law_1.01,0.35215488433837894
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,16,16,power_law_1.01,0.021712639331817628
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,16,16,power_law_1.01,0.4518988800048828
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,16,16,power_law_1.01,0.021457920074462893
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,16,16,power_law_1.01,0.021249279975891114
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,16,16,power_law_1.01,0.021049599647521972
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,16,16,power_law_1.01,0.021518080234527587
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,16,16,power_law_1.01,0.02144000053405762
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,16,16,power_law_1.01,0.020665600299835205
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,16,16,power_law_1.01,0.021062400341033936
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,16,16,power_law_1.01,0.02153856039047241
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,16,16,balanced,0.05092991828918457
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,16,16,power_law_1.01,0.022232320308685303
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,16,16,power_law_1.01,0.022295041084289553
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,16,16,power_law_1.01,0.022495999336242675
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,16,16,power_law_1.01,0.023086080551147463
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,16,16,power_law_1.01,0.02725503921508789
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,16,16,power_law_1.01,0.03709183931350708
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,16,16,power_law_1.01,0.030234880447387695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,16,16,power_law_1.01,0.02489727973937988
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,16,16,power_law_1.01,0.04813568115234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,16,16,power_law_1.01,0.0779750394821167
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,16,16,power_law_1.01,0.05819007873535156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,16,16,power_law_1.2,0.022342400550842283
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,16,16,power_law_1.2,0.023976960182189942
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,16,16,power_law_1.2,0.024447999000549316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,16,16,power_law_1.2,0.025998079776763917
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,16,16,power_law_1.2,0.026819839477539065
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,16,16,power_law_1.01,0.1008627223968506
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,16,16,power_law_1.2,0.04250879764556885
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,16,16,power_law_1.2,0.04316544055938721
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,16,16,power_law_1.2,0.044526081085205074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,16,16,power_law_1.2,0.04642560005187989
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,16,16,power_law_1.2,0.04970111846923828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,16,16,power_law_1.2,0.04711808204650879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,16,16,power_law_1.2,0.052227840423583985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,16,16,power_law_1.2,0.050968317985534674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,16,16,power_law_1.2,0.050922241210937504
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,16,16,power_law_1.2,0.05314688205718994
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,16,16,power_law_1.2,0.05488255977630615
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,16,16,power_law_1.2,0.07495039939880371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,16,16,power_law_1.2,0.08592639923095703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,16,16,power_law_1.2,0.09833855628967285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,16,16,power_law_1.2,0.12094847679138183
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,16,16,power_law_1.2,0.13898624420166014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,16,16,power_law_1.2,0.1804684829711914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,16,16,power_law_1.2,0.2197567939758301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,16,16,power_law_1.2,0.2946931266784668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,16,16,power_law_1.2,0.4708070373535156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,16,16,power_law_1.2,0.6896998596191406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,16,16,power_law_1.2,0.025959680080413817
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,16,16,power_law_1.2,0.9480230712890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,16,16,power_law_1.2,0.029885439872741698
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,16,16,power_law_1.2,0.029306879043579103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,16,16,power_law_1.2,0.03013887882232666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,16,16,power_law_1.2,0.03131392002105713
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,16,16,power_law_1.2,0.036692481040954586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,16,16,power_law_1.2,0.03871360063552857
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,16,16,power_law_1.2,0.03865983963012695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,16,16,power_law_1.2,0.039942400455474855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,16,16,power_law_1.2,0.04263040065765381
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,16,16,power_law_1.2,0.040677118301391604
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,16,16,power_law_1.2,0.04112256050109863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,16,16,power_law_1.2,0.042007040977478025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,16,16,power_law_1.2,0.04202367782592774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,16,16,power_law_1.2,0.04280960083007813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,16,16,power_law_1.2,0.047470078468322754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,16,16,power_law_1.2,0.04421631813049316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,16,16,power_law_1.2,0.05594240188598633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,16,16,power_law_1.2,0.06297215938568115
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,16,16,power_law_1.01,0.13700480461120607
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,16,16,power_law_1.01,0.15035776138305662
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,16,16,power_law_1.2,0.07284863948822021
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,16,16,power_law_1.2,0.12655232429504396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,16,16,power_law_1.2,0.15533696174621583
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,16,16,power_law_1.2,0.23388544082641602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,16,16,power_law_1.2,0.31882112503051757
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,16,16,power_law_1.2,0.466822395324707
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,3072,1536,8,256,16,16,power_law_1.2,0.019792640209198
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,3072,1536,8,256,16,16,power_law_1.2,0.021098239421844484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,3072,1536,8,256,16,16,power_law_1.2,0.020263679027557373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,16,16,power_law_1.2,0.5346611022949219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,3072,1536,8,256,16,16,power_law_1.2,0.021139199733734133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,16,16,power_law_1.2,0.09281536102294921
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,3072,1536,8,256,16,16,power_law_1.2,0.02067456007003784
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,3072,1536,8,256,16,16,power_law_1.2,0.020742399692535402
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,3072,1536,8,256,16,16,power_law_1.2,0.02067456007003784
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,3072,1536,8,256,16,16,power_law_1.2,0.021529600620269776
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,3072,1536,8,256,16,16,power_law_1.2,0.021500160694122313
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,3072,1536,8,256,16,16,power_law_1.2,0.021048319339752198
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,3072,1536,8,256,16,16,power_law_1.2,0.021877760887146
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,3072,1536,8,256,16,16,power_law_1.2,0.021913599967956544
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,3072,1536,8,256,16,16,power_law_1.2,0.022204160690307617
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,3072,1536,8,256,16,16,power_law_1.2,0.02274303913116455
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,3072,1536,8,256,16,16,power_law_1.2,0.021210880279541017
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,3072,1536,8,256,16,16,power_law_1.2,0.02882431983947754
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,3072,1536,8,256,16,16,power_law_1.2,0.022423040866851807
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,3072,1536,8,256,16,16,power_law_1.2,0.024272639751434327
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,3072,1536,8,256,16,16,power_law_1.2,0.025848319530487062
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,3072,1536,8,256,16,16,power_law_1.2,0.036677119731903074
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,3072,1536,8,256,16,16,power_law_1.2,0.043891201019287104
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,3072,1536,8,256,16,16,power_law_1.2,0.06302464008331299
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,3072,1536,8,256,16,16,power_law_1.2,0.12891008377075194
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,3072,1536,8,256,16,16,power_law_1.2,0.11284607887268065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,32,8,balanced,0.021454079151153566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,32,8,balanced,0.02356352090835571
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,32,8,balanced,0.025217280387878415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,32,8,balanced,0.028066558837890627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,32,8,balanced,0.03293312072753906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,32,8,balanced,0.05143807888031006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,32,8,balanced,0.0548415994644165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,32,8,balanced,0.02194943904876709
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,32,8,balanced,0.05672575950622558
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,32,8,balanced,0.050911998748779295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,32,8,balanced,0.05471871852874756
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,32,8,balanced,0.054662399291992195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,32,8,balanced,0.059594240188598636
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,32,8,balanced,0.052737278938293455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,32,8,balanced,0.06237567901611328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,32,8,balanced,0.06590847969055176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,32,8,balanced,0.07292416095733642
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,32,8,balanced,0.09936511993408204
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,3072,1536,8,256,16,16,power_law_1.2,0.24428415298461914
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,32,8,balanced,0.1153536033630371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,32,8,balanced,0.14363903999328614
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,32,8,balanced,0.1764838409423828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,32,8,balanced,0.23430015563964846
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,32,8,balanced,0.4061503982543945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,32,8,balanced,0.29262720108032225
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,3072,1536,8,256,16,16,power_law_1.2,0.18503679275512694
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,32,8,balanced,0.02558720111846924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,32,8,balanced,0.025955839157104494
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,32,8,balanced,0.5233638381958008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,32,8,balanced,0.028456959724426266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,32,8,balanced,0.03012480020523071
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,32,8,balanced,0.03959424018859863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,32,8,balanced,0.04607232093811035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,32,8,balanced,0.7469644927978516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,32,8,balanced,0.9696486663818359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,32,8,balanced,0.04937856197357178
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,3072,1536,8,256,16,16,power_law_1.2,0.07650815963745117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,32,8,balanced,0.05064576148986817
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,32,8,balanced,0.04608511924743652
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,32,8,balanced,0.047678718566894526
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,32,8,balanced,0.050992641448974606
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,32,8,balanced,0.05224319934844971
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,32,8,balanced,0.03566976070404053
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,32,8,balanced,0.05302656173706055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,32,8,balanced,0.056652798652648925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,32,8,balanced,0.06515456199645996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,32,8,balanced,0.07735167980194092
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,32,8,balanced,0.09569536209106445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,32,8,balanced,0.11630720138549804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,32,8,balanced,0.15658111572265626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,32,8,balanced,0.19301376342773438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,32,8,balanced,0.28208896636962894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,32,8,power_law_1.01,0.024143359661102294
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,32,8,balanced,0.5126323318481445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,32,8,balanced,0.6696755218505859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,32,8,power_law_1.01,0.025596160888671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,32,8,power_law_1.01,0.027996160984039307
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,32,8,power_law_1.01,0.030850560665130616
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,32,8,power_law_1.01,0.03240576028823853
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,32,8,power_law_1.01,0.05001728057861328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,32,8,power_law_1.01,0.050993919372558594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,32,8,power_law_1.01,0.05352831840515136
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,32,8,power_law_1.01,0.057055997848510745
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,32,8,power_law_1.01,0.053049597740173336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,32,8,power_law_1.01,0.05534719944000244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,32,8,power_law_1.01,0.05727615833282471
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,32,8,power_law_1.01,0.05407743930816651
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,32,8,power_law_1.01,0.06522880077362062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,32,8,power_law_1.01,0.06298495769500732
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,32,8,power_law_1.01,0.08237567901611328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,32,8,power_law_1.01,0.09568127632141113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,32,8,power_law_1.01,0.13422975540161133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,32,8,power_law_1.01,0.1578867244720459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,32,8,power_law_1.01,0.21030399322509769
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,32,8,power_law_1.01,0.2705075263977051
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,32,8,power_law_1.01,0.3693414306640625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,32,8,power_law_1.01,0.46834175109863285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,32,8,power_law_1.01,0.025544960498809815
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,32,8,balanced,0.046891517639160156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,32,8,power_law_1.01,0.02671871900558472
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,32,8,power_law_1.01,0.6673484802246094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,32,8,power_law_1.01,0.03170687913894653
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,32,8,power_law_1.01,0.0337779188156128
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,32,8,power_law_1.01,0.03678719997406006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,32,8,power_law_1.01,0.9055897521972657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,32,8,power_law_1.01,0.04280704021453857
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,32,8,power_law_1.01,0.04451712131500244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,32,8,power_law_1.01,0.04603775978088379
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,32,8,power_law_1.01,0.04653183937072754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,32,8,power_law_1.01,0.048427519798278806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,32,8,power_law_1.01,0.051813120841979976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,32,8,power_law_1.01,0.048958721160888674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,32,8,power_law_1.01,0.04884863853454589
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,32,8,power_law_1.01,0.04894207954406739
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,32,8,power_law_1.01,0.053059840202331544
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,32,8,power_law_1.01,0.050159358978271486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,32,8,power_law_1.01,0.057158398628234866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,32,8,power_law_1.01,0.06543488025665282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,32,8,power_law_1.01,0.07074048042297362
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,32,8,power_law_1.01,0.08846336364746094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,32,8,power_law_1.01,0.10320768356323243
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,32,8,power_law_1.01,0.14807935714721682
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,32,8,power_law_1.01,0.18608127593994142
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,32,8,power_law_1.01,0.24147327423095702
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,32,8,power_law_1.01,0.3191705513000488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,3072,1536,8,256,32,8,power_law_1.2,0.022776319980621337
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,32,8,power_law_1.01,0.10628095626831055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,3072,1536,8,256,32,8,power_law_1.2,0.024748799800872804
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,32,8,power_law_1.01,0.45691390991210934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,3072,1536,8,256,32,8,power_law_1.2,0.024775679111480715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,32,8,power_law_1.01,0.6245951843261719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,3072,1536,8,256,32,8,power_law_1.2,0.027201280593872067
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,3072,1536,8,256,32,8,power_law_1.2,0.03001600027084351
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,3072,1536,8,256,32,8,power_law_1.2,0.03825279951095581
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,3072,1536,8,256,32,8,power_law_1.2,0.049872641563415525
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,3072,1536,8,256,32,8,power_law_1.2,0.05149312019348144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,3072,1536,8,256,32,8,power_law_1.2,0.05257599830627442
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,3072,1536,8,256,32,8,power_law_1.2,0.053374719619750974
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,3072,1536,8,256,32,8,power_law_1.2,0.055955200195312506
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,3072,1536,8,256,32,8,power_law_1.2,0.05761151790618897
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,3072,1536,8,256,32,8,power_law_1.2,0.0575654411315918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,3072,1536,8,256,32,8,power_law_1.2,0.05504127979278565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,3072,1536,8,256,32,8,power_law_1.2,0.060259838104248044
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,32,8,balanced,0.04579967975616455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,3072,1536,8,256,32,8,power_law_1.2,0.06286719799041748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,3072,1536,8,256,32,8,power_law_1.2,0.07409791946411133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,3072,1536,8,256,32,8,power_law_1.2,0.09175040245056151
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,3072,1536,8,256,32,8,power_law_1.2,0.13195903778076173
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,3072,1536,8,256,32,8,power_law_1.2,0.216549129486084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,3072,1536,8,256,32,8,power_law_1.2,0.2868160057067871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,32,8,balanced,0.3529715347290039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,3072,1536,8,256,32,8,power_law_1.2,0.5042329788208008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,3072,1536,8,256,32,8,power_law_1.2,0.028784639835357666
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,3072,1536,8,256,32,8,power_law_1.2,0.6895410919189453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,3072,1536,8,256,32,8,power_law_1.2,0.11075327873229981
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,3072,1536,8,256,32,8,power_law_1.2,0.030936319828033448
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,32,8,power_law_1.2,0.031718399524688724
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,3072,1536,8,256,32,8,power_law_1.2,0.9404364776611327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,3072,1536,8,256,32,8,power_law_1.2,0.03414400100708008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,3072,1536,8,256,32,8,power_law_1.2,0.037171199321746826
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,3072,1536,8,256,32,8,power_law_1.2,0.39317886352539066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,3072,1536,8,256,32,8,power_law_1.2,0.04364160060882568
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,3072,1536,8,256,32,8,power_law_1.2,0.046033921241760256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,3072,1536,8,256,32,8,power_law_1.2,0.04729343891143799
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,3072,1536,8,256,32,8,power_law_1.2,0.1626367950439453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,3072,1536,8,256,32,8,power_law_1.2,0.047238402366638184
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,3072,1536,8,256,32,8,power_law_1.2,0.05139200210571289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,3072,1536,8,256,32,8,power_law_1.2,0.047642879486083985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,3072,1536,8,256,32,8,power_law_1.2,0.05115007877349853
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,3072,1536,8,256,32,8,power_law_1.2,0.05147007942199707
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,3072,1536,8,256,32,8,power_law_1.2,0.05627520084381103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,3072,1536,8,256,32,8,power_law_1.2,0.06271999835968017
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,3072,1536,8,256,32,8,power_law_1.2,0.0736575984954834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,3072,1536,8,256,32,8,power_law_1.2,0.09537152290344239
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,3072,1536,8,256,32,8,power_law_1.2,0.15555968284606933
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,3072,1536,8,256,32,8,power_law_1.2,0.20523647308349607
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,3072,1536,8,256,32,8,power_law_1.2,0.04525184154510498
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,3072,1536,8,256,32,8,power_law_1.2,0.26968448638916015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,3072,1536,8,256,32,8,power_law_1.2,0.04691711902618408
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,3072,1536,8,256,32,8,power_law_1.2,0.04681856155395508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,3072,1536,8,256,32,8,power_law_1.2,0.3516748809814453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,3072,1536,8,256,32,8,power_law_1.2,0.5394739151000977
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,3072,1536,8,256,32,8,power_law_1.2,0.6294015884399414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,128,balanced,0.022261760234832763
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,128,balanced,0.023946239948272705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,128,balanced,0.020721919536590576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,3072,1536,8,256,32,8,power_law_1.2,0.11176575660705565
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,128,balanced,0.02268671989440918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,128,balanced,0.03699455976486206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,128,balanced,0.037020161151885986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,128,balanced,0.03994879961013794
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,128,balanced,0.037423360347747806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,128,balanced,0.03967616081237793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,128,balanced,0.041103358268737795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,128,balanced,0.03614336013793945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,128,balanced,0.04084735870361328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,128,balanced,0.040759038925170896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,128,balanced,0.043979520797729495
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,128,balanced,0.04476160049438477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,128,balanced,0.022074880599975585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,128,balanced,0.04789247989654541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,128,balanced,0.044843521118164066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,128,balanced,0.05074816226959229
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,128,balanced,0.055671038627624514
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,128,balanced,0.05344895839691162
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,128,balanced,0.055724802017211916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,128,balanced,0.03907711982727051
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,3072,1536,8,256,32,8,power_law_1.01,0.03086335897445679
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,128,balanced,0.07236224174499511
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,128,balanced,0.08784255981445313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,128,balanced,0.06147200107574463
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,128,balanced,0.03671808004379272
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,2,128,balanced,0.016953599452972413
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,2,128,balanced,0.01661311984062195
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,2,128,balanced,0.01691648006439209
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,2,128,balanced,0.01694208025932312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,2,128,balanced,0.016270079612731934
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,2,128,balanced,0.016958719491958617
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,2,128,balanced,0.016611839532852175
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,2,128,balanced,0.016606719493865968
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,2,128,balanced,0.016528639793395996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,2,128,balanced,0.01654528021812439
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,2,128,balanced,0.0166975998878479
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,2,128,balanced,0.01694208025932312
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,2,128,balanced,0.01708415985107422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,2,128,balanced,0.016995840072631836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,2,128,balanced,0.016536320447921755
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,2,128,balanced,0.01697152018547058
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,2,128,balanced,0.016595200300216675
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,2,128,balanced,0.016549119949340822
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,2,128,balanced,0.01658112049102783
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,2,128,balanced,0.016531200408935548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,128,power_law_1.01,0.022359039783477783
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,2,128,balanced,0.016513279676437377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,128,power_law_1.01,0.021834239959716797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,128,power_law_1.01,0.03625344038009644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,128,power_law_1.01,0.037118721008300784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,128,power_law_1.01,0.03806848049163818
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,128,power_law_1.01,0.03911551952362061
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,128,power_law_1.01,0.039654400348663334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,128,power_law_1.01,0.03993855953216553
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,128,power_law_1.01,0.04108799934387207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,128,power_law_1.01,0.04193280220031738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,128,power_law_1.01,0.02110975980758667
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,128,power_law_1.01,0.04436351776123047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,128,power_law_1.01,0.0473689603805542
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,128,power_law_1.01,0.04966015815734863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,128,power_law_1.01,0.05272960186004638
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,128,power_law_1.01,0.054242558479309075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,128,power_law_1.01,0.0549235200881958
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,128,power_law_1.01,0.05876607894897461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,128,power_law_1.01,0.06937088012695312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,128,power_law_1.01,0.0886451244354248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,128,power_law_1.01,0.12204928398132324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,128,power_law_1.01,0.13805824279785156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,128,power_law_1.01,0.045631999969482424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,128,power_law_1.01,0.21347072601318356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,128,power_law_1.01,0.31860095977783204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,128,power_law_1.01,0.4561228942871094
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,2,128,power_law_1.01,0.016153600215911865
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,2,128,power_law_1.01,0.01650048017501831
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,128,power_law_1.01,0.6879462432861329
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,2,128,power_law_1.01,0.016168960332870484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,2,128,power_law_1.01,0.01650431990623474
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,2,128,power_law_1.01,0.016564480066299438
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,2,128,power_law_1.01,0.016780799627304076
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,2,128,power_law_1.01,0.016527359485626218
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,2,128,power_law_1.01,0.0165120005607605
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,2,128,power_law_1.01,0.016593919992446898
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,2,128,power_law_1.01,0.016720639467239378
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,2,128,power_law_1.01,0.01655168056488037
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,2,128,power_law_1.01,0.01661695957183838
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,2,128,power_law_1.01,0.016903680562973023
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,2,128,power_law_1.01,0.01692288041114807
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,2,128,power_law_1.01,0.016573439836502075
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,2,128,power_law_1.01,0.01696128010749817
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,2,128,balanced,0.01699072003364563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,2,128,balanced,0.01818112015724182
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,2,128,balanced,0.017831679582595825
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,2,128,balanced,0.01745408058166504
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,2,128,balanced,0.019569920301437376
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,2,128,balanced,0.020291841030120848
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,2,128,power_law_1.01,0.017354240417480467
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,2,128,power_law_1.01,0.017463040351867676
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,2,128,power_law_1.01,0.018173439502716066
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,2,128,power_law_1.01,0.019815679788589478
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,128,power_law_1.2,0.02103935956954956
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,2,128,power_law_1.01,0.018981120586395263
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,2,128,power_law_1.01,0.025935359001159668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,128,power_law_1.2,0.021073920726776125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,128,power_law_1.2,0.038638079166412355
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,2,128,power_law_1.01,0.029013760089874267
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,128,power_law_1.2,0.03906048059463501
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,2,128,power_law_1.01,0.03647871971130371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,128,power_law_1.2,0.039888639450073246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,128,power_law_1.2,0.0411571216583252
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,128,power_law_1.2,0.041914877891540525
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,128,power_law_1.2,0.044106240272521975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,128,power_law_1.2,0.043208961486816404
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,128,power_law_1.2,0.045377278327941896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,128,power_law_1.2,0.04690688133239746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,128,power_law_1.2,0.05099904060363769
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,128,power_law_1.2,0.05421696186065674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,128,power_law_1.2,0.05453184127807618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,128,power_law_1.2,0.05476096153259278
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,128,power_law_1.2,0.05855103969573975
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,128,power_law_1.2,0.07911295890808105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,128,power_law_1.2,0.09577856063842774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,128,power_law_1.2,0.12747391700744629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,128,power_law_1.2,0.1835750389099121
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,128,power_law_1.2,0.25475967407226563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,128,power_law_1.2,0.022283520698547363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,128,power_law_1.2,0.36197376251220703
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,2,128,power_law_1.2,0.01658751964569092
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,2,128,power_law_1.2,0.01651584029197693
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,2,128,power_law_1.2,0.016602879762649535
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,2,128,power_law_1.2,0.016523519754409792
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,2,128,power_law_1.2,0.016532479524612426
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,2,128,power_law_1.2,0.01652608036994934
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,128,power_law_1.2,0.03701888084411621
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,2,128,power_law_1.2,0.016595200300216675
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,2,128,power_law_1.2,0.01652992010116577
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,2,128,power_law_1.2,0.016591360569000242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,2,128,power_law_1.2,0.0169267201423645
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,2,128,power_law_1.2,0.016944639682769776
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,2,128,power_law_1.2,0.016592639684677123
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,2,128,power_law_1.2,0.016588799953460694
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,2,128,power_law_1.2,0.0170035195350647
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,2,128,power_law_1.2,0.01699712038040161
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,2,128,power_law_1.2,0.01695744037628174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,128,power_law_1.2,0.5979097747802735
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,2,128,power_law_1.2,0.01965824007987976
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,2,128,power_law_1.01,0.044926719665527345
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,2,128,power_law_1.01,0.0566323184967041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,128,power_law_1.2,0.8569561767578125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,128,power_law_1.2,1.0194188690185546
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,2,128,power_law_1.01,0.08002559661865234
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,2,128,power_law_1.2,0.017259520292282105
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,2,128,power_law_1.2,0.01783936023712158
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,2,128,power_law_1.2,0.020220160484313965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,4,64,balanced,0.019805439710617066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,4,64,balanced,0.01982848048210144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,4,64,balanced,0.019786239862442018
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,4,64,balanced,0.021222400665283202
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,4,64,balanced,0.03282560110092163
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,4,64,balanced,0.03414144039154053
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,4,64,balanced,0.03414144039154053
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,4,64,balanced,0.03560703992843628
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,4,64,balanced,0.03594752073287964
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,4,64,balanced,0.036252160072326664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,4,64,balanced,0.036350719928741455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,4,64,balanced,0.03707520008087158
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,4,64,balanced,0.03834111928939819
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,4,64,balanced,0.03872512102127075
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,2,128,power_law_1.2,0.03496831893920899
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,4,64,balanced,0.039950079917907715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,4,64,balanced,0.039493119716644286
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,4,64,balanced,0.04136320114135742
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,2,128,power_law_1.2,0.02499072074890137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,4,64,balanced,0.043655681610107425
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,4,64,balanced,0.04636544227600098
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,4,64,balanced,0.04960256099700928
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,4,64,balanced,0.05217919826507569
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,4,64,balanced,0.05092991828918457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,4,64,balanced,0.054384641647338874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,4,64,balanced,0.06288000106811523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,4,64,balanced,0.07482751846313476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,4,64,balanced,0.10168319702148439
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,4,64,balanced,0.12139904022216796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,4,64,power_law_1.01,0.02064768075942993
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,4,64,power_law_1.01,0.020682239532470705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,4,64,power_law_1.01,0.03337984085083008
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,2,128,power_law_1.2,0.04215680122375488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,4,64,power_law_1.01,0.03546623945236206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,4,64,power_law_1.01,0.034612479209899905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,4,64,power_law_1.01,0.037859840393066405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,4,64,power_law_1.01,0.0366758394241333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,4,64,power_law_1.01,0.03730815887451172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,4,64,power_law_1.01,0.04031231880187988
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,4,64,power_law_1.01,0.0399564790725708
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,4,64,power_law_1.01,0.041414399147033695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,4,64,power_law_1.01,0.02029439926147461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,4,64,power_law_1.01,0.04264448165893554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,4,64,power_law_1.01,0.04456960201263428
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,2,128,power_law_1.2,0.05625343799591065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,4,64,power_law_1.01,0.049724159240722654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,4,64,power_law_1.01,0.051813120841979976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,4,64,power_law_1.01,0.051457281112670894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,4,64,power_law_1.01,0.06883840084075928
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,4,64,power_law_1.01,0.0823846435546875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,4,64,power_law_1.01,0.08747008323669433
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,4,64,power_law_1.01,0.11420160293579103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,4,64,power_law_1.01,0.12880767822265624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,4,64,power_law_1.01,0.16573312759399414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,4,64,power_law_1.01,0.24505727767944335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,4,64,power_law_1.2,0.020257279872894288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,4,64,power_law_1.01,0.3884902572631836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,4,64,power_law_1.2,0.021584639549255373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,4,64,power_law_1.2,0.020193281173706053
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,4,64,power_law_1.01,0.04706175804138184
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,4,64,power_law_1.2,0.03376512050628662
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,4,64,power_law_1.2,0.03584000110626221
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,4,64,power_law_1.2,0.0353766393661499
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,4,64,power_law_1.2,0.03664000034332275
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,4,64,power_law_1.2,0.03677311897277832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,4,64,power_law_1.2,0.0398963189125061
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,4,64,power_law_1.2,0.03829504013061523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,4,64,power_law_1.2,0.041263360977172855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,4,64,power_law_1.2,0.044591360092163086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,4,64,power_law_1.2,0.047169280052185056
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,4,64,power_law_1.2,0.043278079032897945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,4,64,power_law_1.2,0.05107711791992188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,4,64,power_law_1.2,0.05038464069366455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,4,64,power_law_1.2,0.05368959903717041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,4,64,power_law_1.2,0.07277056217193603
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,4,64,power_law_1.2,0.08947711944580078
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,2,128,power_law_1.2,0.06925951957702636
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,4,64,power_law_1.01,0.4969356918334961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,4,64,power_law_1.2,0.11087103843688965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,4,64,power_law_1.2,0.1571059226989746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,4,64,power_law_1.2,0.2088921546936035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,4,64,power_law_1.2,0.2707046318054199
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,8,32,balanced,0.021429760456085207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,8,32,balanced,0.022245121002197266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,8,32,balanced,0.02104192018508911
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,8,32,balanced,0.023086080551147463
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,8,32,balanced,0.037918720245361325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,4,64,power_law_1.2,0.05088511943817139
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,8,32,balanced,0.03920896053314209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,8,32,balanced,0.039950079917907715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,8,32,balanced,0.0415334415435791
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,8,32,balanced,0.0399232006072998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,8,32,balanced,0.04025856018066406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,8,32,balanced,0.044942078590393064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,8,32,balanced,0.045192961692810056
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,8,32,balanced,0.04568064212799072
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,8,32,balanced,0.047322878837585444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,8,32,balanced,0.045411839485168456
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,8,32,balanced,0.04844799995422363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,8,32,balanced,0.04934400081634521
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,8,32,balanced,0.05503615856170654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,8,32,balanced,0.05262335777282715
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,8,32,balanced,0.05461631774902344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,8,32,balanced,0.0693670415878296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,8,32,balanced,0.07501952171325683
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,8,32,balanced,0.05807744026184082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,8,32,balanced,0.10618368148803711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,8,32,balanced,0.13109375953674315
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,8,32,balanced,0.17192960739135743
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,8,32,balanced,0.21633535385131836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,8,32,power_law_1.01,0.024148480892181395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,8,32,power_law_1.01,0.02407167911529541
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,8,32,power_law_1.01,0.03752063989639282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,8,32,power_law_1.01,0.038835198879241944
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,8,32,power_law_1.01,0.022826240062713624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,4,64,power_law_1.2,0.45705215454101566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,8,32,power_law_1.01,0.04114304065704345
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,8,32,power_law_1.01,0.04007167816162109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,8,32,power_law_1.01,0.04353919982910156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,8,32,power_law_1.01,0.04764671802520752
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,8,32,power_law_1.01,0.04583680152893067
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,8,32,power_law_1.01,0.04893439769744873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,8,32,power_law_1.01,0.050193920135498046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,8,32,power_law_1.01,0.051846399307250976
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,8,32,power_law_1.01,0.054711041450500486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,8,32,power_law_1.01,0.056547842025756835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,8,32,power_law_1.01,0.05502592086791992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,8,32,power_law_1.01,0.04166016101837158
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,8,32,power_law_1.01,0.06125184059143066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,8,32,power_law_1.01,0.08477055549621582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,8,32,power_law_1.01,0.09845248222351075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,8,32,power_law_1.01,0.12331775665283203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,8,32,power_law_1.01,0.16450815200805663
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,4,64,power_law_1.2,0.5654476928710938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,8,32,power_law_1.01,0.19028736114501954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,8,32,power_law_1.01,0.26645120620727536
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,8,32,power_law_1.01,0.34247169494628904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,8,32,power_law_1.2,0.0230732798576355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,8,32,power_law_1.2,0.023994879722595213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,8,32,power_law_1.2,0.02350464105606079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,8,32,power_law_1.01,0.47861248016357416
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,8,32,power_law_1.2,0.037425920963287354
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,8,32,power_law_1.2,0.03946752071380615
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,8,32,power_law_1.01,0.6232742309570313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,8,32,power_law_1.2,0.03984256029129028
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,8,32,power_law_1.2,0.04484479904174805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,8,32,power_law_1.2,0.04434175968170166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,8,32,power_law_1.2,0.04607232093811035
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,8,32,power_law_1.2,0.04814591884613037
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,8,32,power_law_1.2,0.05059584140777588
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,8,32,power_law_1.2,0.052737278938293455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,8,32,power_law_1.2,0.054041600227355956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,8,32,power_law_1.2,0.05454976081848144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,8,32,power_law_1.2,0.05660927772521973
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,8,32,power_law_1.2,0.055010561943054204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,8,32,power_law_1.2,0.06805759906768799
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,2,128,power_law_1.2,0.11143168449401855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,8,32,power_law_1.2,0.0927462387084961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,8,32,power_law_1.2,0.10010368347167968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,8,32,power_law_1.2,0.12446463584899903
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,8,32,power_law_1.2,0.18299135208129882
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,8,32,power_law_1.2,0.22387968063354494
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,8,32,power_law_1.2,0.04151296138763428
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,16,16,balanced,0.021639680862426756
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,16,16,balanced,0.021043200492858887
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,16,16,balanced,0.021425919532775877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,16,16,balanced,0.023792641162872313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,16,16,balanced,0.025196800231933592
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,16,16,balanced,0.03948031902313233
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,16,16,balanced,0.04194047927856445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,16,16,balanced,0.041899518966674806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,16,16,balanced,0.04241536140441894
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,2,128,power_law_1.2,0.1311795234680176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,16,16,balanced,0.043782401084899905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,16,16,balanced,0.04486527919769287
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,16,16,balanced,0.045957121849060055
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,16,16,balanced,0.046489601135253904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,16,16,balanced,0.04766592025756836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,16,16,balanced,0.05380735874176026
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,16,16,balanced,0.04987391948699951
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,16,16,balanced,0.05018879890441894
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,16,16,balanced,0.053547520637512204
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,16,16,balanced,0.06576255798339845
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,16,16,balanced,0.07336063861846924
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,16,16,balanced,0.10483712196350096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,16,16,balanced,0.11887231826782227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,8,32,power_law_1.2,0.3344755172729492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,16,16,balanced,0.15925632476806642
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,16,16,balanced,0.19412096023559572
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,16,16,balanced,0.267011833190918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,16,16,power_law_1.01,0.023953919410705564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,16,16,balanced,0.33902591705322266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,16,16,balanced,0.052207360267639155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,16,16,power_law_1.01,0.026521599292755126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,16,16,power_law_1.01,0.026516480445861818
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,16,16,power_law_1.01,0.03879040002822876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,16,16,power_law_1.01,0.04012544155120849
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,16,16,power_law_1.01,0.04194303989410401
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,16,16,power_law_1.01,0.04307072162628174
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,16,16,power_law_1.01,0.044375038146972655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,16,16,power_law_1.01,0.04453120231628418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,16,16,power_law_1.01,0.0475110387802124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,16,16,power_law_1.01,0.04848639965057373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,8,32,power_law_1.2,0.483006706237793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,16,16,power_law_1.01,0.051363840103149414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,16,16,power_law_1.01,0.055902719497680664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,16,16,power_law_1.01,0.05257599830627442
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,16,16,power_law_1.01,0.05355264186859131
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,16,16,power_law_1.01,0.056939520835876466
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,16,16,power_law_1.01,0.062357759475708006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,16,16,power_law_1.01,0.07336575984954834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,16,16,power_law_1.01,0.10130047798156738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,16,16,power_law_1.01,0.11553152084350586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,16,16,power_law_1.01,0.15234175682067871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,16,16,power_law_1.01,0.026750719547271727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,16,16,power_law_1.01,0.17620864868164063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,16,16,power_law_1.01,0.28380287170410157
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,16,16,power_law_1.01,0.34929279327392576
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,16,16,power_law_1.2,0.02347520112991333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,16,16,power_law_1.2,0.025181438922882084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,16,16,power_law_1.2,0.02634880065917969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,16,16,power_law_1.2,0.026776320934295654
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,16,16,power_law_1.01,0.7288050842285156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,16,16,power_law_1.2,0.03863935947418213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,16,16,power_law_1.2,0.04101759910583496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,16,16,power_law_1.2,0.04428287982940674
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,16,16,power_law_1.2,0.043912959098815915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,16,16,power_law_1.2,0.04453375816345215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,16,16,power_law_1.2,0.04650623798370361
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,16,16,power_law_1.2,0.04814464092254639
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,16,16,power_law_1.2,0.05007359981536865
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,16,16,power_law_1.2,0.052203521728515626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,16,16,power_law_1.2,0.05363711833953857
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,16,16,power_law_1.2,0.05358719825744629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,16,16,power_law_1.2,0.05379712104797363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,16,16,power_law_1.2,0.057966079711914066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,16,16,power_law_1.2,0.07502463817596436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,16,16,power_law_1.2,0.08825984001159667
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,16,16,power_law_1.2,0.10329471588134767
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,16,16,power_law_1.2,0.12421119689941407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,16,16,power_law_1.2,0.16972799301147462
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,8,32,power_law_1.2,0.7209651184082031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,16,16,power_law_1.2,0.20766592025756836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,4,64,power_law_1.2,0.8589772796630859
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,32,8,balanced,0.0210150408744812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,32,8,balanced,0.022350080013275146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,32,8,balanced,0.022677760124206543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,32,8,balanced,0.029304320812225344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,32,8,balanced,0.045987839698791506
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,16,16,power_law_1.01,0.5321625518798828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,32,8,balanced,0.04725759983062744
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,16,16,power_law_1.2,0.2851046371459961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,32,8,balanced,0.04803199768066406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,32,8,balanced,0.050196480751037595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,8,32,power_law_1.2,0.9184614562988281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,32,8,balanced,0.05092991828918457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,32,8,balanced,0.055448322296142584
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,32,8,balanced,0.051584000587463374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,32,8,balanced,0.055302400588989255
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,32,8,balanced,0.052225279808044436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,32,8,balanced,0.05551616191864014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,32,8,balanced,0.061882882118225103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,32,8,balanced,0.024766719341278075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,32,8,balanced,0.07080832004547119
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,16,16,power_law_1.2,0.40756351470947266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,32,8,balanced,0.10453887939453126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,32,8,balanced,0.11827584266662598
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,32,8,balanced,0.15534208297729493
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,32,8,balanced,0.18295936584472655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,32,8,balanced,0.25504255294799805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,32,8,balanced,0.048944640159606936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,32,8,balanced,0.3184972763061523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,32,8,power_law_1.01,0.022666239738464357
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,32,8,power_law_1.01,0.026768639087677005
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,32,8,balanced,0.4494579315185547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,32,8,balanced,0.0572160005569458
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,32,8,power_law_1.01,0.025985279083251954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,32,8,balanced,0.5811187362670898
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,32,8,power_law_1.01,0.02939775943756103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,32,8,power_law_1.01,0.028083200454711917
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,32,8,power_law_1.01,0.0456601619720459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,32,8,power_law_1.01,0.0436620807647705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,32,8,power_law_1.01,0.04770559787750244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,32,8,power_law_1.01,0.04886144161224365
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,32,8,power_law_1.01,0.05010816097259522
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,32,8,power_law_1.01,0.05148543834686279
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,32,8,power_law_1.01,0.05464831829071045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,32,8,power_law_1.01,0.05788032054901123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,32,8,power_law_1.01,0.05543424129486084
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,32,8,power_law_1.01,0.05548287868499756
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,32,8,power_law_1.01,0.05592063903808594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,32,8,power_law_1.01,0.06910592079162597
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,32,8,power_law_1.01,0.08513279914855956
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,32,8,power_law_1.01,0.10832256317138671
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,32,8,power_law_1.01,0.13425408363342287
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,32,8,power_law_1.01,0.16632192611694335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,32,8,power_law_1.01,0.21319936752319335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,32,8,power_law_1.01,0.29495935440063475
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,32,8,power_law_1.01,0.3426931381225586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,32,8,power_law_1.2,0.023129599094390867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,32,8,power_law_1.2,0.025139200687408447
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,32,8,power_law_1.2,0.026433279514312746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,32,8,power_law_1.01,0.5733887863159179
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,32,8,power_law_1.2,0.02683903932571411
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,32,8,power_law_1.2,0.02881407976150513
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,32,8,power_law_1.2,0.045698561668396
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,32,8,power_law_1.2,0.04609151840209961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,32,8,power_law_1.01,0.738598403930664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,32,8,power_law_1.2,0.048163838386535644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,16,16,power_law_1.2,0.6405427551269531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,32,8,power_law_1.2,0.04933375835418701
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,32,8,power_law_1.2,0.050603518486022944
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,32,8,power_law_1.2,0.0532748794555664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,32,8,power_law_1.2,0.055406079292297364
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,32,8,power_law_1.2,0.056593918800354005
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,32,8,power_law_1.2,0.05467264175415039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,32,8,power_law_1.2,0.05638912200927735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,32,8,power_law_1.2,0.06204415798187256
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,32,8,power_law_1.2,0.08311296463012695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,32,8,power_law_1.2,0.09021183967590332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,32,8,power_law_1.2,0.10945152282714843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,32,8,power_law_1.2,0.12055168151855469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,32,8,power_law_1.2,0.1823219108581543
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,32,8,power_law_1.2,0.21847679138183596
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,32,8,power_law_1.01,0.05996928215026855
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,256,balanced,0.036607999801635746
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,32,8,power_law_1.2,0.05500927925109863
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,256,balanced,0.03658495903015137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,256,balanced,0.06394624233245849
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,32,8,power_law_1.2,0.3212108612060547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,256,balanced,0.10660863876342774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,256,balanced,0.10701055526733398
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,256,balanced,0.11050623893737792
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,256,balanced,0.11185664176940917
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,256,balanced,0.11028351783752441
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,256,balanced,0.11345791816711426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,256,balanced,0.036287999153137206
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,256,balanced,0.1154700756072998
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,256,balanced,0.1161855983734131
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,256,balanced,0.11844736099243165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,256,balanced,0.1269977569580078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,256,balanced,0.13409152030944824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,256,balanced,0.14041855812072754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,32,8,power_law_1.2,0.39957504272460936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,256,balanced,0.13610239982604982
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,256,balanced,0.14652928352355957
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,256,balanced,0.1679961585998535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,256,balanced,0.1892019271850586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,256,balanced,0.24177919387817384
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,256,balanced,0.32260990142822266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,256,balanced,0.12296832084655762
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,256,balanced,0.40950912475585943
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,256,balanced,0.5507225418090821
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,256,balanced,0.03416960000991821
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,256,balanced,0.0353164792060852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,256,balanced,0.036567039489746093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,256,balanced,0.7028659057617188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,256,balanced,0.043147521018981935
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,256,balanced,0.21085439682006837
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,256,balanced,0.04443391799926758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,256,balanced,0.04440192222595215
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,256,balanced,0.04525568008422852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,256,balanced,0.04563456058502197
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,256,balanced,0.04701056003570557
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,256,balanced,0.047006721496582034
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,256,balanced,0.050981121063232424
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,256,balanced,0.05261824131011963
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,32,8,power_law_1.2,0.6456114959716797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,256,balanced,0.05704959869384766
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,256,balanced,0.06295807838439942
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,256,balanced,0.06327680110931397
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,256,balanced,0.043598079681396486
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,256,balanced,0.07257215976715088
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,256,balanced,0.08318079948425293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,256,balanced,0.09592320442199706
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,256,balanced,0.10798720359802245
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,256,balanced,0.12955648422241212
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,256,balanced,0.17260160446166992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,256,balanced,0.21220735549926756
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,256,balanced,0.29853952407836915
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,256,balanced,0.37593727111816405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,256,balanced,0.03457407951354981
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,16,16,power_law_1.2,0.8835097503662108
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,256,balanced,0.04449408054351807
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,32,8,power_law_1.2,0.8003084564208984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,1,256,balanced,0.0186572802066803
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,1,256,balanced,0.019203840494155883
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,1,256,balanced,0.018650879859924314
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,1,256,balanced,0.019456000328063966
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,1,256,balanced,0.01913599967956543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,1,256,balanced,0.01937280058860779
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,1,256,balanced,0.019285759925842284
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,1,256,balanced,0.01884160041809082
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,1,256,balanced,0.019404799938201905
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,1,256,balanced,0.019449599981307984
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,1,256,balanced,0.019879679679870605
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,1,256,balanced,0.01995136022567749
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,1,256,balanced,0.020211200714111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,1,256,balanced,0.020250880718231203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,1,256,balanced,0.02019711971282959
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,1,256,balanced,0.018982399702072144
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,1,256,balanced,0.01906559944152832
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,1,256,balanced,0.019041279554367067
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,1,256,balanced,0.01922304034233093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,1,256,balanced,0.019351040124893186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,256,power_law_1.01,0.06303487777709961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,256,power_law_1.01,0.06452735900878906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,256,power_law_1.01,0.06434304237365722
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,256,power_law_1.01,0.1058240032196045
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,1,256,balanced,0.01939584016799927
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,256,power_law_1.01,0.10929408073425292
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,256,power_law_1.01,0.11116160392761229
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,256,power_law_1.01,0.11493887901306152
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,256,power_law_1.01,0.11633919715881347
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,256,power_law_1.01,0.12339584350585937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,256,power_law_1.01,0.12701824188232422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,256,power_law_1.01,0.13385984420776367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,256,power_law_1.01,0.1335334396362305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,256,power_law_1.01,0.1372979164123535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,256,power_law_1.01,0.14347904205322265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,256,power_law_1.01,0.15763839721679687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,256,power_law_1.01,0.15957632064819335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,256,power_law_1.01,0.18750080108642578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,256,power_law_1.01,0.22999040603637697
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,256,power_law_1.01,0.26182912826538085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,256,power_law_1.01,0.3510579299926758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,256,power_law_1.01,0.4732096099853516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,256,power_law_1.01,0.6909503936767578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,256,power_law_1.01,0.936874237060547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,256,power_law_1.01,1.3725196838378906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,256,power_law_1.01,1.9882727050781248
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,256,power_law_1.01,3.12732421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,256,power_law_1.01,0.03502079963684082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,256,power_law_1.01,0.035013120174407955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,256,power_law_1.01,0.03500672101974488
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,256,power_law_1.01,0.04331264019012451
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,256,power_law_1.01,0.04486144065856933
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,256,power_law_1.01,0.045679359436035155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,256,power_law_1.01,4.51997314453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,256,power_law_1.01,0.04705535888671875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,256,power_law_1.01,0.048774399757385255
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,256,power_law_1.01,0.050621438026428225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,256,power_law_1.01,0.059111680984497074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,256,power_law_1.01,0.065731840133667
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,256,power_law_1.01,0.06195712089538574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,256,power_law_1.01,0.0649561595916748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,256,power_law_1.01,0.07542143821716309
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,256,power_law_1.01,0.08091903686523437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,256,power_law_1.01,0.08659584045410156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,256,power_law_1.01,0.1031270408630371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,256,power_law_1.01,0.12191871643066406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,256,power_law_1.01,0.144519681930542
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,256,power_law_1.01,0.18536319732666015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,256,power_law_1.01,0.2246937561035156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,256,power_law_1.01,0.3788659286499024
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,256,power_law_1.01,0.48969982147216795
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,256,power_law_1.01,0.8089151763916016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,256,power_law_1.01,1.0927999877929688
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,1,256,balanced,0.03195264101028442
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,1,256,balanced,0.04594560146331787
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,1,256,balanced,0.02601344108581543
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,1,256,balanced,0.025199360847473144
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,1,256,power_law_1.01,0.019288320541381836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,256,power_law_1.01,1.6557798767089842
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,1,256,power_law_1.01,0.019404799938201905
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,1,256,power_law_1.01,0.019006719589233396
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,1,256,balanced,0.03310976028442383
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,1,256,power_law_1.01,0.019037439823150634
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,1,256,power_law_1.01,0.019037439823150634
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,1,256,balanced,0.052203521728515626
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,1,256,power_law_1.01,0.019445760250091555
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,1,256,power_law_1.01,0.01938815951347351
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,1,256,power_law_1.01,0.019393279552459716
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,1,256,power_law_1.01,0.019466240406036377
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,1,256,power_law_1.01,0.019799040555953978
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,1,256,power_law_1.01,0.01988736033439636
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,1,256,power_law_1.01,0.02046207904815674
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,1,256,power_law_1.01,0.01995519995689392
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,1,256,power_law_1.01,0.020117759704589844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,1,256,power_law_1.01,0.021456639766693115
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,1,256,power_law_1.01,0.020213758945465087
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,256,power_law_1.01,2.1053900146484374
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,1,256,power_law_1.01,0.019066879749298098
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,1,256,power_law_1.01,0.06963712215423584
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,1,256,power_law_1.01,0.04324736118316651
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,1,256,power_law_1.01,0.02482815980911255
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,1,256,power_law_1.01,0.03507711887359619
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,1,256,power_law_1.01,0.02768383979797363
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,256,power_law_1.2,0.06368768215179443
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,256,power_law_1.2,0.06281343936920167
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,1,256,power_law_1.01,0.05683328151702881
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,256,power_law_1.2,0.10498432159423828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,256,power_law_1.2,0.1073305606842041
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,1,256,power_law_1.01,0.10000896453857422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,256,power_law_1.2,0.11060352325439453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,256,power_law_1.2,0.11586175918579103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,256,power_law_1.2,0.12060031890869141
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,256,power_law_1.2,0.12583295822143553
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,256,power_law_1.2,0.12962559700012205
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,256,power_law_1.2,0.1377996826171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,256,power_law_1.2,0.1366758441925049
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,256,power_law_1.2,0.14443136215209962
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,256,power_law_1.2,0.06236159801483154
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,256,power_law_1.2,0.15589119911193847
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,256,power_law_1.2,0.17308927536010743
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,256,power_law_1.2,0.18291584014892578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,256,power_law_1.2,0.2109030342102051
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,256,power_law_1.2,0.245294075012207
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,256,power_law_1.2,0.33400577545166016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,256,power_law_1.2,0.4522931289672852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,256,power_law_1.2,0.6311449432373047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,256,power_law_1.2,0.8800627136230469
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,1,256,power_law_1.01,0.1304255962371826
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,256,power_law_1.2,1.389566650390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,256,power_law_1.2,0.03461888074874878
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,256,power_law_1.2,2.3617330932617184
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,256,power_law_1.2,0.03511168003082275
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,256,power_law_1.2,0.035861759185791015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,256,power_law_1.2,0.04399871826171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,256,power_law_1.2,0.044709119796752926
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,256,power_law_1.2,3.2956747436523437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,256,power_law_1.2,0.04723455905914307
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,256,power_law_1.2,0.04528895854949951
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,256,power_law_1.2,0.04936704158782959
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,256,power_law_1.2,0.055048961639404294
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,256,power_law_1.2,0.060756478309631344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,256,power_law_1.2,0.05340799808502197
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,256,power_law_1.2,0.06760831832885741
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,256,power_law_1.2,0.07468160152435302
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,256,power_law_1.2,0.08337535858154296
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,256,power_law_1.2,0.0963315200805664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,256,power_law_1.2,0.11220864295959472
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,256,power_law_1.2,0.13142656326293944
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,256,power_law_1.2,0.17772287368774414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,256,power_law_1.2,0.25061248779296874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,256,power_law_1.2,0.3207398223876953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,256,power_law_1.2,0.06239744186401367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,256,power_law_1.2,0.5051545715332031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,256,power_law_1.2,0.7182361602783203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,256,power_law_1.2,5.30036376953125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,256,power_law_1.2,1.1450406646728515
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,1,256,power_law_1.2,0.019041279554367067
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,1,256,power_law_1.2,0.019304959774017333
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,1,256,power_law_1.2,0.01904255986213684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,256,power_law_1.2,1.6740928649902345
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,1,256,power_law_1.2,0.019038720130920412
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,1,256,power_law_1.2,0.019271680116653443
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,1,256,power_law_1.2,0.019325439929962156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,1,256,power_law_1.2,0.0191103994846344
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,1,256,power_law_1.2,0.01946239948272705
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,1,256,power_law_1.2,0.019444479942321777
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,1,256,power_law_1.2,0.019868160486221316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,256,power_law_1.2,7.4033703613281245
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,1,256,power_law_1.2,0.020195839405059816
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,1,256,power_law_1.2,0.020198400020599368
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,1,256,power_law_1.2,0.020441598892211914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,1,256,power_law_1.2,0.019307520389556885
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,1,256,power_law_1.2,0.019855359792709353
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,256,power_law_1.2,2.508994598388672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,1,256,power_law_1.2,0.041873922348022466
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,1,256,power_law_1.01,0.17761280059814452
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,1,256,power_law_1.01,0.24273279190063474
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,256,power_law_1.2,3.5850329589843755
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,1,256,power_law_1.2,0.028120319843292236
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,1,256,power_law_1.2,0.021122560501098633
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,1,256,power_law_1.2,0.022566399574279784
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,1,256,power_law_1.2,0.032647678852081294
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,1,256,power_law_1.2,0.05431551933288574
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,128,balanced,0.030113279819488525
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,128,balanced,0.0317248010635376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,128,balanced,0.050900478363037106
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,128,balanced,0.09087615966796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,128,balanced,0.09329919815063477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,128,balanced,0.0949516773223877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,128,balanced,0.09921024322509767
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,128,balanced,0.030563840866088866
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,128,balanced,0.09895296096801758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,128,balanced,0.10167039871215819
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,128,balanced,0.10494463920593262
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,128,balanced,0.11561856269836426
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,128,balanced,0.11196672439575195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,128,balanced,0.10805888175964355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,128,balanced,0.10992128372192382
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,128,balanced,0.11406208038330079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,128,balanced,0.12349823951721191
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,128,balanced,0.1388428783416748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,128,balanced,0.15533311843872072
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,128,balanced,0.17074176788330078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,128,balanced,0.19610496520996096
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,128,balanced,0.24299648284912106
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,128,balanced,0.2965030479431152
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,128,balanced,0.4003801727294922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,128,balanced,0.5213926315307618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,128,balanced,0.7311334228515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,128,balanced,0.032381439208984376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,128,balanced,0.032578558921813966
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,128,balanced,0.9383833312988281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,128,balanced,0.03255424022674561
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,128,balanced,0.03174783945083618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,128,balanced,0.040515837669372556
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,128,balanced,0.04193280220031738
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,128,balanced,0.042896637916564936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,128,balanced,0.044615678787231446
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,128,balanced,0.0441267204284668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,128,balanced,0.04684160232543945
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,128,balanced,0.05510655879974365
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,128,balanced,0.06024320125579834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,128,balanced,0.052445440292358404
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,128,balanced,0.05277567863464355
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,128,balanced,0.055015678405761714
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,128,balanced,0.06339712142944336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,128,balanced,0.04772480010986328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,128,balanced,0.07384704113006592
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,128,balanced,0.08463359832763671
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,128,balanced,0.08642687797546386
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,128,balanced,0.10310912132263184
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,128,balanced,0.14042495727539062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,128,balanced,0.1735628890991211
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,128,balanced,0.2336204719543457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,128,balanced,0.29015039443969726
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,128,balanced,0.42436096191406253
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,128,balanced,0.5402073669433594
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,1,256,power_law_1.2,0.07236735820770264
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,1,256,power_law_1.2,0.10566911697387696
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,1,256,power_law_1.2,0.1466483211517334
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,1,256,power_law_1.2,0.18384256362915039
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,1,256,power_law_1.2,0.29142784118652343
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,1,256,power_law_1.2,0.38504318237304686
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,2,128,balanced,0.018659839630126955
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,2,128,balanced,0.01953279972076416
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,2,128,balanced,0.019466240406036377
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,2,128,balanced,0.0186355197429657
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,2,128,balanced,0.019508479833602904
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,2,128,balanced,0.019418879747390747
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,2,128,balanced,0.019443199634552003
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,2,128,balanced,0.01978368043899536
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,2,128,balanced,0.019472639560699462
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,2,128,balanced,0.01821568012237549
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,2,128,balanced,0.01900287985801697
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,2,128,balanced,0.020209920406341553
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,2,128,balanced,0.02014336109161377
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,2,128,balanced,0.020236799716949465
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,2,128,balanced,0.02048511981964111
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,2,128,balanced,0.020696320533752442
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,2,128,balanced,0.019660799503326415
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,2,128,balanced,0.01947648048400879
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,2,128,balanced,0.019399679899215698
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,2,128,balanced,0.019056639671325683
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,128,power_law_1.01,0.05426815986633301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,128,power_law_1.01,0.05609600067138672
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,2,128,balanced,0.019394559860229494
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,128,power_law_1.01,0.056316161155700685
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,128,power_law_1.01,0.08645248413085938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,128,power_law_1.01,0.08997504234313965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,128,power_law_1.01,0.09542655944824219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,128,power_law_1.01,0.10064512252807618
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,128,power_law_1.01,0.10381312370300293
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,128,power_law_1.01,0.10750080108642579
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,128,power_law_1.01,0.1138163185119629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,128,power_law_1.01,0.11429247856140137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,128,power_law_1.01,0.11174015998840332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,128,power_law_1.01,0.11807104110717774
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,128,power_law_1.01,0.12862079620361327
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,128,power_law_1.01,0.144519681930542
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,128,power_law_1.01,0.15895808219909668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,128,power_law_1.01,0.16824703216552733
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,128,power_law_1.01,0.20845567703247073
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,128,power_law_1.01,0.2255206489562988
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,128,power_law_1.01,0.30056703567504883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,128,power_law_1.01,0.43110912322998046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,128,power_law_1.01,0.6668300628662109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,128,power_law_1.01,0.8324211120605469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,128,power_law_1.01,1.3865740966796873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,128,power_law_1.01,1.551146240234375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,128,power_law_1.01,2.5668890380859373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,128,power_law_1.01,0.032647678852081294
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,128,power_law_1.01,0.03365504026412964
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,2,128,balanced,0.029743359088897702
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,128,power_law_1.01,0.03379584074020386
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,128,power_law_1.01,0.04035967826843262
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,2,128,balanced,0.037639679908752444
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,128,power_law_1.01,0.04270592212677002
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,128,power_law_1.01,0.04128128051757812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,128,power_law_1.01,3.7256396484375003
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,128,power_law_1.01,0.04712063789367676
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,128,power_law_1.01,0.045747199058532716
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,128,power_law_1.01,0.05225728034973145
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,2,128,balanced,0.035380480289459226
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,128,power_law_1.01,0.05545983791351319
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,128,power_law_1.01,0.05557375907897949
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,128,power_law_1.01,0.06056960105895996
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,128,power_law_1.01,0.06542208194732665
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,128,power_law_1.01,0.08379391670227052
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,128,power_law_1.01,0.06978816032409668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,128,power_law_1.01,0.08675071716308594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,128,power_law_1.01,0.09150976181030274
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,128,power_law_1.01,0.11504128456115723
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,128,power_law_1.01,0.12676735877990725
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,128,power_law_1.01,0.17612031936645506
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,2,128,balanced,0.05230080127716065
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,128,power_law_1.01,0.2507174491882324
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,128,power_law_1.01,0.4206784057617187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,128,power_law_1.01,0.4887961578369141
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,2,128,balanced,0.028124160766601562
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,2,128,power_law_1.01,0.019453439712524414
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,2,128,power_law_1.01,0.01911296010017395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,2,128,power_law_1.01,0.01911679983139038
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,128,power_law_1.01,0.7412095642089843
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,2,128,power_law_1.01,0.019041279554367067
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,128,power_law_1.01,0.9398361968994141
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,2,128,power_law_1.01,0.019198720455169675
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,2,128,power_law_1.01,0.019383039474487305
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,2,128,power_law_1.01,0.019518719911575316
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,2,128,power_law_1.01,0.019436800479888917
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,2,128,power_law_1.01,0.019467519521713255
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,2,128,power_law_1.01,0.019523839950561523
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,2,128,power_law_1.01,0.01980672001838684
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,128,power_law_1.01,1.349276123046875
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,2,128,power_law_1.01,0.020225279331207276
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,2,128,power_law_1.01,0.019900159835815428
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,2,128,power_law_1.01,0.020168960094451904
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,2,128,power_law_1.01,0.020264959335327147
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,2,128,power_law_1.01,0.02159615993499756
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,2,128,power_law_1.01,0.029128959178924562
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,128,power_law_1.01,1.8145971679687498
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,2,128,power_law_1.01,0.01937280058860779
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,2,128,balanced,0.061452798843383796
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,2,128,power_law_1.01,0.03433471918106079
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,2,128,power_law_1.01,0.04241919994354248
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,2,128,power_law_1.01,0.024376320838928222
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,2,128,power_law_1.01,0.05796224117279053
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,2,128,power_law_1.01,0.06550655841827394
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,128,power_law_1.2,0.05491327762603759
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,128,power_law_1.2,0.05689087867736816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,128,power_law_1.2,0.05798143863677978
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,128,power_law_1.2,0.08949119567871093
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,2,128,power_law_1.01,0.08880127906799316
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,128,power_law_1.2,0.0973043155670166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,128,power_law_1.2,0.09898495674133301
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,128,power_law_1.2,0.10655103683471681
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,128,power_law_1.2,0.11181440353393554
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,128,power_law_1.2,0.11323904037475585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,128,power_law_1.2,0.11376383781433105
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,128,power_law_1.2,0.11462271690368653
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,128,power_law_1.2,0.12805888175964356
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,128,power_law_1.2,0.14436351776123046
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,128,power_law_1.2,0.14965503692626952
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,128,power_law_1.2,0.15981696128845216
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,2,128,power_law_1.01,0.11338496208190918
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,128,power_law_1.2,0.18281728744506837
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,128,power_law_1.2,0.1988198471069336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,128,power_law_1.2,0.28533119201660156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,128,power_law_1.2,0.08978688240051269
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,128,power_law_1.2,0.43446529388427735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,128,power_law_1.2,0.5142182540893555
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,128,power_law_1.2,0.8153778839111329
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,128,power_law_1.2,1.1114137268066406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,128,power_law_1.2,0.03135871887207031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,128,power_law_1.2,0.032321279048919675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,128,power_law_1.2,0.0341593599319458
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,128,power_law_1.2,0.0403391981124878
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,128,power_law_1.2,0.041578240394592285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,128,power_law_1.2,1.9029592895507812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,128,power_law_1.2,0.049108481407165526
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,128,power_law_1.2,0.04458879947662354
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,128,power_law_1.2,0.05547647953033448
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,128,power_law_1.2,2.4548710632324218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,128,power_law_1.2,0.0506931209564209
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,128,power_law_1.2,0.0572057580947876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,128,power_law_1.2,0.0667852783203125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,128,power_law_1.2,0.05630720138549805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,128,power_law_1.2,0.0715558385848999
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,128,power_law_1.2,0.0793395185470581
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,128,power_law_1.2,0.09249279975891114
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,128,power_law_1.2,0.09396991729736329
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,128,power_law_1.2,0.10941823959350586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,128,power_law_1.2,0.044455680847167965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,128,power_law_1.2,0.16476287841796874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,128,power_law_1.2,0.24576000213623045
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,128,power_law_1.2,0.3065203285217285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,128,power_law_1.2,0.4159116744995117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,128,power_law_1.2,0.6314713668823242
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,2,128,power_law_1.01,0.17042688369750975
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,2,128,power_law_1.2,0.019088640213012695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,128,power_law_1.2,1.0485440063476563
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,2,128,power_law_1.2,0.01939072012901306
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,128,power_law_1.2,3.8374462890624996
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,2,128,power_law_1.2,0.019070719480514524
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,128,power_law_1.2,1.3837849426269533
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,2,128,power_law_1.2,0.01904639959335327
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,2,128,power_law_1.2,0.019427839517593384
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,2,128,power_law_1.2,0.019141119718551636
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,2,128,power_law_1.2,0.019477759599685666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,2,128,power_law_1.2,0.01942271947860718
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,2,128,power_law_1.2,0.019375360012054442
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,2,128,power_law_1.2,0.019790079593658447
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,2,128,power_law_1.2,0.01978495955467224
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,2,128,power_law_1.2,0.0200217604637146
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,2,128,power_law_1.2,0.020038399696350098
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,2,128,power_law_1.2,0.020414719581604006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,128,power_law_1.2,5.024215087890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,2,128,power_law_1.2,0.019443199634552003
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,128,power_law_1.2,2.2102886962890627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,128,power_law_1.2,2.848374938964844
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,2,128,power_law_1.01,0.2259507179260254
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,2,128,power_law_1.2,0.02193664073944092
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,2,128,power_law_1.2,0.02392064094543457
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,2,128,power_law_1.2,0.034703359603881836
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,2,128,power_law_1.2,0.029263360500335695
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,2,128,power_law_1.2,0.04031487941741944
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,4,64,balanced,0.026978559494018554
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,2,128,power_law_1.2,0.05035903930664063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,4,64,balanced,0.02759936094284058
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,4,64,balanced,0.029562880992889406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,4,64,balanced,0.052378878593444825
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,4,64,balanced,0.08865792274475097
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,4,64,balanced,0.091211519241333
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,4,64,balanced,0.09568639755249023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,4,64,balanced,0.09971327781677246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,4,64,balanced,0.1027353572845459
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,4,64,balanced,0.11010815620422362
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,4,64,balanced,0.09887743949890136
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,4,64,balanced,0.10416640281677245
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,4,64,balanced,0.10517248153686523
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,4,64,balanced,0.11401344299316407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,4,64,balanced,0.12118528366088868
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,4,64,balanced,0.127903995513916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,4,64,balanced,0.1414796829223633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,4,64,balanced,0.15205760002136232
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,4,64,balanced,0.1715328025817871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,4,64,balanced,0.21315839767456052
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,4,64,balanced,0.25642751693725585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,4,64,balanced,0.33656959533691405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,4,64,balanced,0.4154060745239258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,4,64,balanced,0.588211212158203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,2,128,power_law_1.2,0.06979072093963623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,4,64,balanced,0.759900131225586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,4,64,balanced,0.02845952033996582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,4,64,balanced,0.02983167886734009
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,4,64,balanced,0.03043584108352661
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,4,64,balanced,1.1055564880371094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,4,64,balanced,0.03347840070724487
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,4,64,balanced,0.04276095867156983
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,4,64,balanced,0.04526976108551025
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,4,64,balanced,0.04798463821411133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,4,64,balanced,1.4326463317871094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,4,64,balanced,0.05349631786346436
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,4,64,balanced,0.0511027193069458
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,4,64,balanced,0.0554969596862793
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,4,64,balanced,0.0521676778793335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,4,64,balanced,0.05384960174560547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,4,64,balanced,0.05536255836486816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,4,64,balanced,0.062277121543884276
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,4,64,balanced,0.06609407901763917
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,4,64,balanced,0.06940544128417969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,4,64,balanced,0.07864960193634032
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,4,64,balanced,0.08203776359558106
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,4,64,balanced,0.09421824455261231
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,4,64,balanced,0.11960960388183595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,4,64,balanced,0.14815744400024414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,4,64,balanced,0.20582656860351561
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,4,64,balanced,0.2546329689025879
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,4,64,balanced,0.3633536148071289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,4,64,balanced,0.45795455932617185
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,4,64,balanced,0.6742566680908203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,4,64,balanced,0.8704601287841797
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,2,128,power_law_1.2,0.08951680183410644
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,2,128,power_law_1.2,0.12933631896972658
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,2,128,power_law_1.2,0.1762508773803711
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,2,128,power_law_1.2,0.24540288925170897
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,2,128,power_law_1.2,0.35876735687255856
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,4,64,balanced,0.019014400243759156
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,4,64,balanced,0.021857280731201172
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,4,64,balanced,0.021864960193634035
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,4,64,balanced,0.02160128116607666
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,4,64,balanced,0.022247679233551025
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,4,64,balanced,0.019024640321731567
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,4,64,balanced,0.022279679775238037
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,4,64,balanced,0.02234112024307251
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,4,64,balanced,0.02226047992706299
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,4,64,balanced,0.022736639976501466
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,4,64,balanced,0.01948415994644165
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,4,64,balanced,0.022242560386657714
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,4,64,balanced,0.022310400009155275
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,4,64,balanced,0.023071999549865725
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,4,64,balanced,0.022734079360961914
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,4,64,balanced,0.024453120231628416
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,4,64,balanced,0.021098239421844484
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,4,64,balanced,0.021527040004730224
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,4,64,balanced,0.021870079040527343
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,4,64,balanced,0.021838080883026124
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,4,64,power_law_1.01,0.05221375942230224
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,4,64,power_law_1.01,0.05312128067016601
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,4,64,power_law_1.01,0.057292799949646
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,4,64,power_law_1.01,0.05942783832550049
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,4,64,power_law_1.01,0.08520832061767578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,4,64,power_law_1.01,0.09309568405151367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,4,64,power_law_1.01,0.09724543571472169
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,4,64,power_law_1.01,0.10077823638916014
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,4,64,balanced,0.02153088092803955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,4,64,power_law_1.01,0.10863743782043458
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,4,64,power_law_1.01,0.10018560409545899
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,4,64,power_law_1.01,0.10257920265197754
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,4,64,power_law_1.01,0.1076582431793213
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,4,64,power_law_1.01,0.11682559967041015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,4,64,power_law_1.01,0.12435327529907228
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,4,64,power_law_1.01,0.136814079284668
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,4,64,power_law_1.01,0.15035903930664063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,4,64,power_law_1.01,0.15648384094238282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,4,64,power_law_1.01,0.18839040756225586
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,4,64,power_law_1.01,0.22926336288452148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,4,64,power_law_1.01,0.33807872772216796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,4,64,power_law_1.01,0.3956428909301758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,4,64,power_law_1.01,0.5552384185791015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,4,64,power_law_1.01,0.7967667388916015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,4,64,power_law_1.01,1.1051046752929687
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,4,64,balanced,0.03716223955154419
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,4,64,power_law_1.01,1.4046783447265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,4,64,balanced,0.03589119911193848
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,4,64,balanced,0.04704895973205566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,4,64,power_law_1.01,0.03173376083374023
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,4,64,power_law_1.01,0.03251327991485596
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,4,64,power_law_1.01,0.03347968101501465
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,4,64,power_law_1.01,0.034698240756988526
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,4,64,power_law_1.01,0.041579518318176266
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,4,64,power_law_1.01,0.045329918861389165
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,4,64,power_law_1.01,0.050115838050842285
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,4,64,power_law_1.01,0.057784318923950195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,4,64,power_law_1.01,0.054580478668212896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,4,64,power_law_1.01,0.05279744148254395
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,4,64,power_law_1.01,0.058048000335693364
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,4,64,power_law_1.01,0.06144512176513672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,4,64,power_law_1.01,0.06601471900939941
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,4,64,power_law_1.01,0.07500415802001953
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,4,64,power_law_1.01,2.2317747497558593
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,4,64,power_law_1.01,0.08152704238891602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,4,64,power_law_1.01,0.08638208389282227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,4,64,power_law_1.01,0.09307904243469238
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,4,64,power_law_1.01,0.10796928405761719
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,4,64,balanced,0.04984320163726806
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,4,64,power_law_1.01,0.14179840087890624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,4,64,power_law_1.01,0.18316543579101563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,4,64,power_law_1.01,0.2398591995239258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,4,64,power_law_1.01,0.38342784881591796
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,4,64,power_law_1.01,0.48047233581542964
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,4,64,power_law_1.01,3.116435241699219
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,4,64,power_law_1.01,0.6917107391357422
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,4,64,power_law_1.01,0.021560320854187014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,4,64,power_law_1.01,0.7913804626464843
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,4,64,power_law_1.01,0.02150399923324585
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,4,64,power_law_1.01,0.02156543970108032
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,4,64,power_law_1.01,0.02191744089126587
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,4,64,power_law_1.01,0.02152064085006714
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,4,64,power_law_1.01,0.021539840698242187
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,4,64,power_law_1.01,0.021893119812011717
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,4,64,power_law_1.01,0.021934080123901366
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,4,64,power_law_1.01,0.02183295965194702
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,4,64,power_law_1.01,1.4004019165039063
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,4,64,power_law_1.01,0.021908481121063233
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,4,64,power_law_1.01,0.02189440011978149
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,4,64,power_law_1.01,0.022008318901062012
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,4,64,power_law_1.01,0.022296319007873537
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,4,64,power_law_1.01,0.022277119159698485
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,4,64,power_law_1.01,0.02248703956604004
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,4,64,power_law_1.01,0.02427903890609741
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,4,64,power_law_1.01,1.7625779724121096
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,4,64,power_law_1.01,0.02194175958633423
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,4,64,balanced,0.06934400081634522
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,4,64,balanced,0.08126079559326171
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,4,64,power_law_1.01,0.02639103889465332
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,4,64,power_law_1.01,0.04720767974853516
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,4,64,power_law_1.01,0.04002943992614746
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,4,64,power_law_1.01,0.05830016136169434
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,4,64,power_law_1.2,0.05262720108032226
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,4,64,power_law_1.01,0.030687999725341798
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,4,64,power_law_1.2,0.053427200317382816
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,4,64,power_law_1.2,0.0583897590637207
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,4,64,power_law_1.01,0.06615935802459717
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,4,64,power_law_1.2,0.08595456123352051
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,4,64,power_law_1.2,0.09981311798095703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,4,64,power_law_1.2,0.10270463943481445
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,4,64,power_law_1.2,0.10911616325378418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,4,64,power_law_1.2,0.10657792091369629
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,4,64,power_law_1.2,0.10517888069152832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,4,64,power_law_1.2,0.0577177619934082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,4,64,power_law_1.2,0.1113036823272705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,4,64,power_law_1.2,0.12024448394775392
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,4,64,power_law_1.2,0.1378764820098877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,4,64,power_law_1.2,0.147706880569458
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,4,64,power_law_1.2,0.09011584281921386
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,4,64,power_law_1.2,0.15150336265563963
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,4,64,power_law_1.2,0.1712895965576172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,4,64,power_law_1.2,0.18430591583251954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,4,64,power_law_1.2,0.25972351074218747
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,4,64,power_law_1.2,0.4075519943237305
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,4,64,power_law_1.2,0.4899903869628906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,4,64,power_law_1.2,0.7133478546142579
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,4,64,power_law_1.2,1.0325337219238282
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,4,64,power_law_1.2,1.4355148315429687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,4,64,power_law_1.2,0.03172863960266113
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,4,64,power_law_1.2,1.9965670776367186
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,4,64,power_law_1.2,0.032282879352569585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,4,64,power_law_1.2,0.03373823881149292
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,4,64,power_law_1.2,0.03469311952590942
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,4,64,power_law_1.01,0.10006272315979003
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,4,64,power_law_1.2,0.04218751907348633
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,4,64,power_law_1.2,2.7311001586914063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,4,64,power_law_1.2,0.04517375946044922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,4,64,power_law_1.01,0.12298751831054688
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,4,64,power_law_1.2,0.052609281539916994
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,4,64,power_law_1.2,0.04948224067687988
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,4,64,power_law_1.2,0.0602291202545166
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,4,64,power_law_1.2,0.05335936069488525
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,4,64,power_law_1.2,0.059526400566101076
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,4,64,power_law_1.2,0.0642739200592041
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,4,64,power_law_1.2,0.06532224178314208
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,4,64,power_law_1.2,0.07971583843231202
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,4,64,power_law_1.2,0.08791040420532227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,4,64,power_law_1.2,0.08660479545593261
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,4,64,power_law_1.2,0.09474687576293946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,4,64,power_law_1.2,0.10601984024047853
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,4,64,power_law_1.2,0.15870464324951172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,4,64,power_law_1.2,0.21951744079589847
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,4,64,power_law_1.2,0.2645235252380371
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,4,64,power_law_1.2,0.39106689453125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,4,64,power_law_1.2,0.5560665512084961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,4,64,power_law_1.2,0.8958937835693359
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,4,64,power_law_1.2,0.021858561038970947
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,4,64,power_law_1.2,1.1327718353271483
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,4,64,power_law_1.2,0.021848320960998535
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,4,64,power_law_1.2,0.0214899206161499
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,4,64,power_law_1.2,0.02167423963546753
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,4,64,power_law_1.2,0.021502718925476075
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,4,64,power_law_1.2,3.9797222900390623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,4,64,power_law_1.2,1.7645504760742188
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,4,64,power_law_1.2,0.021583359241485595
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,4,64,power_law_1.2,0.021475839614868163
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,4,64,power_law_1.2,0.021488640308380127
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,4,64,power_law_1.2,0.02211967945098877
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,4,64,power_law_1.2,0.021962239742279052
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,4,64,power_law_1.2,0.021904640197753907
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,4,64,power_law_1.2,0.021856000423431395
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,4,64,power_law_1.2,0.021893119812011717
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,4,64,power_law_1.2,0.022338559627532957
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,4,64,power_law_1.2,0.022353920936584473
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,4,64,power_law_1.2,2.463603210449219
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,4,64,power_law_1.01,0.18224895477294922
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,4,64,power_law_1.01,0.21841152191162108
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,4,64,power_law_1.2,0.023983359336853027
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,4,64,power_law_1.2,0.025501439571380614
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,4,64,power_law_1.2,0.03002880096435547
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,4,64,power_law_1.2,0.03652607917785645
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,4,64,power_law_1.2,0.04491007804870605
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,8,32,balanced,0.024880640506744385
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,8,32,balanced,0.02678272008895874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,8,32,balanced,0.029187839031219482
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,8,32,balanced,0.05049344062805176
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,8,32,balanced,0.06376704216003418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,8,32,balanced,0.09505151748657227
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,8,32,balanced,0.11086336135864258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,8,32,balanced,0.09782143592834472
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,8,32,balanced,0.10003328323364258
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,8,32,balanced,0.10225024223327636
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,8,32,balanced,0.11118080139160155
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,8,32,balanced,0.11707519531249999
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,8,32,balanced,0.12389887809753417
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,8,32,balanced,0.1368396759033203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,8,32,balanced,0.1470400047302246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,8,32,balanced,0.14173824310302735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,8,32,balanced,0.15928064346313478
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,8,32,balanced,0.19345279693603518
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,8,32,balanced,0.23180416107177731
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,8,32,balanced,0.3044710350036621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,8,32,balanced,0.37785217285156253
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,4,64,power_law_1.2,0.05497087955474853
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,8,32,balanced,0.6596685028076171
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,8,32,balanced,0.5236723327636719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,8,32,balanced,0.9551654052734374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,8,32,balanced,1.2466393280029298
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,8,32,balanced,1.8447769165039063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,8,32,balanced,0.028430080413818358
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,8,32,balanced,0.02962048053741455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,8,32,balanced,0.03176448106765747
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,4,64,power_law_1.2,0.07368192195892334
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,8,32,balanced,0.03735039949417114
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,8,32,balanced,0.041532158851623535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,8,32,balanced,0.05142655849456788
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,8,32,balanced,0.059852800369262694
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,8,32,balanced,0.05428736209869385
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,8,32,balanced,2.4142796325683595
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,8,32,balanced,0.05607039928436279
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,8,32,balanced,0.05714943885803223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,8,32,balanced,0.06239999771118164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,8,32,balanced,0.06539904117584229
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,8,32,balanced,0.08009984016418456
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,8,32,balanced,0.08707327842712402
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,8,32,balanced,0.08072959899902343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,8,32,balanced,0.09207039833068847
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,8,32,balanced,0.1134937572479248
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,4,64,power_law_1.2,0.09069696426391602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,8,32,balanced,0.1401561641693115
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,8,32,balanced,0.18070528030395508
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,8,32,balanced,0.2283635139465332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,8,32,balanced,0.3314547348022461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,8,32,balanced,0.06976128101348877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,8,32,balanced,0.41659519195556644
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,8,32,balanced,0.5972159957885742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,8,32,balanced,0.7664934539794922
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,8,32,power_law_1.01,0.055230717658996585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,8,32,balanced,1.139678726196289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,8,32,balanced,1.4885554504394531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,8,32,power_law_1.01,0.05929599761962891
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,8,32,power_law_1.01,0.05994751930236817
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,8,32,power_law_1.01,0.09220992088317871
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,8,32,power_law_1.01,0.05407104015350341
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,8,32,power_law_1.01,0.10087295532226563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,8,32,power_law_1.01,0.05586048126220703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,8,32,power_law_1.01,0.10353407859802247
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,8,32,power_law_1.01,0.10206720352172852
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,8,32,power_law_1.01,0.10663552284240722
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,8,32,power_law_1.01,0.11429375648498534
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,8,32,power_law_1.01,0.12134271621704103
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,8,32,power_law_1.01,0.13760640144348144
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,8,32,power_law_1.01,0.1500915241241455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,8,32,power_law_1.01,0.15432576179504393
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,8,32,power_law_1.01,0.16354175567626955
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,8,32,power_law_1.01,0.2035481643676758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,8,32,power_law_1.01,0.24836736679077148
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,8,32,power_law_1.01,0.3598348617553711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,8,32,power_law_1.01,0.4191603088378907
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,8,32,power_law_1.01,0.10841343879699708
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,8,32,power_law_1.01,0.6055910491943359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,8,32,power_law_1.01,0.8141311645507813
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,8,32,power_law_1.01,1.1949440002441407
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,8,32,power_law_1.01,0.03595391988754272
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,8,32,power_law_1.01,1.435116729736328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,8,32,power_law_1.01,0.036610560417175295
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,8,32,power_law_1.01,0.03913088083267212
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,8,32,power_law_1.01,2.410147857666016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,8,32,power_law_1.01,0.04168704032897949
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,4,64,power_law_1.2,0.14805248260498047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,8,32,power_law_1.01,0.05036416053771973
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,8,32,power_law_1.01,0.05808896064758301
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,4,64,power_law_1.2,0.19110015869140623
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,8,32,power_law_1.01,0.037920000553131106
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,8,32,power_law_1.01,0.05874815940856933
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,8,32,power_law_1.01,0.058311681747436526
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,8,32,power_law_1.01,0.05741824150085449
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,2688,22,512,4,64,power_law_1.2,0.2573132705688477
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,8,32,power_law_1.01,0.06242432117462158
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,8,32,power_law_1.01,0.07267327785491944
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,8,32,power_law_1.01,0.08123647689819335
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,8,32,power_law_1.01,2.9779583740234377
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,8,32,power_law_1.01,0.08519807815551758
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,8,32,power_law_1.01,0.08204928398132325
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,8,32,power_law_1.01,0.09799551963806152
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,8,32,power_law_1.01,0.1237990379333496
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,8,32,power_law_1.01,0.15591423988342284
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,8,32,power_law_1.01,0.22239616394042966
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,8,32,power_law_1.01,0.25788288116455077
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,8,32,power_law_1.01,0.39785854339599613
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,8,32,power_law_1.01,0.47873920440673834
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,8,32,power_law_1.2,0.05531007766723632
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,8,32,power_law_1.01,0.06821887969970704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,8,32,power_law_1.01,0.7650240325927735
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,8,32,power_law_1.01,1.0052301025390624
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,8,32,power_law_1.2,0.05928703784942627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,8,32,power_law_1.2,0.06283775806427003
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,8,32,power_law_1.2,0.0939788818359375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,8,32,power_law_1.2,0.10315263748168946
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,8,32,power_law_1.2,0.10461824417114259
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,8,32,power_law_1.01,1.5537855529785156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,8,32,power_law_1.2,0.09996416091918944
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,8,32,power_law_1.2,0.10350463867187501
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,8,32,power_law_1.2,0.11388416290283203
SGLang,0.5.10,NVIDIA B200,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,2688,22,512,4,64,power_law_1.2,0.3827046585083008
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,8,32,power_law_1.01,1.7448524475097655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,8,32,power_law_1.2,0.1188633632659912
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,8,32,power_law_1.2,0.1294374370574951
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,8,32,power_law_1.2,0.14729984283447267
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,8,32,power_law_1.2,0.14747520446777346
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,8,32,power_law_1.2,0.05663616180419921
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,8,32,power_law_1.2,0.1526848030090332
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,8,32,power_law_1.2,0.17787647247314453
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,8,32,power_law_1.2,0.22487936019897461
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,8,32,power_law_1.2,0.26784639358520507
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,8,32,power_law_1.2,0.05690368175506592
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,8,32,power_law_1.2,0.3783180618286133
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,8,32,power_law_1.2,0.46162689208984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,8,32,power_law_1.2,0.035870718955993655
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,8,32,power_law_1.2,0.7033100891113281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,8,32,power_law_1.2,0.03665791988372803
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,8,32,power_law_1.2,0.03830399990081787
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,8,32,power_law_1.2,0.9092070770263672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,8,32,power_law_1.2,0.03913216114044189
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,8,32,power_law_1.2,0.04173439979553223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,8,32,power_law_1.2,0.05111551761627198
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,8,32,power_law_1.2,0.05607168197631836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,8,32,power_law_1.2,0.057529602050781246
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,8,32,power_law_1.2,1.3655130004882812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,8,32,power_law_1.2,0.05887487888336181
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,8,32,power_law_1.2,0.06507520198822021
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,8,32,power_law_1.2,0.06868864059448242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,8,32,power_law_1.2,0.07556863784790038
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,8,32,power_law_1.2,0.0840998363494873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,8,32,power_law_1.2,0.08958463668823242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,8,32,power_law_1.2,0.08563072204589843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,8,32,power_law_1.2,0.10423551559448244
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,8,32,power_law_1.2,0.13577471733093263
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,8,32,power_law_1.2,0.16831615447998047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,8,32,power_law_1.2,0.233251838684082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,8,32,power_law_1.2,0.31363712310791014
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,8,32,power_law_1.2,2.7204736328125003
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,8,32,power_law_1.2,0.46542209625244146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,8,32,power_law_1.2,0.5761459350585938
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,8,32,power_law_1.2,1.8223948669433594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,16,16,balanced,0.027215359210968015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,16,16,balanced,0.03214080095291137
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,8,32,power_law_1.2,0.8399705505371093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,16,16,balanced,0.03707648038864136
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,8,32,power_law_1.2,0.05844351768493652
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,8,32,power_law_1.2,3.5237799072265625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,16,16,balanced,0.08642304420471192
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,16,16,balanced,0.10805760383605958
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,16,16,balanced,0.13272704124450682
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,16,16,balanced,0.1439782428741455
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,16,16,balanced,0.14680831909179687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,16,16,balanced,0.15670271873474123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,8,32,power_law_1.2,1.7481932067871093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,16,16,balanced,0.16815103530883788
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,16,16,balanced,0.17428863525390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,16,16,balanced,0.1853811264038086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,16,16,balanced,0.19566335678100585
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,16,16,balanced,0.054873600006103515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,8,32,power_law_1.2,1.0993971252441406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,16,16,balanced,0.21997568130493167
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,16,16,balanced,0.2711039924621582
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,16,16,balanced,0.13696255683898925
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,16,16,balanced,0.3187046432495117
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,16,16,balanced,0.46023681640624997
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,16,16,balanced,0.5571366500854492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,16,16,balanced,0.7867072296142579
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,16,16,balanced,0.9812569427490235
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,16,16,balanced,0.19819007873535158
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,16,16,balanced,0.06819583892822265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,16,16,balanced,0.09666175842285156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,16,16,balanced,0.17182464599609376
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,16,16,balanced,1.4515621948242188
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,16,16,balanced,0.24558591842651367
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,16,16,balanced,0.3259929656982422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,16,16,balanced,1.8953753662109374
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,16,16,balanced,0.3987161636352539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,16,16,balanced,0.3615871810913086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,16,16,balanced,0.37826560974121093
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,16,16,balanced,0.38443008422851566
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,16,16,balanced,0.40841087341308596
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,16,16,balanced,0.4218956756591797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,16,16,balanced,0.39994239807128906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,16,16,balanced,0.4346688079833984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,16,16,balanced,0.4277030563354492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,16,16,balanced,0.4333171081542969
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,16,16,balanced,2.798917236328125
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,16,16,balanced,0.43347969055175783
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,16,16,balanced,0.4704064178466797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,16,16,balanced,0.536069107055664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,16,16,balanced,0.601154556274414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,16,16,balanced,0.6689753723144531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,16,16,balanced,3.6469671630859373
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,16,16,balanced,1.0543641662597656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,16,16,power_law_1.01,0.06539008140563965
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,16,16,balanced,1.214871063232422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,16,16,power_law_1.01,0.08487039566040039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,16,16,power_law_1.01,0.08634112358093263
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,8,32,power_law_1.2,2.241222381591797
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,16,16,balanced,1.9010342407226564
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,16,16,power_law_1.01,0.09552384376525878
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,16,16,power_law_1.01,0.08926719665527344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,16,16,power_law_1.01,0.14024191856384277
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,16,16,power_law_1.01,0.13149696350097656
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,16,16,balanced,2.180564422607422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,16,16,power_law_1.01,0.13691136360168457
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,16,16,power_law_1.01,0.14577024459838867
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,16,16,power_law_1.01,0.15622783660888673
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,16,16,power_law_1.01,0.1646758460998535
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,16,16,power_law_1.01,0.18272512435913085
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,16,16,power_law_1.01,0.18083967208862303
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,16,16,power_law_1.01,0.19232383728027344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,16,16,power_law_1.01,0.2163315200805664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,16,16,power_law_1.01,0.2813324737548828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,16,16,balanced,3.2640640258789064
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,16,16,power_law_1.01,0.3302988815307617
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,16,16,power_law_1.01,0.4509491348266602
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,16,16,power_law_1.01,0.5495347213745118
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,16,16,power_law_1.01,0.09956352233886719
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,16,16,power_law_1.01,0.7805427551269531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,16,16,balanced,4.273061218261718
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,16,16,power_law_1.01,0.9936768341064454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,16,16,power_law_1.01,0.13282048225402832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,16,16,power_law_1.01,1.5545408630371094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,16,16,power_law_1.01,0.3449369430541992
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,16,16,power_law_1.01,1.8886399841308594
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,16,16,power_law_1.01,0.33291904449462895
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,16,16,power_law_1.01,0.26471040725708006
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,16,16,power_law_1.01,0.28542720794677734
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,16,16,power_law_1.01,0.3282547378540039
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,16,16,power_law_1.01,0.3539763259887695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,16,16,power_law_1.01,2.7466021728515626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,16,16,power_law_1.01,0.36449024200439456
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,16,16,power_law_1.01,0.3353919982910156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,16,16,power_law_1.01,0.3570355224609375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,16,16,power_law_1.01,0.3501990509033203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,16,16,power_law_1.01,0.4040627288818359
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,16,16,power_law_1.01,0.3799257659912109
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,16,16,power_law_1.01,0.3793753433227539
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,16,16,power_law_1.01,3.6736691284179686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,16,16,power_law_1.01,0.4215564727783203
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,16,16,power_law_1.01,0.41243518829345704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,16,16,power_law_1.01,0.485847053527832
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,16,16,power_law_1.01,0.5604800033569336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,16,16,power_law_1.01,0.7693977355957031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,16,16,power_law_1.01,0.8164224243164062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,16,16,power_law_1.01,1.1467021179199217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,16,16,power_law_1.01,1.3988006591796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,16,16,power_law_1.2,0.06247039794921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,16,16,power_law_1.2,0.08294912338256835
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,16,16,power_law_1.01,2.3071859741210936
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,16,16,power_law_1.2,0.08444671630859375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,16,16,power_law_1.01,2.355951385498047
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,16,16,power_law_1.2,0.09428607940673828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,16,16,power_law_1.2,0.10291328430175781
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,16,16,power_law_1.2,0.14011391639709472
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,16,16,power_law_1.2,0.12865535736083983
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,16,16,power_law_1.2,0.12960384368896485
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,16,16,power_law_1.2,0.13707776069641114
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,16,16,power_law_1.2,0.15018624305725098
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,16,16,power_law_1.2,0.15401984214782716
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,16,16,power_law_1.2,0.08959487915039063
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,16,16,power_law_1.2,0.1690559959411621
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,16,16,power_law_1.2,0.18792831420898437
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,16,16,power_law_1.2,0.18470272064208984
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,16,16,power_law_1.2,0.194703369140625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,16,16,power_law_1.01,4.026056823730469
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,16,16,power_law_1.2,0.22026880264282225
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,16,16,power_law_1.2,0.28311552047729494
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,16,16,power_law_1.2,0.32919551849365236
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,16,16,power_law_1.2,0.4653043365478515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,16,16,power_law_1.2,0.565802230834961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,16,16,power_law_1.01,4.596949768066406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,16,16,power_law_1.2,0.8052057647705079
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,16,16,power_law_1.2,1.0991539001464843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,16,16,power_law_1.2,0.20627199172973634
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,16,16,power_law_1.2,1.6178495788574216
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,16,16,power_law_1.2,0.23992191314697267
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,16,16,power_law_1.2,0.3476057434082031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,16,16,power_law_1.2,2.112857666015625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,16,16,power_law_1.2,0.2658726310729981
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,16,16,power_law_1.2,0.3516672134399414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,16,16,power_law_1.2,0.2757785606384277
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,16,16,power_law_1.2,0.3473420715332031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,16,16,power_law_1.2,3.1162008666992187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,16,16,power_law_1.2,0.3281280136108399
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,16,16,power_law_1.2,0.3421427154541016
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,16,16,power_law_1.2,0.35879425048828123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,16,16,power_law_1.2,0.35074302673339847
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,16,16,power_law_1.2,0.3724800109863281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,16,16,power_law_1.2,0.3754048156738281
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,16,16,power_law_1.2,0.4170035171508789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,16,16,power_law_1.2,0.38524417877197265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,16,16,power_law_1.2,0.39211776733398435
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,16,16,power_law_1.2,0.44265598297119146
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,16,16,power_law_1.2,0.5064166259765626
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,16,16,power_law_1.2,0.5557939147949218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,16,16,power_law_1.2,0.7353395080566406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,16,16,power_law_1.2,1.0511244964599609
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,16,16,power_law_1.2,1.248590087890625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,16,16,power_law_1.2,1.4794355773925782
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,16,16,power_law_1.2,4.128416137695313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,16,16,power_law_1.2,2.2554150390625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,32,8,balanced,0.03414144039154053
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,32,8,balanced,0.042004480361938476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,32,8,balanced,0.05550079822540284
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,32,8,balanced,0.0932863998413086
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,16,16,balanced,0.06945024013519288
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,32,8,balanced,0.12061440467834472
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,16,16,power_law_1.2,3.105383605957031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,32,8,balanced,0.13239551544189454
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,32,8,balanced,0.15653632164001466
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,32,8,balanced,0.028401920795440672
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,32,8,balanced,0.1620966339111328
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,32,8,balanced,0.16936063766479492
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,32,8,balanced,0.17796991348266603
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,32,8,balanced,0.18779647827148438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,32,8,balanced,0.20546304702758791
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,32,8,balanced,0.1970649528503418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,32,8,balanced,0.20738687515258789
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,32,8,balanced,0.23084800720214843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,32,8,balanced,0.28246912002563473
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,32,8,balanced,0.32861183166503904
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,32,8,balanced,0.46932350158691405
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,32,8,balanced,0.12684800148010253
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,16,16,power_law_1.2,4.491573791503906
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,32,8,balanced,0.5670131301879883
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,32,8,balanced,0.8425843048095704
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,32,8,balanced,1.0333209228515625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,32,8,balanced,0.0543782377243042
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,32,8,balanced,0.06835584163665771
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,32,8,balanced,1.9682304382324218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,32,8,balanced,0.15558015823364257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,32,8,balanced,0.24950016021728513
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,16,16,power_law_1.01,0.24641151428222657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,32,8,balanced,2.897469482421875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,32,8,balanced,0.32516864776611326
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,32,8,balanced,0.4398668670654297
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,32,8,balanced,0.3360319900512695
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,16,16,power_law_1.01,0.25979904174804686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,32,8,balanced,0.5198720169067383
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,32,8,balanced,1.5409011840820312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,32,8,balanced,3.7815206909179686
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,32,8,balanced,0.4843008041381836
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,32,8,balanced,0.5244441604614257
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,32,8,balanced,0.5008806228637696
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,32,8,balanced,0.5699763107299805
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,32,8,balanced,0.5458240127563476
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,32,8,balanced,0.5878732681274415
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,32,8,balanced,0.5349017715454101
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,32,8,balanced,0.5956607818603515
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,32,8,balanced,0.645274887084961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,32,8,balanced,0.7974540710449218
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,32,8,balanced,0.7621209716796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,16,16,power_law_1.2,5.275082397460937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,32,8,balanced,1.4214041137695312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,32,8,balanced,0.13144831657409667
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,32,8,power_law_1.01,0.045377278327941896
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,32,8,balanced,1.6802584838867187
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,32,8,power_law_1.01,0.08096511840820313
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,32,8,power_law_1.01,0.0758784008026123
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,32,8,power_law_1.01,0.08577152252197265
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,32,8,power_law_1.01,0.10050687789916993
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,32,8,power_law_1.01,0.11323136329650878
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,32,8,balanced,2.3029644775390627
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,32,8,power_law_1.01,0.12614784240722657
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,32,8,power_law_1.01,0.151978235244751
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,32,8,power_law_1.01,0.15252863883972168
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,32,8,power_law_1.01,0.16151424407958986
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,32,8,power_law_1.01,0.1740083122253418
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,32,8,power_law_1.01,0.17959423065185548
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,32,8,balanced,2.842920837402344
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,32,8,power_law_1.01,0.20081151962280272
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,32,8,power_law_1.01,0.19371519088745118
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,32,8,power_law_1.01,0.2078016090393066
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,32,8,power_law_1.01,0.2941555213928223
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,32,8,power_law_1.01,0.346874885559082
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,32,8,power_law_1.01,0.45838977813720705
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,32,8,power_law_1.01,0.1193075180053711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,32,8,balanced,3.994027404785156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,32,8,power_law_1.01,0.6289292907714843
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,32,8,power_law_1.01,0.8344102478027343
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,32,8,balanced,4.33322509765625
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,32,8,power_law_1.01,1.1144921875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,32,8,power_law_1.01,0.1820800018310547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,32,8,power_law_1.01,1.5853094482421874
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,32,8,power_law_1.01,0.24176000595092773
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,32,8,power_law_1.01,0.29097215652465824
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,32,8,power_law_1.01,0.34249343872070315
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,32,8,power_law_1.01,2.1032780456542968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,32,8,power_law_1.01,0.3214924621582031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,32,8,power_law_1.01,3.036423645019531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,32,8,power_law_1.01,0.4472012710571289
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,32,8,balanced,0.5873932647705078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,32,8,power_law_1.01,0.44935039520263675
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,32,8,power_law_1.01,0.5061222457885742
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,32,8,power_law_1.01,0.4697612762451172
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,32,8,power_law_1.01,4.084958801269531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,32,8,power_law_1.01,0.5358041763305664
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,32,8,power_law_1.01,0.5108339309692382
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,32,8,power_law_1.01,0.2329510307312012
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,32,8,power_law_1.01,0.5362931060791015
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,32,8,power_law_1.01,0.6259404754638671
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,32,8,power_law_1.01,0.9949798583984375
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,32,8,power_law_1.01,0.47949951171875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,32,8,power_law_1.01,0.28061695098876954
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,32,8,power_law_1.01,0.9996902465820312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,32,8,power_law_1.01,0.37952384948730467
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,32,8,power_law_1.01,0.34649726867675784
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,32,8,power_law_1.01,1.5498278808593748
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,32,8,power_law_1.01,1.571029815673828
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,32,8,power_law_1.2,0.04489471912384033
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,32,8,power_law_1.2,0.07029759883880615
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,32,8,power_law_1.2,0.07680255889892577
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,32,8,power_law_1.2,0.09029631614685059
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,32,8,power_law_1.2,0.10528639793395995
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,32,8,power_law_1.2,0.10814592361450195
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,32,8,power_law_1.2,0.12018431663513182
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,32,8,power_law_1.2,0.12463999748229981
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,32,8,power_law_1.2,0.1497152042388916
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,32,8,power_law_1.2,0.15453056335449217
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,32,8,power_law_1.2,0.16277503967285156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,32,8,power_law_1.2,0.17421184539794923
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,32,8,power_law_1.01,3.3187762451171876
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,32,8,power_law_1.2,0.18170495986938479
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,32,8,power_law_1.2,0.20442495346069336
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,32,8,power_law_1.2,0.1994367980957031
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,32,8,power_law_1.2,0.21212415695190429
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,32,8,power_law_1.2,0.23968896865844727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,32,8,power_law_1.2,0.29436927795410156
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,32,8,power_law_1.2,0.3446988677978516
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,32,8,power_law_1.2,0.48104705810546877
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,32,8,power_law_1.2,0.6362009429931641
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,32,8,power_law_1.01,5.006785278320312
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,32,8,power_law_1.2,0.8672038269042968
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,32,8,power_law_1.2,1.1475929260253905
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,32,8,power_law_1.2,1.6385804748535158
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,32,8,power_law_1.2,2.155740203857422
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,32,8,power_law_1.2,0.15737215995788573
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,32,8,power_law_1.01,0.827267837524414
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,32,8,power_law_1.2,0.20407039642333985
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,32,8,power_law_1.2,3.1793637084960937
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,32,8,power_law_1.2,0.3419059371948242
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,32,8,power_law_1.2,0.3935897445678711
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,32,8,power_law_1.2,0.28336896896362307
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,32,8,power_law_1.2,0.3716236877441406
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,32,8,power_law_1.2,0.40162303924560544
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,32,8,power_law_1.2,0.40897663116455074
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,32,8,power_law_1.2,0.535235824584961
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,32,8,power_law_1.2,0.42044544219970703
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,32,8,power_law_1.2,0.5258303833007812
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,32,8,power_law_1.2,4.207257690429687
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,32,8,power_law_1.01,2.1045733642578126
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,32,8,power_law_1.2,0.4906201553344727
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,32,8,power_law_1.2,0.5795750427246094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,32,8,power_law_1.2,0.5819123077392578
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,32,8,power_law_1.2,0.5124057769775391
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,32,8,power_law_1.2,0.5429452896118164
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,32,8,power_law_1.2,0.6512947082519531
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,32,8,power_law_1.2,0.8128652954101563
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,32,8,power_law_1.2,0.9945075225830078
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,32,8,power_law_1.2,0.9550975799560547
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,32,8,power_law_1.2,1.5496678161621094
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,32,8,power_law_1.2,1.966796875
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,32,8,power_law_1.2,2.9351141357421873
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,32,8,power_law_1.01,5.749761352539062
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,32,8,power_law_1.2,3.9705624389648433
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,32,8,power_law_1.2,3.990982360839844
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,32,8,power_law_1.2,6.685761108398438
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,32,8,power_law_1.01,0.5898380661010743
SGLang,0.5.10,NVIDIA B200,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,32,8,power_law_1.2,0.21701375961303712
